CiderHouse (SIHDR)

Ciderhouse is going to southern France.

2010-03-16T21:01:00.000-07:00

Ciderhouse is unavailable for a month since S and I are going to Barcelona and then into southern France (Arles, Nimes, Nice, Aix de Provence, etc.). We leave tomorrow and will be back on April 19. I'd love to see fans of Squinchpix or Ciderhouse on the way and we'll be twittering in the hope of making contact. Best to all of you until we come back and, in the meantime, you might like to look at some blog entries that I think might be helpful to you.
Real HDR vs. Pseudo HDR
FDRTools and the Compressor Algorithm
Dynamic Photo HDR and Photomatix: Color Compare
Why Photomatix is a poor choice
Dynamic Photo HDR and the Halo-Matix Algorithm
Doing HDR in pieces
A Simple HDR Workshop
Single Pixel shift in HDR alignment. What does it mean?
Noise in Single Image HDR. A Workshop.
Manipulating L-channel in Lab with HDR

Until we get back ...

Lars Roglin writes

2010-03-02T16:56:00.000-08:00

One of my readers, Mr. Lars Roglin, of (I think) the Netherlands (see his gallery here) writes the following in response to one of my old posts:

Hi Bob,

interesting calculations, but I think you have one major flaw in your assumptions. You state: "Let’s say that the time to make all these exposures is 10 seconds."

With modern DSLRs this might be the case for long exposures and night shots but e.g. my EOS 40D is able to shoot at ~6 fps (the EOS 1D Mark IV even at ~12 fps) at full speed. So your 10 seconds change to .5 seconds for a series of -2EV,0EV,+2EV (0.25 or 1/4s respectively)!

In addition, I would assume that for most typical landscape shots clouds might be further away than a mile, so proper alignment should avoid stuttering.

If you want to push it to the limit, you could perhaps limit yourself to 2 exposures, which would push it to 1/8s which should even allow portrait shots (although perhaps with a higher reject rate - but shutter speeds of 1/8s and even 1/4s is not unusual in available light photography!)

Curious what you think about that.

Cheers,
Lars

My answer is as follows:

Hi Lars,

Whatever you're doing your gallery shows that it's successful. I did use a large number for the time it would take to make multiple exposures. Great cameras like yours allow for much less time to accomplish the same number of exposures so the problems I outlined would be minimized for you.

Two points:

1. I shouldn't allow myself to be drawn into the 'average cloud thing' but I can't resist. So, in the spirit of fun, let's say that the average cloud height is about 2000 feet. And let's say that the angle to the nearest cloud in the landscape photo that you're taking is 45 degrees. Simple trig tells us that the hypotenuse of the resulting triangle (the hypotenuse is the direct distance from the camera to the nearest cloud as defined above) is 2828.4 feet. If that cloud is moving faster than about 1.92 miles per hour (in parallel with the sensor plane) then that's a displacement of 1 pixel in half a second. (I'll leave the math to you, Lars). It's not unusual for a cloud to travel about 20 mph and those conditions would translate into a 10 pixel offset even in half a second. These numbers are full of assumptions and I'm not trying to prove that you will necessarily have a problem with your equipment. But the fact is that we see a lot of cloud displacement (and foliage displacement) in multiple image HDR. It must be caused by something (perhaps it's not coming from Canons?) and the math (even highly speculative math like this) suggests what it is. I'm just not personally optimistic about such a technique but that means nothing if you can make it work.

2. I want to approach this from a different angle. Let's say that you wrote me a letter and said "Bob, I invented this great new way to shoot photographs! Instead of using f8 at 250th of a second for normal lighting, I've pioneered what I call the 'slow shutter'(tm) technique. The benefits of a small f-stop are so obvious to me that I always shoot at f78 at one/half second even, I repeat, in normal lighting conditions." And then you would carry on, perhaps, about the incredible depth-of-field and that everyone should always shoot at speeds of 0.5 seconds or even longer. Well, if I received (or if you received) such a letter we would suspect that our correspondent was cracked. But that's what you seem to actually be advocating.

However, I know that the f64 school of American photographers (Weston and Adams and the like) actually did shoot this way so maybe I need to go back and review what they actually did and what problems they encountered. Perhaps you would be interested enough to look into it and write back about what you find? Thanks for your letter.

Best,

Bob

Real HDR vs. Pseudo HDR

2010-03-01T17:53:00.000-08:00

There's a lot of discussion on the web about the difference between 'Real' HDR and something called 'Pseudo' HDR. You know how it goes; 'Real' HDR is made with more than one image and 'Pseudo' HDR is made with only one image. To all of which I say 'Balls!' If you use HDR tone-mapping algorithms to get everything you can out of a single image then it's as real as HDR can be. In fact, not a lot rides on the outcome of this discussion (it's a variation of the Essentialist fallacy) but I had an experience recently that made me think again about what 'pseudo' HDR might really mean.

I was on Crete last year and, at the palace of Knossos, I took about 600 shots and all of them were more or less defective. It is the Mediterranean after all and so all the shots were either over-exposed or under-exposed but usually both. And yet, while looking at them, I felt that all the detail was really there although pushed to the edges. If only I could find a way to retrieve the detail on both ends of the brightness range. I process my pictures in stages; the first stage employs Lightroom 2.6. So the problem was to find a method of working in Lightroom that would take the rough edges off. Here, for example, is a shot of a wall and a buttress in Knossos that will illustrate a typical starting point.

In this shot the darks are not too bad but what to do about the highlights?

The first thing I did was set the 'Recover' slider to 100. Here's the result:

'Recover' moves brightnesses on the extreme bright end back towards the middle. In other words the effect is to manipulate one end of the histogram without modifying the other .. or at least not very much. By doing this we can see that there is some detail in the brightest stones on the right. They're not completely burnt out. The next step is to turn the brightnesses all the way to the bottom. Here we go:

By doing this we've regained all the detail in the highlights that we're likely to get. In the next step we turn up the fill slider as much as possible consistent with the current state of the brights. Here it is:

Now we have a pretty dim image but all the brightnesses are in agreement with each other. The shadows are suitably dark with respect to the brights. There's full detail in the darks AND full detail in the lights. We're light-years away from where we started. Now we're going to turn up the brightness gradually until the brightest areas of the photo are just under where they should be. Here we go:

Now the steps that I've outlined typically degrade the contrast. The fill slider and the contrast slider are really opposites. The contrast has to be fixed particularly in the lights. I do that in parts. I begin by moving the 'Black' slider up to about 15 (way too high for an ordinary photo). Here's the result:

Before we fix the rest of the Contrast problem I'm going to correct the color cast. If I don't fix the blue cast in the shadows now then changing the contrast will only worsen it. In Lightroom I just decided to use the 'Saturation' slider and I pulled the cursor downward in the shadows on the tops of the rocks on the left. Here's the result:

I present the whole Lightroom screen so you can see what the effect was of the Saturation change. The shadows are now monochromatic (pretty much) and upping the Contrast in the next step shouldn't worsen that. Now it's time to fix the contrast. By moving the Contrast or the Clarity sliders (or both) to the right we push the brights up and the shadows down so the trick in doing this is to increase the contrast without finally burning out the lights and pushing the shadows down so that you're back where you started from.

Here's my final version.

Now this photograph will clearly never interest very many people. Even now it's not really successful but it could be of some small interest to an archaeologist whose specialty happens to be this particular wall. It could use a lot more work. The local operator brush could probably be employed to good effect in the highlights. But what really interested me after doing this more than 550 times was how much like HDR the results of this kind of processing actually look. The final example looks like HDR (at least the shadowed areas do) and next I present another example of before and after shots in which the kind of processing outlined above gave the result an 'HDR' look.

Here's a 'before' shot of Knossos' famous colonnade as it came from the camera (complete with sensor dirt).

Here's the 'after' shot in which I used an aggressive 'fill' as described above.

And also this example. Here's the before picture.

The same picture after processing as I described. If I told you that this 'after' picture was HDR you'd probably have believed me. It just looks like HDR.

The following photo actually was processed with HDR software (FDRTools). The non-HDR version actually looks more like HDR (or our common expectation of it) than the one that actually is HDR.

HDR is always described as having a 'unique look'. But what is that look? Lots of people suppose that it has something to do with aggressive saturation of the image. Sometimes we're told that it makes photos look like paintings although the people saying this have no idea what makes a painting look like a painting. So let's answer that question first. What does make a painting look like a painting? Simple. A reduced tonal range. For technical reasons most painters never, and have never, employed pure blacks or whites. So that giving up those extreme tones right away reduces the tonal range. Also painters do not paint to obscure. They will never (or rarely) paint shadows as dark as they are in real life. In fact one of the chief goals of the painter is not only to paint clearly in the shadows but to paint full color in the shadows. Therefore any picture that appears to have a reduced tonal range will look painterly. (As in everything there are many exceptions. There are certain chiaroscuro painters who DO employ pure blacks and whites and virtually nothing in between. Such paintings gain in drama but relatively few painters actually paint this way.)

In photography HDR algorithms produce a reduced tonal range; in many cases a shorter range than can actually be handled by the output media. The resulting images feature weakened shadows; shadows no darker than if one had simply poured water on the ground. Doing this darkens the ground but does not obscure it and many shadows in HDR look this way. Also many HDR shots seem to show us a perpetual dawn.

So the moral of the story is that if you want 'pseudo' HDR pump up the fill slider.

FDRTools and the Compressor Algorithm: An approach

2010-02-02T21:49:00.000-08:00

In one of the recent posts I promised to explore FDRTools' Compressor Algorithm. The 'Compressor' is, according to FDRTools, a local toning operator. That means that it covers the surface point by point and it tries to adjust each pixel according to the brightnesses of neighboring pixels within a certain radius. The FDRT manual says this "The Compressor algorithm works as follows: the intensity of each pixel is regulated individually and depends on the intensities of neighbouring pixels. If neighbouring pixels differ strongly in their respective intensity, e.g. in the surroundings of a strong light source, then these differences are strongly attenuated. Where the differences are small, e.g. in shadowy areas, there is little or no attenuation, evtl. the difference is even increased. As a result the dynamic range of the HDR image is strongly reduced while local tonal value differences are preserved. Local contrast and thus the perceptibility of details are considerably improved compared to global tone mapping methods like Simplex or Receptor."

The authors of this manual are, I think, Germans and I'd like to rephrase this as follows: "Compressor examines each individual pixel and changes it depending on what it finds in neighboring pixels (within a certain radius). If the immediate neighborhood has a big dynamic range then it is reduced. If the neighborhood has a relatively narrow dynamic range then it is increased. The result is that the dynamic range of the entire image is significantly reduced but in such a way that local relative differences are preserved and local contrast is even improved relative to Simplex or Receptor."

Good enough. But is it true and how do we use it? Let's concentrate on a small number of details just at first. We start with the gross image of the column drum that I used last time. Here it is:

Yow.

What we see with defaults is this:

Now this is another low-contrast image. All the brightnesses are squeezed into the center of the histogram as we see. We know that we need to draw in the left and right sides (black point and white point) to more nearly enclose the actual values. The result is this:

Now it's no longer grossly low-contrast. In fact the gravel looks pretty good. But the brights have lost all the definition even of the default settings and somehow that needs to be restored. Another way of saying this is that the brights are mapped into too narrow a range and to spread out that range you steepen the curve on the bright/right side, like this:

You may have to experiment but, at least in this case, the right side (RHS) of the curve had to be steepened so that there would be more output grays in a very small horizontal brightness distance. That's a very clumsy way of putting it but it's the best way that I can think to explain it. Let's pretend that the brights are o.k. and examine the darks. In this particular case I really couldn't find any better setting for the darks (the gravel) and so I left it alone. There are several more important controls and let's look at the Compressor slider. This slider, according to the tool tips, regulates the strength of tonal length compression. I confess that I don't know what that means in practical terms so we're going to have to approach that empirically. When I moved it all the way to the left (0.0) the histogram shrunk symmetrically around its center. When I moved it all the way to 10.0 the histogram expanded around the center. The effect was, at 0.0, to lower contrast, at 10.0 to heighten contrast. The tool tips say that you should use values from 0.0 to 5.0 for a single raw (which this is) and 5.0 to 10.0 for multiple images. I found that this all looked best when I moved it up to 8.0 or even 10.0. Here it is:

HDR gets details out of burned-out or from under-exposed areas. Once you get that it's always in low contrast. The whole problem of HDR, as I'm sure you're aware, is to increase the contrast (that is, to increase the range of grays) on the result of HDR processing without going all the way back to the original picture. That was our challenge with the simplex method. But FDRTools Compressor seems to do the trick. What do you think? Let me know in the comments section if I've grossly mishandled this or if you have better ideas. I'm going to finish up by processing this result and then post-processing in Photoshop. I gave it a lick with Curves to increase the contrast, sharpened the L layer, and here's the final result:

It'll never be a great picture but it's a perfectly good illustration of a column drum. If I say so myself there's nothing at all wrong now. It's a clear and adequate picture with a full and reasonably distributed range of grays. Perfect for squinchpix.com. The sunlit wedge in the right foreground is a potential source of trouble (just after the top of the drum itself) but here it gave no problem.

Also the purple grid marks are what FDRTools puts over pictures when you don't have a license.

So:
1. bring in the curve ends so that they enclose the histogram.
2. Evaluate the gray range in the brights. If it's too narrow (which it was) then increase the range there by steepening the right end of the curve. Steeper means more grays. Shallower means fewer grays.
3. Evaluate the gray range in the darks and increase the gray range if you need to by making the left end of the curve steeper.
4. Experiment with an optimal setting of the 'Compressor' slider. By this stage you should have an reasonably distributed tonal range of grays. That's your real goal for HDR processing. If you get to that point you're successful.
5. Finish up in Photoshop to get complete control over the result in the way you're used to. You can adjust contrast, correct color casts, etc.

I'm mindful that color is a distraction and what I'd like to do next is to use FDRTools on just the L-layer of a Lab version of this picture. I've done things like this in the past and it's time to do that again. In the next day or two.

And there's something that I now realize that I assumed throughout but never actually stated:
With HDR Always think in terms of brightness - Never think in terms of color. HDR is not an idea about color. It's an idea about brightness. Always, always, always!!!

Till next time...

Bob

HDRCreme shot of the day: A Review

2010-01-30T20:27:00.000-08:00

When listing the areas in which multiple-image HDR has a competency I did leave one out: the still life. Nowhere is that borne out better than in the image which graces HDRCreme's front page. See it here. It is a photo of an old spinning wheel placed against a background of colored wool swatches and hand-embroidered fabrics. The photographer hasn't given us any details about it and I'm assuming that it originated with multiple images. It is very low-key and even-contrast. When I downloaded the largest version and ran it through Photoshop it produced this histogram.

The spike on the right is the over-exposed swatch of wool. Take a good look at this histogram. This smooth and even shape, just barely bell-shaped, is the mark of many HDR photos that I've seen. In fact when we see a histogram as smooth as this one we can be pretty sure that the photo which produced it was created with HDR. I wish I could pin this specific smooth histogram to a specific HDR tone-mapper but I can't yet. Perhaps my readers know more than I do about this? I may devote entries in the future to relating histograms to specific tone mappers.
The creator of this photo identifies him (her?) self as 'Akun2500' and there's a small gallery of his work on HDRCreme. You can ignore the motorcycles which are beginning efforts but the others are splendid. Sighisoara is a popular tourist destination in Romania which is where 'Akun2500' lives. Clearly people go there (among other things it is the home of Vlad the Impaler - the original Dracula) to look at the antiques. The delicacy of this rendering, the absence of any visible HDR artifacts, the over-all violet tonality lend it an inexpressible charm and it is a strong advertisement for multiple-image HDR (if that's in fact what it is) - at least under controlled conditions. Vermeer would have been proud to do this one.
Over and over I've tried to make the point that HDR does away with the subject and creates a fully articulated surface over which the eye may wander with pleasure. This picture is what I mean. The spinning wheel is deemphasized by the calm and even light, a light which (I'll bet) is purely imaginary - the product of HDR alone; it (the wheel) fades into the background. The eye takes in the sublime needlework,the old wood, the textures on the walls and the wool. It is a decorated surface and nothing more. This picture restores pleasure to the act of looking at the surface of a photograph. We learn nothing - we enjoy everything.
I won't give it a '12 out of 10' but I can hardly see how it could be improved.

Bob

FDR Tools. The Simplex Algorithm

2010-01-30T09:28:00.000-08:00

I've been intrigued by the performance of FDRTools and I put up a small post two days ago to show how it handled the color sheet. Now I want to consider the Simplex algorithm which, according to FDR, is a simple tone-mapping operator. Let's start with a black and white test sheet and see how it performs. Straight out of the box with default settings it produces this:

This has good separation in the blacks and not so good separation in the whites. No problems with the lines or their intersections and - until it reaches the brights - realistic rendition of the gradient with some banding.

I looked for a simple manipulation of the contrast curve that would improve things and so I steepened the slope on the left side hoping to increase separation among blacks. That worked pretty well so I did the same thing on the light side. That resulted, of course, in the classic low-contrast reverse S-curve and, well, see what it did:

This simple change produced low contrast (see previous) but a huge improvement overall in terms of rendition and separation. Now that I felt that I was on the right track I tried to optimize it:

I brought in the outer bounds so that they more tightly bounded the observable values on the histogram and I steepened the sides of the curves in the darkest and brightest areas. I also brought the middle of the curve way down so that it produced the darker band in the center of the gradient - in so doing it helped to intensify the separation in the lights (you probably couldn't do this with a real photo). I think the result looks really good but, at least, by knowing that we're searching for some reverse S-curve to improve the scene we have a clue as to how to proceed.

Now keep this reverse S-curve in mind because the next thing I wanted to do was to see if I could apply just this one thing to a real-life scene. I chose the following awful picture that I took in Argos, in Greece.

Here's a column-drum lying on the ground. The lights are on the edge of being burnt-out but everything else, almost, is in inky blackness. I'm going to use FDR tools on this and just take the default settings first. What does Simplex do to this without my doing anything?

Lots of detail here and overall very low contrast. Here's something to work with. Let's apply the reverse S-curve and see where we are but, it's clear, we have to move in the upper and lower curve bounds on the histogram itself. Here are the settings that I found and the resulting image.

I found that it's not just a matter of applying any old reverse S curve. You have to experiment with the shape on each end until it looks as good as possible. I moved in the upper and lower bounds (do that first) on the curve until they just touched the values on the highest and lowest end and that makes a crucial difference. So the thought for today is, when using FDR tools, the Simplex processor will do a reasonable job and very quickly. Once you've applied it inspect the histogram and move in the upper and lower bounds appropriately. After that you need to find the right reverse S-curve to bring out the shadow detail with adequate contrast and then the same on the lights in a way that doesn't burn them out. In most circumstances you should be able to find a reasonable compromise. The resulting image won't be perfect. It will probably be low-contrast relative to the ideal and you'll need to correct that in Photoshop.

But that should get you started with FDR tools. Next time I'll look at the other two tone-mapping algorithms, Receptor and Compressor.

Till next time,

Bob

Look what FDR tools can do:

2010-01-28T17:34:00.001-08:00

Just a quick post to show what FDRTools (here and free trial download of FDRTools Advanced) can do to the colored test sheet that I used previously.

It does way better than Photomatix, doesn't it? See here.

HDRCreme: 'Echo Chamber of Love'

2010-01-28T16:56:00.000-08:00

Recently someone on HDRCreme posted a multiple-image HDR photo which depicted a steamboat at its mooring during a sunrise. View it here.

The sky above it was covered with grossly stuttered clouds and the picture was really nothing more than a great illustration of the inescapable pitfalls of multiple image HDR but, nonetheless, the sycophants on HDRCreme fell over themselves to say how great it was; '12 out of 10' was one of the milder examples. To read all the praise you would have supposed that this photo represented the Second Coming of Ansel Adams.

I, too, wanted to join the fun and so I suggested, in a nice way, that perhaps (just perhaps) the photo wasn't quite as perfect as the claque had declared and that there were one or two tiny points about the photo that ought to be improved. I did not mention that, in fact, its creator had been grossly sloppy and technically negligent. I also recommended a reading (of my own) which would point the original creator of this photo in a helpful direction.

Specifically I said this:

"A perfect example of 'stuttered' clouds. See this: http://sihdr.blogspot.com/2009/11/in-my-recent-statement-at-hdrcreme.html for a technical explanation of how slow objects need to be travelling in order not to have their leading edges reduplicated. This effort is a FAIL. Bob"

Well, then the sky fell in. I was called names that don't belong in a family blog like this. I was impeached for daring to criticize the darling of the hour and, reading the furious invective that resulted, you'd think that I had proposed a stiff Federal tax on digital sensors.

Now to say that HDRCreme is crawling with amateurs, sycophants, bullies and boot-lickers is just stating the obvious. It is the Internet after all. But there's more going on here than just the fury of incompetence when confronted with honest criticism. What I want to know is this:

How can it be that many people (people who are actually motivated enough to write in and comment on the picture) can look at this picture and rate it at the highest level (HDRCreme is a competitive site) and clearly miss its obvious technical flaws? More, they act as though this particular picture is as good as HDR can get.

This is a question of real interest.

The people who looked at this picture (including the person who made it) only looked at the boat, loved it, and ignored the rest of the picture. They literally did not see the flaws because, hey man, who looks at clouds? In other words the maker used HDR to bring out details in the clouds and then never looked at them.

And the claque on HDRCreme still treats these pictures as though they have subjects and only the subject is important.

But HDR eliminates subjects; it delivers a fully articulated surface and why would we want an articulated surface (i.e., use HDR) unless we wanted to look at it? Why would we fabricate a subject out of a picture surface that's completely articulated?

This results from what I call 'aesthetic inertia'. We can only conceive of pictures that have subjects (because for hundreds of years that's what Western art has produced); we have difficulty conceiving of a decorated surface that we enjoy for its own sake. But our ancestors did not have this difficulty; we've simply lost the habit that's all. HDR has a lot more in common with Medieval painting and painting before about 1423 than it does to anything produced since. Now we have a way of creating a completely articulated surface but we're still looking at the resulting pictures with eyes that just extract a portion of the surface as interesting. This is the way we've been looking at pictures, all pictures, since the Counter Reformation.

It's because of this selective attention, this 'aesthetic inertia', that multiple-image HDR practitioners are able to fool themselves about the insuperable technical difficulties of their craft. These drawbacks are many and I've pointed them out before.

1. Inherently limited subject matter.

(It's a matter of great amusement to me to hear practitioners extol the virtues of multiple image HDR and then complain about how the resulting pictures are only ever of cars, buildings, and machinery. They literally can't understand that the two are opposite sides of the same coin.)

How will we apply HDR to sports, to candids, to child photography, to portraits, fashion photography, to racing, to action pictures, weddings and even to most landscapes if there's any kind of actual weather involved if all HDR has to be from multiple images? To suggest that all these areas are off limits to HDR because we can't make moving subjects hold still is just silly.

2. Subject Motion. It's been well said (by me) that multiple-image HDR is like making a movie in which you hope to God that nothing moves. If multiple-image HDR ever really catches on we'll be going back 100 years to use neck braces on the subjects of portrait photography.

3. Confinement to a tripod. If we're really going to go to all this trouble we really should go back to the 8x10 cameras and have done with it.

4. Overall blurring and fuzziness consequent to misalignment in software. This is a serious problem for many products of HDR. As a result there's no way that many HDR pictures could ever be printed. If the photographer of the Southern Empress had printed his picture first he would have understood what was wrong and he never would have submitted it to HDRCreme.

5. Duplication of leading edges of moving objects. And, again I recommend this reading (http://sihdr.blogspot.com/2009/11/in-my-recent-statement-at-hdrcreme.html) to those who actually care about HDR.

6. Ghosting and ghosting removal which renders multiple image HDR back into single image HDR anyway.

7. Dishonesty and Subterfuge:

These insuperable difficulties result in a lot of lying by their makers about how HDR pictures are actually made. 'Multiple-image' photos are, in a lot of cases, really single-image photos. I won't point the finger (even I don't have that much courage) but I'm not the only one to have noticed many examples on HDRCreme and, par extenso, a lot of other HDR sites. Hint: look at some of the most highly resolved outdoor photos which don't show any traces of movement at all. There's one beautiful landscape in particular in which tiny people in the background have no blurring at all (and it's not because of anti-ghosting). Also look for landscapes in which clouds are razor sharp. Clouds actually travel surprisingly quickly relative to the time it takes to make multiple

exposures.

What we need in the HDR world is honest criticism. HDR needs critics that will look at the whole picture surface and critique it. HDR does not need forums for lickspittles who do nothing but praise each others work - 'echo chambers of love' such as HDRCreme.

What we need in the HDR world is more printing of HDR photos by their makers before posting on the web. Things that look good on the web are very often obvious failures when printed.

I think it would be good if we could abandon the fetish of multiple photo techniques. Multiple photo HDR creates more problems than it solves. This wouldn't involve much of a sacrifice because many people who claim to be doing multiple HDR are really practicing single-image HDR anyway. Either by subterfuge or through the practice of anti-ghosting. They should come out of the closet.

Most of all: We need to abandon the aesthetics of the recent past and learn to see the entire articulated surface in order to critique it. Without that HDR will always be just a gimmick.

Dynamic Photo HDR 4 and Photomatix: How do they handle color?

2010-01-25T18:27:00.000-08:00

It's time to look at a color test sheet and so I prepared one. Here's the original.

For this I chose a gradient from green to red and, over that, I imposed two sets of concentric disks, one green and one red. Both of them consist of smaller and smaller disks starting with pure red (255,0,0) or green (0,255,0) and graduated in 5% steps. If you care to examine this you'll see that even the human eye can't distinguish 100%, 95%, 90% green or red from each other. We'll see if the software can. Over the disks I drew sets of thicker and thinner red and green lines with red mostly over red and green mostly over green. There's a certain softness here from converting the tif to jpeg (uncompressed) and then posting on the blog. In the original the circles are crisp and sharp (where the human eye can distinguish between them) and they're evenly colored without blotching.

Let's see what the Photomatix Details Enhancer default settings can make out of this:

There's good separation here between nearly all the red and green disks. The gradient is even and without banding. But there's severe haloing, pinching, and flaring. And once again Photomatix is fooled by straight lines and performs the usual compartmentalization on unitary background objects.

I tried to fix some of these problems and here are the settings along with the results:

The haloing here is somewhat reduced and the separation of the disks (where they're not compartmentalized) is very good. But I could find no setting that would get rid of the flaring and pinching.

Many of the Dynamic Photo HDR results resembled each other. Let's start with 'Photographic':

We've seen in the past that Dynamic Photo HDR (DPH) tends to band on the gradient and that's what we see here. The green and red gradient is also somewhat reduced in saturation from the original (PX did the same thing). The red and green disks are well-separated but not as good as they are in PX. What we don't see here is the compartmentalization, the haloes, and the flaring that characterises PX's results.

Photomatix vs. Dynamic Photo HDR. The Big Tone-Off.

2010-01-22T11:11:00.000-08:00

I've created another test sheet and I'm devoting this blog entry to showing how Dynamic Photo HDR performs against Photomatix. Here's the new test sheet:

It consist of a stepped gradient in the background. That's overlaid with three overlapping circle figures. The one on the left starts at black and moves to more bright in 3% steps (HSL). The round figure on the right starts at black and moves upward in 5% steps. The circles in the center start at white and move downward in 5% steps. Lying over all this is a matrix of black and gray lines on the left and top. And there is a matrix of many white lines on the right. Let's see what Photomatix does with this on its default settings. This is Photomatix Pro version 3.2.7 for a Windows 7 machine. Here it is with default settings.

When I first saw this I thought it was a quilt. We see the same flaring that we saw in the last post. I found another site where this behavior was identified as 'pinching'. On that site, Outback photo, they said this:

"Pinching artifacts reduced (slight change in the Details Enhancer algorithm to avoid the problem of "pinching" artifacts showing along straight lines against a uniform background."

It looks like Outback posted this in the middle of 2008 when they were reviewing 3.0. If this is improvement, well, ...

Also we see the exaggerated shift in toning emphasis whenever a straight line is crossed. This is what I called 'compartmentalization' in the last post. I thought some more about this and I have some ideas about it. Details Enhancer is a local toning algorithm. Speaking in general such algorithms use a radius parameter to define how large an area to process; they then move and, using the same radius, start again and process the new area. The drawback to this algorithm is that sometimes the radius is large enough to lap over from one distinctive area to another. Bands of sky get processed in the same way as an adjoining mountain range. That leaves a light band (or halo) in the sky where it gets inappropriately lightened. One way to combat this problem is to modify the radius. In Dynamic Photo HDR (DPH) that's easy to do and users are sometimes encouraged to make the radius as large as possible so that its effects are spread out over a larger area and, thus, attenuated. The problem with that, in turn, is that you lose some flexibility in fine-tuning your results.

A little thought should convince us that the way to combat this haloing behavior is to precede the toning operation with one that breaks up the entire image into discrete regions. By using some edge-finder algorithm or the like the photo is decomposed into a number (probably a large number) of polygons. Now each polygonal region can be tone-mapped separately and stopped at the polygon edge so that no haloing results in other regions. So far so obvious. It seems clear that the Halo-Matix algorithm of DPH does not do that. The secret of using the Halo-Matix algorithm is to manage the halos; you're not going to get rid of them completely. But the Details Enhancer algorithm of PX does seem to do that. It seems as though that algorithm finds natural lines or edges in the photo and treats the contained areas separately from each other. Here it appears as though PX DE is using the dark and light lines of my test sheet to help break the sheet up into separate regions. Since each region is processed separately we would expect each region to have a different look. We actually see this in the results. I don't know that everything I've said is correct but that's the way it appears to be. The drawback with PX is that there are occasions when PX should ignore the lines because they simply cross over an underlying large feature that should be treated in a uniform fashion. This is probably related to the problem of distinguishing high-frequency features from low-frequency features. The low-frequency features in this sheet consist of the two or three 'towers' of circles (and the steps of the stepped gradient). The high-frequency features are the little compartments carved out by the underlying black and white straight lines (and sometimes the lines themselves as you can see in the thicker lines).

That's probably as far as we can go in interpreting this particular result.

I tried to improve this performance on the part of PX but I was only able to do this:

Again, I had to set the strength to the lowest setting and turn 'Luminosity' down in order to get an acceptable result. The compartmentalization is almost gone (naturally, 'Strength' is at 0) but no combination of settings would allow me to get rid of all the flaring or pinching. This is as good as it's going to get and, in fact, aside from the flaring the separation of dark and light disks is acceptable and a bit more visible than it is on the original (which is the whole point). We've lost the highest gradient brightness (is it my imagination but do the algorithms on both these products do better with darker values?). There's no point in pursuing this; we should turn to the 'Tone Compressor' algorithm. Here it is.

This is not a local algorithm. It addresses the entire picture surface and just squeezes everything in until it fits. The low-contrast inherent in HDR is most visible in such algorithms. You can see better into the shadows (but not the lights, apparently) but the area in which you might most be interested is squeezed into fewer bits and so contrast is lost.
This particular result isn't too bad, really. There's adequate separation in the left and right disks but, on the settings I chose (I couldn't find better), we've lost separation in the highlights.

Now let's compare this performance with Dynamic Photo HDR. DPH has seven algorithms and I processed the test sheet for each of them and I include them along with my settings (I tried to optimize each). Here's the 'Ultra Contrast' algorithm:

This represents a distinctive improvement over the test sheet original with better separation in the right and left towers and a solid performance on the gradient (without adequately separating the highest values). Not bad. Now the 'Smooth Compressor' which I take from its name to be a global algorithm:

Pretty good although it lacks something for the brightest values. Compare this to Photomatix's 'Tone Compressor' version above. Here's the 'Photographic' version from DPH:

This is the best we've seen so far. There's good separation in all three disks and every step of the gradient is at least indicated. What we've noticed so far is that the straight lines don't have any effect on the processing of the underlying figures. No 'compartmentalization'. DPH seems to do a good job of finding the high-frequency shapes and ignoring them. Now for DPH's 'Human Eye' algorithm.

Best performance so far on the gradient but the settings I chose reduced separation in the darkest disks (they're only separated by 3% as I said above).

Let's see if the 'Halo-Matix' algorithm lives up to our expectations:

Wow. It sure does. The gradient is rendered clearly (notice that the darkest zone has been lightened). All the disks are clearly distinguished. No flaring or compartmentalization (but there isn't anything like that in any DPH algorithm). Over all this might be the best. It sure looks like the place to start when you're doing HDR work.

DPH first became notorious because of its 'Eye-Catching' algorithm. It produced a lot of the hot and colorful results of the early phase of HDR. What's also clear is that it performs really well in black and white:

Not as good, on the whole, as 'Halo-Matix' but reasonable well. To wrap things up here is 'Auto-Adaptive'.

Not too good on the gradient but the clearest separation of the disks on the left that we've seen so far.

What do we learn from all this? So far I would recommend Dynamic Photo HDR for several reasons. It works better overall. It doesn't suffer from some of the obvious problems of Photomatix (which I and others have identified as 'pinching' or 'flaring' and also 'compartmentalization'). The product is more responsive to the controls. It has MANY more features than Photomatix both in terms of available algorithms and light controls. And the price is about two thirds of Photomatix' price.

I want to emphasize that we've only looked at black and white test sheets but, still, HDR is an idea about brightness and not about color. Also one legitimate use of HDR is just to take the brightness layer from Lab versions of photos and process that before returning it. So although these tests are limited they point the way to some important uses.

Till next time,

Bob

Photomatix Pro 3.2 and the Details Enhancer

2010-01-19T21:32:00.000-08:00

I have a copy of Photomatix Pro 3.2 (64 bit) that I've been running on a new Windows 7 machine. I subjected my test sheet from two posts ago to Photomatix to see what would happen. The results were interesting to say the least. The unmodified test sheet is here:

As you can see it's an overall gradient with two regions of nested dark and light squares. In addition there are 1-pixel wide black and white lines running the length of the picture. At the lower left quadrant there is a set of white and black lines intersecting with similar gray lines that are nearly invisible. What happens when we expose this to Photomatix?

Frankly there isn't any combination of settings in Photomatix Pro that can make this look acceptable. Most of the settings produce terrible results. Let's look at the default settings for 'Details Enhancer' and what they do to the sheet. Here it is:

This is hideous. To be sure, the background is an even gray without artifacts - acceptable. The dark squares have a reasonable, even good, degree of separation - acceptable. But the white squares are all mushed together at about the same level of gray: It appears to be a zone lower than middle gray. And they are primarily distinguished by their edges. But what about all the artifacting clustered around the straight lines? They look like the Northern Lights. I don't think there is a way to produce that in Dynamic Photo HDR 4. Not only are there flares all along the straight lines but there's a weird compartmentalized effect in the black squares. The straight lines overlaying the squares cause the toning algorithm to radically shift gears. The even performance of DPHDR4 is missing. And, worst of all, the huge flare in the white squares starting at the second black line and flaring upward in a bright band. Let's see if we can fix this.

If we turn the strength and the luminosity all the way down we can fix most of the straight-line flaring and much of the compartmentalizing behavior.

Frankly, this is about as good as it's going to get. In order to get any reasonable result we had to turn the strength and the luminosity down to the minimum and that still does nothing for the gross flaring in the center of the white squares. What we have at this point is, basically, minimal HDR. We have good separation in the black squares and barely distinguished white squares. These are separated by what looks to be about 2% decreases in brightness from the edge to the center over a whole brightness range of about 10%. And we still haven't been able to touch the flaring that originates at the black line that crosses the white squares.

I couldn't find a combination of settings for Details Enhancer that would eliminate the flaring. In order to get rid of the gross flaring in the white squares I had to set 'Light Mode' to 'Max':

But now the flaring has returned to the lines passing through the dark areas of the field.

These are, I believe, the most important controls for Details Enhancer. My readers may know more about how to set these controls in an optimal manner. But for now the performance of Photomatix Pro lags significantly behind Dynamic-Photo HDR 4 at least for this sheet. The flaring around the thin straight lines bothers me. I've always thought I could detect a certain mushiness in this product and if it's true that any setting of 'Strength' above 0 leads to artifacts then that might (and I mean 'might') form part of an explanation.

Next time I'll go into more detail about the 'Tone Compressor' algorithm.

Bob Consoli

Feedback!

2009-12-19T12:30:00.000-08:00

By the way, I'd love to have feedback on these things so don't hesitate to leave a comment at the bottom of each article and/or click on the follow button to let me know if you're interested in what I'm writing. Unlike most bloggers I'm well aware that you all are smarter than I am (and you probably have better equipment) so I want your feedback. It's easy; it's informative for me; you won't get any spam, and it's not like I'm asking for money for God's sake! By the way... Merry Christmas!

A Letter from ColdwaterJohn

2009-12-19T10:42:00.000-08:00

My good friend ColdwaterJohn has recommended this article on HDR to me. I went there and found a very well written basic multiple image HDR tutorial. Not quite our thing, I know, but we must keep informed. I was intrigued by this gentleman's (Harold Davis) previous article on using a single RAW image processed in different ways and then combining the versions with a layer mask and a gradient to reveal portions of underlying layers. I know that that's really off the track of SIHDR but you should go there and read his article anyway. I'll deal with his HDR ideas in time but first let's explore his ideas about multi-processed RAW images. In short (but read it for yourself) he describes a basic method of processing a single RAW in different ways (in his example he's correcting for different brightness zones) and then combining these versions in Photoshop. He takes one image in which the left is underexposed and, in Adobe Raw, he makes a version of that image in which the left is properly exposed. He saves that first image. He then makes a second version in Adobe Raw in which the overexposed portion on the right is corrected. He now has two partially correct images and he layers the bright image over the dark image and then exposes part of the dark image by using Layer Mask Hide All and then erasing part of the mask (the bright image) with a gradient.

A gradient.

Hmmmmm. It's the old neutral density filter trick but moved way up into post-processing. There are circumstances in which this approach is effective (Mr. Davis demonstrates one) but for most circumstances it introduces more problems than it solves. The basic problem is that the neutral density area is simply not the required shape. Here's an example where something else has to be tried:

In Athens recently S and I went to the Kerameikos which is a kind of ancient cemetery (not far from Plato's Academy) and I made this ghastly shot.

This is a beautiful funeral stele but the shot is totally ruined because I was shooting into the sun so that the temporary roof in the foreground is way too bright and the rest is murky and dark. A failed shot (Not to boast but I have a built-in advantage as a photo-blogger because I take so many bad shots that I always have something to fix).

Mr. Davis tells us to take the RAW and make a light and a dark version of it. No sooner said than done. Here's the light version:

The stele and the background look good but the roof is over-exposed. Here's the dark version in which the roof will be properly exposed:

...and of course everything else is murk.

Now we put them together. Of course it's very difficult to design a gradient that will combine these two. There's a dark zone on the top, a bright zone in the middle, and another dark zone on the bottom. Mr. Davis' shot had only a dark zone on the left and a light zone on the right. His shot was made for a gradient. This one isn't. But just to show that I'm a good sport I tried it his way. I layered the light layer over the dark layer and used a layer mask to hide the upper layer until the application of the gradient. Here's the shot again with a gradient drawn from the upper left edge through the peak of the shadow and ending at the edge on the lower right.

In fairness this looks pretty good. The shadow of the stele is more realistic than in any other of my versions but the stele sculpture is a little dark and the foreground shadow is way too dark (given that you were interested in it at all which I'm pretending that I am).

For my alternative version I approached this shot in the following way. I layered the dark layer over the light layer and then I used the history brush in Photoshop to brush away everything except for the roof (and the shadow on the roof - very critical point). The final result is this:

This is a highly-compressed jpg. You have to take my word for it that the original TIF had more detail in the roof but I think you get the idea. The only odd note here is that the shadow seems way too dark for the situation but I'd be interested in what my readers think about that.

In short, read Mr. Davis' article on RAW processing (it reads a little like my tutorial on dividing up an HDR into pieces and using each piece separately) but I'd take the gradient thing with a grain of salt.

Until next time,

Bob

Dynamic-Photo HDR 4 and Halo-Matix

2009-12-18T18:23:00.000-08:00

I promised another session with the HDR test sheets. Here's a new sheet I designed. It consists of some complicated black to white gradients. Upon those I layered two articulated squares. The first, in the upper left consists of a bottom layer of pure black followed by more layers each one brighter by 5% (the B in HSB) than the one preceeding. The second, in the lower right, consists of a base of pure white and then additional layers; each one darker by 5% than the one underneath. I then drew a number of pure white or black lines from one zone to the other. Many of these lines were one line width apart. At the place in the center lower left where the white lines and black lines come together I drew more lines in near-middle gray (136, 136, 136). You can just barely see them. Here's the chart:

It all looks very Piet Mondrian. I then fed this to the HDR software which happens, in this case, to be Dynamic Photo HDR 4. I used 'Auto Adaptive' and full HDR. Then I used the Halo-Matix tone mapper and that produced this with default settings:

I tried Halo Matix because of the good separation between similar tones that it showed in the last test and here we are not disappointed. All the dark squares are clearly separated; even the 100% and the 95% black square inside of it. These could not be seen in the original. The effect of the transformation has been to progressively exaggerate the lightness of the nested dark squares. Contrariwise it has exaggerated the darkness of the nested white squares so that they, also, are separated. The one pixel-width lines, black and white, are all much more clearly visible in this version than in the original. If you go to the Picasa album here and enlarge the Halo-Matix version (click on 'Full Screen') you can also see some gray lines clustered just above the short black lines in the lower left center. These gray lines are also more visible after HDR processing (particularly the vertically-oriented ones).

Now let's look at the same image when made bright and the halos increased to the maximum size:

Here the dark squares have maximum separation but the increased brightness has merged the brightest squares into each other with generally poor separation of tones - but better as they get darker. Turning down the brightness of the image improves things as we see in the next image. It is as dim as the settings allow but with max halo size:

The two lowest black squares blend into each other here to a much greater degree but the white squares are mostly clearly separated. They are given a much greater range of darks.

Now let's reduce the size of the halo setting to a minimum:

This image was produced with the brightest setting but with a minimum halo size. Again the bright setting forces the brightest squares into each other and the small halo is clearly seen especially around the lines. In the darkest squares each seems to run through exactly the same brightness ranges and the separation is achieved by enhancing the differences at the edges of the squares.

Last of all comes the dim setting with minimum halo size:

One thing we notice in these low-halo settings is that they will certainly degrade tiny details that are close together. Because each line has a halo around it these halos run together where the lines are close together and thus degrading the entire area.
I created a chart to sum up what I think I've found and here it is:

If your eyes are like mine you'll have to click through to read it. So it turns out that brightness has a relationship to tone separation for darks (bright) or lights (dim) and tonal separation can be achieved by radically altering the tones or by leaving the tones more or less the same and simply emphasizing the edges. The chart should help you to pick the settings you need for whatever it is you're trying to achieve. I think that for the next post I should find a desaturated image and experiment with it by trying these different settings to see what difference it makes 'in real life'.

Until next time...

HDR Test Sheets

2009-12-15T18:23:00.000-08:00

Let's try to get this HDR thing on firmer ground. We need a repeatable way to predict the performance of HDR for any set of parameters. I'm going to use Dynamic Photo HDR 4 from MediaChance and I created the following simple test sheet with Photoshop CS4. It consists of two black to white gradients at right angles. In the dark portion of the test sheet I typed the word 'abracadabra' in black at three different locations. In the light portion of the test sheet I typed the same word in white at three different locations. We should be able to see, at least, some set of parameters that would let us read all six words. Without further ado here is the test sheet (and if you can do better let me know):

I transformed this in Dynamic Photo HDR 4 using the Auto-Adaptive method of re-mapping and I selected 'Full HDR Process'. When that was done I tone-mapped it using all seven algorithms at the default settings. Here they are. The first is 'Eye-Catching':

The next is 'Ultra Contrast':

Now comes the real surprise: 'Halo Matix':

I can't resist a gasp of surprise. I would never have guessed that this algorithm (which seems to feature garish haloes and color distortions) is far and away the best at separating details from a similarly colored background. Look how even the background tone is. If anything it has slightly reversed the gradient. And all six words visible as plain as day. And notice that the white versions of 'abracadabra' are rendered in something like real white. This distinguishes it from 'Human Eye' which picks out all three white 'Abracadabra's but in shades of gray.

Now comes 'Smooth Compressor':

What's going on with 'Smooth Compressor' also seems clear. I believe that this algorithm simply takes the entire tonal range and compresses it into the number of output bits available (7 or 8). We'd expect this to be a low-contrast solution which would make the words hard to read. That's what we actually seem to see. See this post for my ideas about reduction of contrast under some compressive HDR algorithms.

After that comes 'Auto Adaptive':

'Photographic' is next:

And, finally, 'Human Eye':

I'm going to put this on-line right now. Next post I'll present the seven results from performing what MediaChance calls 'Eye-Catching' remapping.

Breaking up the HDR-processed image

2009-12-11T10:07:00.000-08:00

We were in southern Greece in the middle of October and, one day, Sue and I found ourselves in an old Mycenaean tomb. These are sometimes called 'tholos' or 'beehive' tombs; they're made of rock, domed, and covered with earth except for the entrance. This particular tomb had fallen in centuries ago and nothing was left except the circular wall up to a height of about four feet and the entrance which had a gigantic lintel over it. I pointed the camera upward and took a shot of the lintel but because the sky was overcast and bright the underside of the lintel was underexposed. What to do? I could see enough texture under the lintel so I supposed that HDR might retrieve something. Here's the original:

Pretty grim.

I then proceeded to expose this mess to my HDR software. I used Dynamic Photo HDR 4 and the settings were as follows: I chose 'Eye-catching' over 'Adaptive' and then selected 'Full HDR Process'. Once that was complete I chose the 'Auto Adaptive' tone-mapping process with the following parameters: Vivid Colors: 0.27, Light Radius: 4.1, Light Strength: 0.52. Other choices are possible.

I felt that I came out too bright but this is still usable. But I want to use the underside of the lintel separately from the rest of the picture. The underside of the lintel is closer to what I really want but the rest has to be toned way down. First I'm going to move the under side of the lintel and modify it by setting the opacity. Then I'm going to use a different opacity for the rest of the picture.

First I used a mask to select just the underside of the lintel. Here's a Photoshop screen shot.

Once it was a selection I then copied it and pasted it over the original. It was way too bright there so I moved the opacity of the underside of the lintel down to 64%. Here's the result.

Now the underside of the lintel is where it should be but the rest of the photograph could benefit by having the rest of the HDR-processed version layered over it. I had saved the selection for the underside of the lintel. I got that back and inversed it to select the rest of the picture without the lintel. I then copied that and pasted it over the ongoing composite. Here's a screen-shot of the result; it shows the composited lintel after opacity adjustment along with the rest of the picture before opacity adjustment. If you can make out the layer menu in this next shot you can see that the original is the bottom layer, the lintel (only) is the middle layer, and everything else in the HDR shot is the top layer. Here it is at this stage:

Believe it or not this is progress. Yes the composite is garishly too bright but we're going to fix that by adjusting the opacity just for the part of the image that isn't the lintel underside. I took the lower two layers (original and lintel underside) and merged them. That leaves two layers. The composite is on the bottom and the HDR shot minus the lintel is on the top layer. Now simply adjust the opacity of this top layer down to 48%. That leaves enough brightness so that details can be made out and it is appropriate to the lintel underside. Here's what it looked like after I adjusted opacity. Remember my goal in these sessions is clarity, not beauty. Because I've been able to rescue the clarity this picture is now suitable for Squinchpix where you can see what I actually uploaded.

...and here's the original again so that you can compare:

The moral of the story is that you can cut up the product of any HDR transform any way you like and re-composite it back on the original with any adjustments you choose.

And because you've been such good students here is your reward: This is a shot of the west coast of the Mani peninsula in southern Greece on a stormy day. It was also processed with a modified HDR procedure. You're looking south here towards Areopolis.

Until next time,

Bob

2009-11-28T18:03:00.000-08:00

Well, I'm finally back from Greece and I have my head reasonably together. Now I can turn my attention to the blog and this post takes the form of a work session with one of the pictures I brought back from Athens. In the middle of the city is an ancient church, the Kapnikarea. It's a small Byzantine church that sits right in the middle of the city and down from the main level by about two or three feet. The result is that you're almost at the level of the roof before you descend the stairs to the church itself. On the south side is a little porch with an apsidal ceiling which is covered with a beautiful mosaic of Mary ('theotokos') and the Christ Child. I stood just to the south of that and, on a beautiful day in early October, took this ghastly shot. What's wrong with it? Let me count the ways.

What's wrong with it is that it's simultaneously underexposed and overexposed. The overexposure shows up in the washed-out appearance and lowered contrast of the brick and ashlar facing; sure it's an old building but THAT colorless? It looks like it needs a good dusting. At the same time the mosaic of Virgin and Child is lost in the dark muck under the porch roof. I thought, 'Aha! HDR!' Maybe there is a way to punch up the brightness of the mosaic and simultaneously push back the brightness of the building itself.

In my Dynamic Photo HDR software I took a single image of this scene, a TIF and selected 'Eye-catching' and 'Full HDR process'. The first good news was the result of the primary HDR process. In the preview I could see that every tessera in the mosaic was razor-sharp. After the tone-mapping step was done I tried the various preliminary methods that Dynamic Photo provides. 'Eye-catching' with 'brightness' set to 20 seemed o.k. The mosaic jumped out at you but the building was still washed out. I couldn't find any setting that would make 'Ultra-Contrast' work (but then I never can). 'Halo-Matix' with Color Saturation set to 0.22 and Light Strength set to 0.75 seemed acceptable. I couldn't find any settings for 'Smooth Compressor' that would work. I finally settled on 'Human Eye with Color Saturation set to 0.62 and Light Strength set to 0.64. This seemed to have an acceptable balance between a lightened and colorful mosaic along with a building that isn't totally washed out. I'm not trying for the finished result here; just trying to get something that I can work on. Here's the HDR result:

This is a huge improvement over the original. Now we have something, as I said, that I can work with. Before I didn't. The building is close to what I want (it wasn't beautiful to start with). I need to lighten up the mosaic. I went into Photoshop and went into Quick Mask mode. I placed a mask over the mosaic and exitted Quick Mask. I then selected 'Select'->'Inverse' to actually select the mosaic. I then used 'Image Adjust' -> 'Curves' to manipulate just this area which I made lighter and a little contrastier. The result is this:

I still wasn't satisfied with the building even though the mosaic was coming along just fine. I got back the mask I had drawn and inversed it to select the building. Using Curves again I tightened up the contrast and adjusted the brightness. Here's the final result (not sharpened):

At this point the image looks completely unrealistic but in this situation that's not a concern. My goal on Squinchpix is photographs which show you what's there and I don't mind tweaking them. You've got to admit: the mosaic looks a lot better.

Here it is sharpened:

...and here's the original so that you can compare them directly.

Till next time...

Single Pixel shift in HDR alignment

2009-11-02T07:43:00.000-08:00

In my recent statement at hdrcreme.com I stated that multiple-image HDR was difficult to do and, for that reason, had a limited application future. Specifically I said this:
“Multiple-image HDR is difficult to do right. It’s almost impossible to align multiple digital exposures correctly as a look at most multiple-image HDR photos will clearly show. Even one pixel off will degrade the picture.”
I instanced pictures of people, of sports, and even of landscape where there was anything moving such as foliage or clouds. My statements elicited a reply from ColdwaterJohn (CWJ) whom I know to be a great exponent of multiple image HDR and I know that he is a solid and capable worker in this area. CWJ heartily defends multiple image HDR (as he should) and supposes that a ‘rock-solid’ tripod is the answer in most instances. I don’t share CWJ’s sanguine view of this technique but let’s look at it a bit more rigorously. So here’s a question: What difference does a single pixel shift make for a subject at any given distance? In other words, if two images, A and B, are shifted by a single pixel with respect to each other then how much of a difference does that make? Let us turn to our famous conceptual camera and stipulate that it has a viewing angle of two radians (to make the math easy, this is about 114 degrees) and that it supports 4000 pixels horizontally. Now imagine that our camera is at the center of a circle and we make two exposures. The first is a reference exposure but the second exposure is misaligned in the horizontal by one pixel. How large an error is that? Let’s take 100 yards as the radius of our circle. Simple trigonometry says that the width of such a pixel at 100 yards is the tangent of the angle times 3600 (inches in 100 yards). This will give us our answer in inches. So tan(1/2000 radians) * 3600 = 1.8 inches which is accurate to five places. So a misalignment of 1 pixel for our conceptual camera at 100 yards is a misalignment of 1.8 inches. This means that an object of width 1.8 inches at 100 yards will appear to be 3.6 inches in width under ideal conditions. Thus the following table:

Distance                Error
---------------------------------------
1 inch                   0.0005 inch
12 inches              0.006 inch
100 inches            0.05 inch
1000 inches          0.5 inch
2000 inches          1.0 inch
3600 inches          1.8 inch
1 mile               31.68 inches or 2 feet and 7.68 inches
10 miles        316.8 inches or 26 feet and (about) 5 inches

The first thing this suggests is that the smaller the linear distance covered by a single pixel the smaller the error so, for any given subject and at any given distance, the HDR photographer should attempt to box the intended subject as tightly as possible consistent with his or her purpose. That way the number of pixels (which is fixed) covers the smallest angle and any misalignment is minimized. If our conceptual camera had a zoom lens that would let me box in only one radian at 100 yards the error would be immediately halved. Another thing is that you should shoot at the highest number of pixels that your camera supports. The error for 4000 pixels is half the error for 2000 pixels.

The second thing is that you should absolutely keep the number of exposures to the bare minimum to fulfill your artistic goals. (see Consoli’s conjecture in a previous post). I think, (but I’m not sure), that the probability of a single misalignment grows only arithmetically with the increase in the number of separate exposures. For two exposures let’s say that the probability m of a misalignment is 1*m. For three exposures the probability of one misalignment is 2*m. For four exposures the probability of one misalignment among the four should be 3*m. If I’m right then that’s good news for multiple image practitioners. But I’m not sure and I invite my readers to consider this problem.
What is the probability of making such a misalignment given that the subject is absolutely still? I don’t know. Whatever it is it’s a function of the stability of the camera, whether the camera has a mirror or not, whether the mirror locks up or not, whether the shutter is released manually or by radio release etc., etc. (CWJ has already pointed out some of these factors) I confess that I have no way of approaching that probability using numbers.

If the camera is absolutely still but the subject is moving then (for our conceptual camera and for any given distance) the subject has to move more slowly than the distances given from the time of the first shutter release to the time of the final shutter closure. Let’s say that the time to make all these exposures is 10 seconds. An object a mile away must be moving more slowly than 3.168 inches per second in order not to stutter after 10 seconds. This is 3600 * 3.168 inches or 11, 404 inches per hour or 0.18 mph. The average cloud probably moves much faster than that. That’s why we see so many HDR landscapes with stuttered clouds.
So far I’ve only considered yaw. The numbers for pitch would be of the same magnitude (since it’s my conceptual camera I can specify that it supports 4000 pixels vertically). Each practitioner should work out the details for his or her own camera. Since I’ve forgotten all my spherical trigonometry I can’t work out what the error for roll would be but I do know that any error in roll would be more severe at the edges than at the center. I sincerely hope that my readers can help me out in this area. And although each of you should, as I said, work out the details for your own setup I think you get the idea about the kinds and sizes of errors that we’re talking about. Can multiple-image HDR be done at all? Of course it can. Practitioners like CWJ are teaching the rest of us how it’s done and sharing their knowledge. For that we are definitely grateful. But the multiple-image HDR perspective is skewed. The thing is, HDR is not an end in itself. It’s a means of extending the tonal range of camera capture beyond the pitifully narrow range that it currently supports. The magic was never multiple exposures (which are as old as photography itself. See Michael Freeman’s book on Digital Black and White Photography for a good and effective non-HDR approach using multiple images); the new magic is the tone-mapping algorithms themselves. They can be widely applied; they don’t require multiple exposures. There are other means to this goal of tonal extension. Support by the camera manufacturers of more capable sensors is the ultimate winner. There’s some evidence that they’re starting to realize this. Multiple image HDR will always be the province of the enthusiast but I sincerely believe that due to its own severe limitations it doesn’t have much of a future. (There are too many subjects that just can’t be approached using multiple exposures and I find CWJ’s remarks about Photoshop unconvincing – although perhaps he can tell us in more detail exactly how that’s done.) It could be, however, that multiple image HDR does have a future in the studio. I’ve seen demonstrations (you probably have too) where exposures are made, lights are shifted, more exposures are made, etc. and all the exposures combined using Photomatix or the like. Great results can be achieved like this.
For a future post I (or we) need to consider sharpening as a cure for small misalignments.

Best to all and I look forward to CWJ’s reply,

(A version of this was published at http://hdrcreme.com/forums/4/topics/53?page=1#posts-198.)

Bob Consoli

CiderHouse is Going to Greece

2009-09-27T14:15:00.000-07:00

On October 1 CiderHouse is taking a long-awaited photo vacation in Greece. We'll be gone until November 1 so don't look for any updates during that time. On the other hand, if you'd like to submit a piece to CiderHouse then just leave a comment on this post and I promise I'll get back to you somehow. For example, just tell us what you're doing with HDR. Others will be interested!

Have a nice October,

CiderHouse Gang

The problem of contrast in SIHDR

2009-09-08T19:59:00.000-07:00

How many images do we require for effective HDR in any given situation? I'm going to break away from the purpose of this blog for a moment and consider multiple-image HDR. So let's do a thought experiment. Let's say that, for a 32-bit target space, we take 32 pictures and each one is one f-stop apart. And let's say that we're using a conceptual camera which takes them all simultaneously on, oh, 32 different sensors. Now we dump these 32 pictures into Photomatix or similar and process them onto a 32-bit space. Nirvana, yes? Of course. We guarantee that we have all brightness ranges covered multiple times. Photomatix churns and produces the desired mapping of which, as usual, we can see about 8 or 9 bits worth at any one time.
Oh. Problem.
The range which we really desired to see only covered about 16 bits and now that 16-bits will be tone mapped into about half of the 8 or 9 available tone-mapping target bits. To sum up; the part we were interested in is now occupying about 4.5 bits. Can you say 'low contrast'?
It's exactly this tonal compression that lies at the heart of the 'low-contrast HDR problem'. Is it possible to have too many images for HDR processing? It certainly IS possible.
I call the following 'Consoli's Conjecture'. The optimal number of exposures for multiple-image HDR is approximately equal to (B - 6) / 2. I justify this in the following way. B is the number of brightness ranges over which we desire visible exposure. To make the math easy let's say that we desire 16 brightness ranges. So B is 16. I say that we only need exposures two stops apart. Therefore we don't need 16 exposures for B = 16. But neither do we need them on either end. Let's say that the tail of the lightest and darkest exposures extends to the edges of what's desired and that the lowest three zones of the darkest exposure and the upper three zones of the brightest exposure will give adequate rendering of the dark and light edges of the range. That accounts for the 'minus 6' part of the expression. Therefore, for 16 brightness ranges we have (16 - 6) / 2 or 5 exposures. And these are exposures 2 f-stops apart with the middle exposure centered right in the center of the range of interest. Other people have certainly worked this out; if my readers know more I wish that they would enlighten me.
Of course, all of this assumes a smooth global compressor tone-mapping function. This would be one which would take 16 brightness ranges and smoothly map them on to an 8 bit space. That is, half what we would normally expect 16 exposure zones to require.
It is exactly this that leads to the low-contrast problem in HDR. Here's an example.

This is a picture of the Badia Tower in Florence and it's much as it came from the camera. See the disgustingly underexposed expanses on left and right.

I processed this through Photomatix (which I obviously don't own) and that rendered this amazing result:

The right and left sides have opened up amazingly but notice the dramatically lowered contrast in the center; in the tower itself. It's as though a veil of gauze had been drawn over the scene. Let's detour a moment and look at the following version which started with the original and just used Photoshop Levels control to bring the right side up to about where it was in the HDR version:

Here the right side is about the same as the HDR version but, of course, the center of interest is completely washed out.

But, in the end for Squinchpix, I didn't use the HDR version. I used the following version because it was easier using traditional Photoshop techniques to overcome the problem by dividing up the picture space:

This is it. I created a brightness map in Photoshop CS3 and used that to select the various areas for brightness adjustments. It never lost contrast. Compare it to the others. A brightness mask is a kind of local tone mapper of the kind you find in HDR software. One way to overcome the severe contrast problem in HDR is not to use it at all.

To my friends on HDRCreme: A Manifesto on HDR

2009-09-08T19:27:00.001-07:00

HDR is the most important development in photography since the discipline was invented.

That’s because photography and photographers have always lived with a dirty little secret; the brightness range reproducible by straight photographic techniques is laughably small. HDR has now expanded that by, let’s say, doubling the capturable brightness range.

HDR is an idea about brightness.

It is not an idea about color. Color is lovely and because of that people always assume that color is more important in image-making (painting or photography) than it really is. But, compared to the importance of brightness in these disciplines, color is nothing. All painting and photography is a meditation about brightness. Nearly all changes in these disciplines involve a new way of thinking about how to render shadows.

HDR processing consists of two phases. In the first phase multiple pictures are mapped from a 12- or 14-bit capture space onto a 32-bit picture space. In the second phase the entire brightness range actually used in the 32-bit space is tone-mapped back onto a smaller space; one which is, perhaps, as wide as nine bits. This is done because no viewable medium can display more than nine bits of information (most display less). This is starting to change with certain (scarce) wide-range display devices but it doesn’t affect us yet. For us nine bits is still a hard parameter. If you want to see it then it has to fit into about nine bits of displayable space.

As I mentioned above people currently are mapping multiple images into 32-bit space in order to extend the brightness range. This is a time-honored technique and it is as old as photography itself. I honor those who are working in this area; they are getting amazing results and teaching us a great deal.

A case can also be made for single-image HDR. Multiple-image HDR is difficult to do right. It's almost impossible to align multiple digital exposures correctly as a look at most multiple-image HDR photos will clearly show. Even one pixel off will degrade the picture. The insistence on multiple-image HDR must exclude certain important types of photography (sports, racing, nature photography, children, candids: anything with rapidly moving objects). More affirmatively I think that tone-mapping techniques have revealed to us just exactly what we’ve missed in our single images. Tone-mapping techniques have led to the potential rejuvenation of every photograph ever taken as long as we have it in its original form (RAW file, TIF file, photo negative, etc). Personally I was astounded to discover the actual brightness limitations I’d been living with over the years. Single-image HDR does have its challenges and I think there’s a lot of work to carefully define those challenges and how to overcome them. That is the purpose of this blog. None of this is intended to denigrate people working in multiple-image HDR. I recognize both the hard work and the astounding results. There’s room for everyone. People often use the term ‘pseudo-HDR’ when referring to single-image HDR. I, for one, think that we should eschew the use of this term. Why? Because it doesn’t aid the thinking process. There’s nothing phony or false about single-image HDR. It’s simply an HDR that skips the 32-bit space mapping function. Or if it is done then it is idempotent. With single-image HDR, in effect, we go straight to the tone-mapping phase. It’s just my personal hobby-horse but I know that the single-image HDR workers also have something to tell us.

Remember, though, that multiple-image HDR and single-image HDR are simply rapidly passing technological phases. But multiple-image HDR will pass first for the simple reason that there are billions more single images out there that would benefit from HDR processing (a business opportunity for someone). It’s inevitable that the camera manufacturers will support an HDR mode directly and they’re going to do it sooner rather than later. The ideal solution would be a sensor that captures, say, 16 brightness zones and a camera that incorporates it and which makes the number of desired capture zones user-selectable. Does anyone doubt that the next three years will produce such a camera? I don’t.

Extended brightness range is not the answer to every pictorial problem. Many photographs are taken expressly because much of the picture space is effectively blank through over-exposure or under-exposure. This is the modern photographic aesthetic and we should remember that HDR flies directly in its teeth.

HDR practitioners sometimes suppose that extended range is always desirable.

It isn’t.

To support the continuously articulated picture surface that HDR promises will require a new aesthetic. One that’s more characteristic of modern art (or art before the fifteenth century) than ‘photographic realism’. (There are good remarks about all this in Hockney’s Secret Knowledge, 2001). At the very least HDR (in whatever form it comes) will force us to re-learn that every picture is an artifact; a completely artificial creation. For example, there’s a photograph on HDRCreme of the Spanish town Ronda (http://hdrcreme.com/photos/3236-Ronda-Panorama). When confronted with this picture one of the anonymous reviewers said: “I'm not sure it needed the full HDR treatment, which in my mind adds a level of un-reality to what is otherwise a good shot.” Inadvertently Mr. Anonymous uncovered a profound truth in his remark. The only thing is that Mr. Anonymous didn't realize that all pictures are profoundly unreal. There is no purity in art. Every image, no matter how it is made, is completely artificial. There is no ‘realism’ in photography as a good look at that Ronda picture will tell you.

So here is my manifesto:

a. Every picture is an artifact pure and simple.

b. Multiple-image and single-image HDR are just passing phases and they will be replaced with a camera that supports extended brightness ranges.

c. We should investigate the older decorative aesthetics in order to find guidance to this new world of the continuously articulated picture space.

d. Stop the flames. Flames will not help us. There is nothing sillier than grown men (and we are mostly men) arguing about the appearance of a picture. We can learn from everyone.

e. Let’s stop beating up on the extreme-color people. This nearly out-of-gamut color is a mark of the earliest phases of HDR. No one, even the designers, really knows how to use these sliders for optimal results. Most books on HDR recommend that you move the sliders around and ‘play with them’ in order to learn how it’s done. There’s a place for expressionist color in photography as much as there is a place for it in painting.

f. We have much to learn from our painter cousins. They have been practicing local brightness optimization for centuries before Chevreul pronounced the Law of Simultaneous Contrast of Tones. They could instruct us if we had the humility to listen.

g. We have an important task in hand and that is: How is HDR to be controlled? How, for example, do we modify brightnesses without modifying colors? What shall we do about the noise in single-image HDR? What shall we do about the lack of contrast inherent in all HDR? What does it mean (if it means anything) to tone-map color? How many images is optimal in multiple-image HDR before the lowered contrast degrades tones in the brightness ranges of interest? We need to know the answers to these and a myriad of other basic problems before we can convince others that HDR is the advance which it really is.

h. I mentioned it above but, above all, humility. Wynton Marsalis says: ‘The humble get better.’

Robert Consoli

Another review about the Sony a550

2009-09-07T11:03:00.003-07:00

For some more information about the Sony alpha 550 see this more wide-ranging (but uncritical) review:
Sony Alpha 550 Digital SLR Camera

It appears that, in addition to what we already knew, the new Sony not only can take two pictures in HDR mode but will align them before combining them.

A Typical Problem in SIHDR - noise.

2009-09-07T09:39:00.000-07:00

I want to devote this post to the problem of noise in SIHDR. It's more of a problem in single- than it is in multiple-image HDR because, of course, there's less information to work with. Here's a typical example. It's the Art Academy in Dresden and, taken in the early morning, it's underexposed. There's almost no information under the intrados of the arch. Here's the original:

Under the arch it was easy to find (RGB) 1,0,0 points and lots of other points like 5,3,4, etc. I just know that this is going to be noisy after an HDR transform. But before we try HDR let's look at the noise by turning the Levels slider in Photoshop all the way up. Let's really look at that shadow.

Here there was a huge gap at the top end so I moved the white point down to 189. The middle tone slider was moved to 2.22. Yes. It's pretty bad. Not only chromatic noise but luminance noise. What to do?

This (above) is an HDR transform made with Dynamic Photo HDR and it's the one I'm going to use. The color noise is not as gross as if we'd just used 'Levels' but the luminance noise is still a real problem.

In Photoshop I converted the image to grayscale. The color noise is now gone but, having done that, I'm committed to using this as the L channel in the final image. Let's do something about the luminance noise.

To cure the luminance noise I went into quick mask mode and painted the area under the arch. Having selected just that area I removed the noise with Noise Ninja. In Noise Ninja the contrast setting under luminance had the greatest effect. There's now slight blurring here but, in the context of the whole picture it won't matter much.

Here's the composite (above) after using the HDR-transformed image as the new L-channel.

This has been sharpened using Unsharp Mask. I used a high radius, low amount setting (HIRALOAM). Radius 49.8, Amount 15. This put the contrast right where I wanted it. So the moral of the story is that you may have to use aggressive noise-riddance techniques if you use single image HDR.

One last point. There is still a little luminance noise directly under the arch (not on the glass). Sometimes you can get rid of that minor noise with the blur operator in Photoshop. That's only for very minor noise on smooth surfaces. Otherwise you're better off with the control afforded by Noise Ninja or the like.

HDR from a painter's perspective

2009-09-04T10:16:00.000-07:00

This is a painting by Gentile da Fabriano. Fabriano was a master painter who was working at the very beginning of the Renaissance; a master of the so-called 'International Gothic' style. This Nativity, which is now in the Uffizi in Florence, was painted about 1428. As we look at his painting we see that it was created according to a very different aesthetic from that which we are accustomed to now.

It is an aesthetic that emphasizes the artificial nature of a picture; every part of the surface is to be used. Everything is to be articulated. There is a sort of 'horror vacui' in this aesthetic. Fabriano's typical products have a close relationship, for example, to the rug.

And there are corresponding ideas elsewhere in Islamic art as we see in this next picture of the Mihrab in the mosque in Cordoba, Spain.

In this aesthetic empty space is, well, wasted.
In the work of artists like da Fabriano no part of the picture space is emphasized over any other part.

Some artists have this tendency to a very great degree - I would number Albrecht Altdorfer among these. Here is an Altdorfer painted about 1526 which depicts Susannah, the Old Testament heroine, at her bath. It bears an uncanny resemblance to a photograph processed in HDR. In just a moment we're going to see such a photograph. The eye wanders over the surface of such paintings thoroughly enjoying whatever it comes across. Altdorfer's delight in the pictorial is so pronounced that we are hard put to it to find Susannah herself (the red-head in the lower left corner whose hair is being combed by the maid). We get the feeling that the title itself is a condescension of later generations. Clearly Susannah's encounter with the elderly lechers is the last thing Altdorfer really cares about.

I encourage you all to look at this in more detail:

http://upload.wikimedia.org/wikipedia/commons/8/8b/Albrecht_Altdorfer_037.jpg

The elderly lechers I mentioned may be seen at the very base of the tree on the far left.

Here is another famous example. 'The Conversion of Saint Paul' by Brueghel (1567). What's remarkable about this and other pictures of this type is the attenuation of the idea of subject. This depicts Saint Paul's famous conversion on the road to Damascus. The subject, Saint Paul himself, is nearly impossible to spot in this painting. If we didn’t know better we would suppose that the figure with the yellow tunic in the right foreground was the real subject. For lack of a focus the eye wanders around the painting discovering, with some delight, the distant sky and sea, the dark forest, the ragged rocky landscape, the strange lowering clouds in the pass, not to mention the plethora of men and horses, flags, burdens, and the pikes of the soldiers travelling with the Saint.

Let's look at one more picture of this type.

This is another example of the same aesthetic orientation. It is Saint Jerome reading in the wilderness. This painting is by Giovanni Bellini who painted it late in his life, in 1505. Here the subject is a little more obvious but Bellini's delight in the picture surface is still very apparent. His 'subject' is squeezed into the lower right corner. The eye travels across the picture space discovering rocks, a walled town, a rabbit, and any number of accurately rendered plants.

Now let's look at the same subject rendered in a completely different way...

This is a painting of Saint Jerome in his cell. It was painted by Caravaggio (Michelangelo Merisi) shortly after 1600. Clearly this painting arose from a very different set of aesthetic conceptions. Here there is no 'horror vacui' - indeed, large parts of the picture space are, we would say, 'seriously underexposed'. Caravaggio was a pioneer of this aesthetic, reacting against the eirenic and serenely pale pastels of the Carracci family. This darkness is characteristic of all his work; for that reason he and his followers (which include Rembrandt) are referred to as 'tenebristi' or painters 'in the dark method'. Most of us feel rather more at home with this aesthetic. We are constantly exposed to pictures of this type, both paintings and photographs. They are characterised by being organized around broad areas of overexposure or underexposure.

Such pictures are organized around a specific 'subject'. They have 'emphasis' in a way that the previous pictures do not. Such pictures are always 'about' something. No longer do we experience a continually articulated space as we do with the productions of a da Fabriano, a Brueghel, or an Altdorfer. No longer does the eye wander around the picture space in a leisurely fashion. The space is organized by vacuums so as to emphasize the artist's intent. It is no accident that Caravaggio was the great propagandist of the Counter-Reformation. Pictures in this aesthetic are meant to instruct; things are too serious to be otherwise.

I propose to refer to these organizing vacuums as 'scotoma' or 'deficits'. When I use these words I mean precisely these light or dark picture vacuum spaces which have organizing intent.

Pictures with a continuously articulated picture space I will call 'Type A' pictures. Pictures which use deficits as their organizing principle I will call 'Type B' pictures. No ordering or relative worth is implied. Altdorfer and da Fabriano were producers of 'Type A' pictures; Caravaggio, Honthorst, Rembrandt and others were producers of 'Type B' pictures.

Let's move these concepts into the world of photography. Here, for example, is a perfectly charming photo of two girls feeding pigeons in St. Mark's Square in Venice.

(http://www.hdrsoft.com/images/stmark/before.html)

There are two organizing principles at work here. The over-exposed sky directs our attention to the lower two-thirds of the picture. And the subjects are wearing orange and yellow clothes which helps to draw our attention to them. A Type B picture, surely.

This picture has now been transformed with HDR to bring out the details in the clouds.

http://www.hdrsoft.com/images/stmark/after.html

It is a very different picture from the foregoing. Now the deficits have been removed. Our eye is still drawn to the orange and yellow-clad girls but our experience of the picture is completely altered. There is no organizing vacuum. Dare I say it? The clouds are even more interesting than the pigeons. This is a Type A picture; one with a continuously articulated picture space.

The moral? Traditional photography produces Type B pictures; they are organized around deficits.

But HDR has the potential to produce Type A pictures; pictures with a continuously articulated picture surface.

More later.

Manipulating the K-channel in CMYK.

2009-09-04T09:42:00.000-07:00

In a previous post I showed that you can manipulate just the L-channel in Lab and leave the colors unchanged. Can we do the same thing in CMYK? Yes we can.

This is the original of a photograph that I took in Dresden, Germany.

The usual unsalvageable deep shadows of early morning along with burnt highlights where the sun is striking. Let's convert this to CMYK and take out the K channel and look at it.

Here's the K channel (Black) which I extracted from the previous photograph. This is where brightness info is encoded in the foregoing picture. (N.B. You can also sharpen in this channel as you can in the L channel of Lab.) Now I'm going to transform this K channel in a series of HDR tone-mapping operations and put it back into the original photo.

Here's the K channel tone-mapped with Dynamic Photo HDR's Auto-Adaptive algorithm. Now let's put it back into the original photograph.

Here it is. Most of the bad shadows are gone and there are decent retrieved highlights. There are halos, though, because this is a local operator. You can see them better if you click on this photo and look under the cars and around the planters at the front of the picture-space. Still, we could definitely start from here and salvage the rest in Photoshop.
Let's try the 'eye-catching' variant next.

This is the K channel again. It's been tone-mapped with the 'eye-catching' tone mapper from Dynamic Photo HDR.

When it replaces the K-channel of the CMYK original it produces this (above). It's nearly identical to the previous except that the halos are under control here.

Here's the K channel tone mapped with the Halo tonemapper. The halos are pretty obvious and this seems the worst of the four variants.

Original with the halo tone-mapped K channel. The halos are quite obvious, particularly if you click through.

Dynamic Photo HDR's 'human eye' algorithm is supposed to 'mimic the rods and the cones of the human eye'. No. I don't know what that means either. This is the K-channel (above) as transformed by that algorithm.

Above see the finished product. Notice the amazingly unchanged color. HDR is an idea about brightness. It is NOT an idea about color. That becomes obvious when we isolate HDR tonemapping just to the brightness layer of the original (either L channel or K channel). I've said that the color layer doesn't change. That's true as long as you don't manipulate the C,M, or Y channels. If you made this CMYK from an RGB, say, and then perform these manipulations and convert it back to RGB then the RGB representation of the colors WILL change. But as long as you're transforming Lab to Lab or CMYK to CMYK then they will NOT change. You may like the garish colors of HDR transformation but this tutorial will allow you to hold the colors still if that's what you want.

I've compared Lab versions of the original to Lab versions where the L channel has been modified with HDR. The L channel varies significantly but the a and b channels (color) stay within 1/2 of 1% of each other (and even that is a rounding artifact). So I say that L or K will vary, but not a or b (or CMY) and this is what we expected. You can easily replicate this for yourselves.

I took the final 'human eye' version and produced this from it with some brightness map work and a touch of contrast. No sharpening. Compare it to the original. From a photographic disaster zone it has become a reasonable picture.

By the way, their 'orange-juicer' dome is famous.

(signed) Robert Consoli

Sony alpha 550 supports HDR directly

2009-09-02T13:25:00.000-07:00

Here's a review of the not-yet-released Sony alpha 550 that supports HDR directly.

http://www.luminous-landscape.com/reviews/cameras/a550-first.shtml

Apparently it takes two shots in rapid succession at differing ISO(? with, perhaps, fixed shutter and diaphragm). It then blends these with HDR tone-mapping algorithms in-camera and produces a jpg. See the sample shots at the link. The review doesn't mention how close together the two images are taken. It does say that you can vary the difference between the two shots by up to 3 f-stops.

The reviewer is always good but he falls down here. He thinks of HDR as a gadget and not as the most revolutionary development in photography in 150 years. But, still, the review will get you orientated.

HDR without changing the color. The L-channel of Lab.

2009-09-02T12:05:00.001-07:00

Can we use HDR without affecting the color? Yes. This tutorial shows how. The first picture is a badly underexposed photo taken at the old cemetery in San Juan Bautista. Oh my.
The real problem is the California brightness range. It's either catastrophically underexposed or it's about to burn out the highlights.

I converted it to Lab and then just captured the L channel and formed a new image with it. I left the color channels (a and b) alone.

I then ran that black and white image through Dynamic HDR selecting 'eye-catching' and using pseudo HDR (just the tone-mapping part). This result already shows a lot of promise with wide-open shadows and what appear to be non blown highlights.

I pasted that transformed black and white picture back onto the L layer of the original. Voila! But notice that the colors are not changed or distorted. It still needs some work but now, at least, that further work might actually succeed.

To finish off I made a brightness mask with a Gaussian filter and a radius of about 7. I then selected the dark areas and lightened them with levels. I pumped up the saturation a bit. No contrast enhancements or sharpening. There was never any color in the grass or the ground and so none was introduced. This is a long way from the original.

What's really curious about HDR is that shadows have lost all their power. Shadows are crucial for any kind of pictorial realism but they don't have to be inky to work. In HDR the shadows are there, the eye knows that they're there. They define the scene just fine - the way that they're supposed to. But they've lost their power to obscure.

One nice thing about this variant is that there appears to be full detail in the brightest and in the darkest parts of the picture. Compare to original.

HDR-processed Dresden photos

2009-09-02T11:44:00.000-07:00

I have a collection of Dresden photos that are processed in HDR. Here's a sample from the Opera House (the Semperoper). It is a grotesque head (Dionysus) used as an arch keystone. This was an HDR composite. It was generated with Dynamic HDR from a single image and then layered back on the original in Photoshop at an opacity of 50%. For the remainder see:
http://picasaweb.google.com/edificia00/DresdenHDR#

Squinchpix.com

2009-09-02T11:07:00.000-07:00

My own domain is

http://www.squinchpix.com

This domain is devoted to an archive of photographs of Europe. Featured there are pictures of buildings, art works, and archaeological sites from various parts of Europe. There are now about 5000 photographs on the site and every last one is completely tagged. From the HDR perspective search on that site for 'Dresden' or 'Dresden Brul*'. Many of those photographs are either HDR or HDR composites. I invite you to look at them and post your reactions.

Introduction to CIDER

2009-09-02T09:40:00.000-07:00

Welcome to CiderHouse!

The purpose of this blog is to explore every aspect of High Dynamic Range photography and, in particular, the use of HDR on the single image. [single-image HDR = SIHDR = Cider] We think that multiple image HDR has many technical advantages but if we restrict HDR tone-mapping to multiple images then many areas of photography will never benefit from these techniques. There will be, for example, no sports HDR, no NASCAR HDR, no candid HDR, etc. Single Image HDR has not yet been adequately explored and there is no convincing work-flow to deal with its unique problems. This site is devoted to examining those questions.
In the meantime the reader of this blog is also recommended to look at my Picasa Albums where topics on HDR are featured: http://picasaweb.google.com/edificia00