Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: Marshallarts on August 11, 2009, 04:50:16 pm
-
I've been asking this question on DPReview here (http://forums.dpreview.com/forums/read.asp?forum=1014&message=32654721) and have wasted days trying to understand my concern. To quote Luminous Landscape Merge to HDR in Photoshop CS2: A First Look (http://www.luminous-landscape.com/tutorials/hdr.shtml) "A more contemporary approach is to take multiple exposures of the same scene, varying (usually) just the shutter speed. Take a "normal" exposure, and then a few more at 1-2 stops over and under that point. Then, in Photoshop, blend these exposures, using the parts from each one that capture properly the part of the scene that you want.This can work very well, but to look convincing it needs to be done with some considerable skill, and usually requires quite a bit of work with masks and brushes. My blending tutorial looks at this technique. Also popular, when multiple exposures aren't possible (for example when there is movement in the scene), is to process the RAW file twice, once for the highlights and once for the shadows. This can't, of course, extract information that isn't in the actual file, but it can do a better job than using the usual raw processing tools currently available. The same blending techniques as are used for merging multiple exposures are also used here, with similar issues arising."
I'm trying to understand why the usual raw processing tools can't do as decent a job, since after all the processed RAW files to be combined were created in the usual raw processor. What makes the multiple processed raw files better when combined? Should the usual raw tools know how to isolate the regions that need adjusting in a whole image? I completely understand the principle of exposure blending. My questions is more about what happens during usual processing that makes it so it's not as good of a way to do it.
Also, I've been using Photomatrix to create HDR from single raw images. Does this automatically create the different exposures necessary? I also use the program for true HDR photos from multiple exposures on location. But I've never learned how to properly adjust single RAW images to prepare them for blending manually (not using Photomatrix). Does one only adjust the exposure slider and not touch Blacks, recovery, or fill when making the separate exposures?
Also, when combining photos, especially in generating HDR from bracketed exposures, is it best to not adjust any other perimeters and save image editing to after it is fully created? (except for exposing once for highlights and once for shadows in the single RAW file method)
Here is my first attempt at HDR from bracketed exposures (-2EV, 0, +2EV). I know many people don't like this effect. I'm working on making it look more natural. But this was definitely a situation where dynamic range was very high. I have another question for checking when is appropriate to use HDR since it's pointless when DR isn't high. Looking at my histogram, if I'm still getting a lot on both the highlight edge and the shadow edge wouldn't that indicate bracketing and exposure blending is necessary if I want to include those lost regions to clipping? thanks!
-
True 32bit HDR has to come from more than one unique file, since one file has only 12~16bits depending on camera/back used. The same file processed two or three times with different parameters is still the same bits.
-
True 32bit HDR has to come from more than one unique file, since one file has only 12~16bits depending on camera/back used. The same file processed two or three times with different parameters is still the same bits.
True, but it may be easier to adjust the high values and low values separately and then combine the two. This seems to be what Mark Segal is doing in today's (August 12, 2009) essay. What's New.
Read it and see if it answers any of your questions.
-
I'm trying to understand why the usual raw processing tools can't do as decent a job, since after all the processed RAW files to be combined were created in the usual raw processor. What makes the multiple processed raw files better when combined? Should the usual raw tools know how to isolate the regions that need adjusting in a whole image?
How would the raw processor know what you want to achieve?
The problem with the single pass raw processing is often just what you pointed at: different parts of the image require different adjustments. How to develop which part may be straightforward, but it may be complicated, subject to taste, goal, scenery, etc.
I just finished a pano (I shot it for over two years ago in Zion and today was the time to process it). It is a difficult location because of the valley between the high mountains. If I develop the frames for the foliage, the rock illuminated directly by the sun gets blown (the red is particularly a problem). If I develop it for the rock, the trees are far too dark; see the attachments. So, I developed the same frames twice. Blending was simple: I loaded them as layers and added a vertically graduated mask to one of them. The result is Zion pano (http://www.panopeeper.com/panorama/Zion_2.jpg) (the sky comes from different frames, which were exposed lower, but I did not use any HDR technique to put them together).
The blending can be much more complicated in other cases, it may require tone mapping as well.
-
The problem with the single pass raw processing is often just what you pointed at: different parts of the image require different adjustments.
Please forgive my misunderstanding. This seems to hint at what people are missing in what I'm saying, maybe I'm the one confused. You stated the examples between the rocks and trees and how you need to process for both which have different exposure values, and then blended them. But isn't that the point with RAW processing, that one can distinguish between exposure values and manipulate them independently? That I can raise my fill light in the case of the darker trees and independently adjust the exposure for the brighter rocks. I understand that the results from multiple processing of a single image speaks for itself, but I'm finding it hard to understand how since it seems a RAW editor with specific perimeters for different portions of your EV exist (ie. Recovery, Exposure, Fill Light, Blacks). What about processing multiple versions of a picture makes it better?
-
True, but it may be easier to adjust the high values and low values separately and then combine the two. This seems to be what Mark Segal is doing in today's (August 12, 2009) essay. What's New.
Excellent essay!! Thanks!
Also, to jjlphoto, I hear what you're saying. In my original thread at DPReview my questioning sparked a tirade about HDR and how from a single image you can't create anymore dynamic range than was originally captured--an insightful reminder but not relevant to my question. Then came the comments about HDR looking fake and also how it's pointless to attempt HDR when your scene's dynamic range isn't high to begin with (if your scene is only 8 stops and within the dynamic range capable of your sensor). Also a good reminder but sidestepping my questions.
Despite all the opinions people still continue to use these techniques and I'm always trying to understand the best methods to learn which techniques are most correct. I've heard for awhile about people processing multiple versions of a single image in a form of HDR. Truly this is not the same as multiple bracketed exposures so ignore the semantics. Yes no additional DR is created but people still do this and call it what you will is still a form a dynamic range manipulation. Many people don't like the way real HDR looks. I'm interested in exploring this more myself. I'm sure my photo looks very poor to someone with more HDR experience.
Maybe I should attempt it with the single RAW image method described in the essay. Is that typically how this method works--exposing for specific areas in the photo then blending them? I think my confusion at first was how one goes about this versus the more common HDR methods of simply combining evenly bracketed photographer in HDR software.
Maybe a part of my confusion is in my approach to begin with. Looking at a histogram if there are areas peaking/clipping on either side of my chart, I blindly assume bracketing my exposure will allow for those areas to be better exposed. Honestly I never concerned myself with where in the image those areas are peaking. Is this a poor way of deciding something requires HDR bracketing?
-
Please forgive my misunderstanding. This seems to hint at what people are missing in what I'm saying, maybe I'm the one confused. You stated the examples between the rocks and trees and how you need to process for both which have different exposure values, and then blended them. But isn't that the point with RAW processing, that one can distinguish between exposure values and manipulate them independently? That I can raise my fill light in the case of the darker trees and independently adjust the exposure for the brighter rocks.
The problem is that the controls in RAW converter are for the most part global to the whole image. Lightroom and ACR do have some localized edits, but they're pretty limited (for instance no fill light or recovery sliders, to use your example). What happens is that if you try to boost the shadows and recover the highlights in a single RAW conversion, you end up with a dull and lifeless image. A pleasing image needs contrast, both local and global, and if you just linearly compress a high-contrast scene to fit within a reduced dynamic range the result is pretty ugly.
By processing the image twice, you can get target your RAW conversions for specific areas of the image to get a more pleasing result.
-
You stated the examples between the rocks and trees and how you need to process for both which have different exposure values, and then blended them. But isn't that the point with RAW processing, that one can distinguish between exposure values and manipulate them independently?
This is not always so straighforward. I can adjust the mapping in the raw conversion (fill light, recovery, curves, etc.), but that will effect the entire image. However, that is not working in the example I showed above: parts of the rocky areas are quite dark, and they are supposed to remain so. Lifting the foliage out of the shadows ruins the contrast on the rock surface.
ADDED:
I just see Jeff posted exactly what I wanted to answer. Sorry for the repetition, I was typing too slowly.
-
I'm always trying to understand the best methods to learn which techniques are most correct.
The most correct technique is the one that gives the best image, period!
In many cases, an efficient DR mitigation can be made from one raw exposure in LR/ACR, by the simple use of fill light/blacks (and to the expense of noise in the shadows).
However, if eg extensive highlight recovery is needed with the tone curve (as in the excellent Mark Segal's essay), then you hit the fact that the tone curve is not a "masking" tool (ie it affects all pixels of the image), and you have to rely on masking in PS or other pixel-level editing app (until LR's brush can do tone curve adjustments).
(edit : we've said the same thing 3 times, sorry for the echo)
Maybe I should attempt it with the single RAW image method described in the essay.
If shadow noise is not a crucial problem, one exposure at base ISO taken for the highlights often does the trick.
-
To say it a fourth time, but different:
a pixel with for instance 50/50/50% R/G/B somewhere in the shadow area needs to move in a different direction (eg lighter) than another 50/50/50% pixel in the lighter area (eg. needs to become darker). A raw converter cannot do both in one pass since it only changes similar toned pixels in the same direction globally.
-
a pixel with for instance 50/50/50% R/G/B somewhere in the shadow area needs to move in a different direction (eg lighter) than another 50/50/50% pixel in the lighter area (eg. needs to become darker). A raw converter cannot do both in one pass since it only changes similar toned pixels in the same direction globally.
That's a good description of the problem!
However (disclaimer : my answer only does cut hairs in quarters, as we say in French) keep in mind that some tools in some raw converters already do that based on the surroundings pixels : eg, Fill Light will probably lighten it if there are only darker pixels around, not if there are only lighter pixels.
There are already mask-based tools. The problem is when you hit their limit!
-
I'm trying to understand why the usual raw processing tools can't do as decent a job, since after all the processed RAW files to be combined were created in the usual raw processor. What makes the multiple processed raw files better when combined? Should the usual raw tools know how to isolate the regions that need adjusting in a whole image? I completely understand the principle of exposure blending. My questions is more about what happens during usual processing that makes it so it's not as good of a way to do it.
It's a question of which software tool or method is more adequate to ease the task. The information contained in the RAW file (or files) is the same for all of them:
- A RAW developer with shadows/highlight and so forth sliders and options
- Two (or more versions) of the same RAW developed at different exposure values and manually blended in PS layers
- A specific tone mapping tool designed to enhance local contrast while reducing global contrast like Photomatix
...
Using 2 (or more) versions of the same RAW file developed at different exposures allows to easily obtain the desired result, and meanwhile there is no other clear winner option this will be one of the methods to use.
Let me show you a test I did some days ago to find out how well Photomatix can blend RAW files with a big exposure gap (4EV): it showed to be a total dissapointment!. A simple algorithm (implemented both in Zero Noise (http://www.guillermoluijk.com/tutorial/zeronoise/index.htm) or through this simple tutorial (http://jtrujillo.net/qpix/) in PS) consisting in selecting the pixels with the highest exposure performs much better than Photomatix's information blending:
(http://www.guillermoluijk.com/tutorial/hdr/resultadolite10.jpg)
(http://img19.imageshack.us/img19/8554/comp1g.jpg)
(http://img19.imageshack.us/img19/4519/comp2o.jpg)
(http://img190.imageshack.us/img190/4452/comp3p.jpg)
If not even a specialized HDR blending software manages to achieve better results than a proper PS blending, why should we forget manual solutions and think a software tool should always be preferred?
Regards
-
What about processing multiple versions of a picture makes it better?
My tutorial on this very topic was published on this site yesterday. It's not HDR but it works. How and to what extent would vary from image to image.
Mark
-
Wow! I feel like I just got tutored by some real pros! Thank you all for coming to the rescue. Your answers really helped me better understand this. They did raise some questions that I would like to independently ask. It seems you all noticed you schooled me with the same answer but each way told was different. This was helpful showing me different ways of explaining it in that I see a bigger picture, but I do have a few specifics I wouldn't mind cleared up.
The problem is that the controls in RAW converter are for the most part global to the whole image. Lightroom and ACR do have some localized edits, but they're pretty limited (for instance no fill light or recovery sliders, to use your example). What happens is that if you try to boost the shadows and recover the highlights in a single RAW conversion, you end up with a dull and lifeless image. A pleasing image needs contrast, both local and global...
but LR does have fill light and recovery sliders...? Perhaps you didn't mean to write "no" and was using fill light & recovery to explain your later explanation that when you adjust one you effect the other--making the image dull and lifeless. This struck me like lighting as I finally realized how such global control would effect the curve. Indeed we want to keep the contrast, and local control makes sense!
This is not always so straighforward. I can adjust the mapping in the raw conversion (fill light, recovery, curves, etc.), but that will effect the entire image. However, that is not working in the example I showed above: parts of the rocky areas are quite dark, and they are supposed to remain so. Lifting the foliage out of the shadows ruins the contrast on the rock surface.
another great explanation... Then I wonder "when creating HDR (or WDR, let's not focus on HDR) how does it know which part of the scene is rock or foliage without complete manual control (ie. masking in photoshop)".. This is partially answered by more great replies I will quote below but I'd like your take on it. Unless it's part of a scene you can easily mask (i.e. the sky) how can you lift foliage that's mixed in with rocks you wish to stay the same? Let me comment on the other replies that discuss this but your specific example seems particularly unique and difficult and I'd like your take on it.
...In many cases, an efficient DR mitigation can be made from one raw exposure in LR/ACR, by the simple use of fill light/blacks (and to the expense of noise in the shadows).
However, if eg extensive highlight recovery is needed with the tone curve (as in the excellent Mark Segal's essay), then you hit the fact that the tone curve is not a "masking" tool (ie it affects all pixels of the image), and you have to rely on masking in PS or other pixel-level editing app (until LR's brush can do tone curve adjustments)....
If shadow noise is not a crucial problem, one exposure at base ISO taken for the highlights often does the trick.
this is where I realized how by masking using 2 differently developed photos really makes sense. And Mark's essay was a perfect example with that blown sky. I completely see how by recovering the sky you would also recover the building and loose contrast there. But this is easy to mask as it's an isolated part of the scene. In other cases the two could be mixed (like Panopeer's).
I know LR can do curve adjustments, what do you mean by LR's brush do tone curve adjustments? Also I'm having trouble understanding your last sentence. "...one exposure at base ISA.." Do you mean process the RAW file bringing down exposure for the sky? and how would that effect shadows? or do you mean simply take the photo originally at base ISO unless you're worried about bringing up the shadows? Could you please explain that? (I didn't think you're talking about blending an HDR because we're talking about one exposure)
To say it a fourth time, but different:
a pixel with for instance 50/50/50% R/G/B somewhere in the shadow area needs to move in a different direction (eg lighter) than another 50/50/50% pixel in the lighter area (eg. needs to become darker). A raw converter cannot do both in one pass since it only changes similar toned pixels in the same direction globally.
Great explanation also. But having two similarly shaded/colored pixels in two other regions would require the editor to know the difference between them.. Answered below but brings other questions.
That's a good description of the problem! However... keep in mind that some tools in some raw converters already do that based on the surroundings pixels : eg, Fill Light will probably lighten it if there are only darker pixels around, not if there are only lighter pixels.
There are already mask-based tools. The problem is when you hit their limit!
I did not realize this! How can I find out which program does this and what the limit is. Essentially this could mean processing one version would be enough so long as these limits work for the image. What are some mask-based tools I could learn about!
It's a question of which software tool or method is more adequate to ease the task. The information contained in the RAW file (or files) is the same for all of them:
- A RAW developer with shadows/highlight and so forth sliders and options
- Two (or more versions) of the same RAW developed at different exposure values and manually blended in PS layers
- A specific tone mapping tool designed to enhance local contrast while reducing global contrast like Photomatix
...
Using 2 (or more) versions of the same RAW file developed at different exposures allows to easily obtain the desired result, and meanwhile there is no other clear winner option this will be one of the methods to use....
If not even a specialized HDR blending software manages to achieve better results than a proper PS blending, why should we forget manual solutions and think a software tool should always be preferred?
Great examples, thank you for sharing! In fact after posting yesterday I spent more time better understanding Photomatix and learning that in generating HDR the choices in processing yield far different results. Naively I thought HDR was one thing, as if a tool to be used globally. Now I'm really seeing the bigger picture. How in Photomatix you can go the tone compressor route or the details enhancer route. But how you can also simply go the exposure blending route....and then the manual masking route like in Mark's essay. I'd like to learn your method and will look at that tutorial. Bottom line, it depends on how you want it to look.
If I've learned correctly, the tone mapping software route like Photomatix would be good in global situations where you want enhanced DR without flattening out the curve...!! ding, ding, ding---is that correct!? But in other situations where global adjustments are unwanted this may not be the correct route. In those cases it seems manual blending (via masking) may be best..
Under what situations have you learned each method should be used?
My tutorial on this very topic was published on this site yesterday. It's not HDR but it works. How and to what extent would vary from image to image.
Mark
...and what an excellent essay it was! Thank you so much for your tutorial!
I know this reply is long but I can't be more happy with any thread I've ever started! This has really helped me a lot. I really look forward to hearing back from my last questions! Then I really think I'll have a well rounded understanding! (This is now my preferred forum!)
-
but LR does have fill light and recovery sliders...? Perhaps you didn't mean to write "no" and was using fill light & recovery to explain your later explanation that when you adjust one you effect the other--making the image dull and lifeless. This struck me like lighting as I finally realized how such global control would effect the curve. Indeed we want to keep the contrast, and local control makes sense!
It has them, but they're global. You can't limit the effect to just part of the image.
-
The problem is that the controls in RAW converter are for the most part global to the whole image. Lightroom and ACR do have some localized edits, but they're pretty limited (for instance no fill light or recovery sliders, to use your example).
Other than that, I find the highlight/shadow capability of C1 Pro to be doing the most convincing job at extracting detail from shadows without making the image look flat. I believe that their algo is not just global.
Globally, the simple answer to the original poster's question is that DSLR do in fact have too much DR for many scenes, and some form of tone mapping is required to map these informations to a small bit space like the one of our screens or papers. Most raw converters are poor tone mappers with the notable exception of those developped specifically for MF cameras (which explains why C1 Pro works so well on DSLRs as well).
Cheers,
Bernard
-
It has them, but they're global. You can't limit the effect to just part of the image.
Jeff, yes you are correct thery are global, and it would be nice to see in a future release of LR/ACR that more of these controls become available under the Adjustment Brush, which as you know allows localized adjustment of tone and colour For now, we can use the Exposure, Brightness and Contrast controls and the brush density to achieve some of these effects locally - not ideal but still often quite useful.
-
Other than that, I find the highlight/shadow capability of C1 Pro to be doing the most convincing job at extracting detail from shadows without making the image look flat. I believe that their algo is not just global.
Globally, the simple answer to the original poster's question is that DSLR do in fact have too much DR for many scenes, and some form of tone mapping is required to map these informations to a small bit space like the one of our screens or papers. Most raw converters are poor tone mappers with the notable exception of those developped specifically for MF cameras (which explains why C1 Pro works so well on DSLRs as well).
Cheers,
Bernard
Bernard, perhaps you meant that DSLRs do not have enough DR for many scenes.
Mark
-
If I've learned correctly, the tone mapping software route like Photomatix would be good in global situations where you want enhanced DR without flattening out the curve...!! ding, ding, ding---is that correct!? But in other situations where global adjustments are unwanted this may not be the correct route. In those cases it seems manual blending (via masking) may be best..
Under what situations have you learned each method should be used?
In my particular case I always use the same method: properly blend the captured information with pixel selection (ZN or the tutorial I linked which is basically the same), and then use curves with layer masks to manually tone map that information in PS.
The only automated software that I have found providing fairly good results (natural looking, no or very few unexpected artifacts) for HDR situations are the Enfuse/Tufuse (http://www.tawbaware.com/tufuse.htm) implementations of this Exposure Fusion (http://research.edm.uhasselt.be/~tmertens/papers/exposure_fusion_reduced.pdf) algorithm, and still it can't match a proper manual pp in PS.
I just did the Photomatix tests to find out how bad the software was at fusing the information fed into it. Photomatix's results depend strongly on how many images you provide, even if they are just copies of the same original RAW file with different exposure. From the point of view of information fusion, this is unacceptable.
Regards
-
It has them, but they're global. You can't limit the effect to just part of the image.
Probably a lack of controls (e.g. tonal width beside of strength).
Fill light as well as the Recovery slider seem to be selection based.
So strictly seen, the term “global” might be wrong, even though that’s what we see.
For me, Fill Light typically works well enough, whereas Recovery can be easily outperformed by “HDR blending (http://luminous-landscape.com/forum/index.php?showtopic=35412)”.
It’s not only a lack of controls, Recovery seems to be placed too late in the processing chain so that compression of highlight details and loss of color saturation beneath the upper shoulder of the usual tone curve applied can’t be restored or better prevented adequately. Highlight details and colors which ARE in the Raw image are wasted. IMO.
The question of the threadopener is certainly valid – and should be addressed to the software engineers.
Peter
--
-
DPL:
My understanding of the Recovery tool in LR/ACR is that it builds back detail by calculating replacement tones for blown primaries as long as at least one of the primaries is still present in the area to be "recovered".
I'd like to hear more about your proposition that "Recovery seems to be placed too late in the processing chain so that compression of highlight details and loss of color saturation beneath the upper shoulder of the usual tone curve applied can’t be restored or better prevented adequately. Highlight details and colors which ARE in the Raw image are wasted."
I don't think Adobe has published details about the processing sequence of LR/ACR instructions when exporting an adjusted raw image to Photoshop, nor have they published the details of the algorithms which would allows one to be confident of this statement. I am also having trouble understanding exactly what this statement means: what do you mean by "usual tone curve applied"? What is it that can't be "restored or better prevented adequately? And where in the processing cycle are you thinking of: ACR or PS?
I also have trouble believing, based on my understanding and experience, how the Recovery tool can "waste" highlight detail and colors. Those always remain in the raw image. It is of course possible to overdo a Recovery Adjustment and emerge with overly dulled highlights, but stuff like that can happen with any wrongly applied procedure.
-
DPL:
My understanding of the Recovery tool in LR/ACR is that it builds back detail by calculating replacement tones for blown primaries as long as at least one of the primaries is still present in the area to be "recovered".
Mark:
The problem is easy to see, but let’s support it by some numbers. Starting point is a "linear rendition", means all relevant controls zero in ACR, no curve applied. Then, Brightness & Contrast as well as Recovery are raised. The file was handled as Smart Object in Photoshop in order to have HSB readings (referring to ProPhoto RGB). Two sampling points were placed in the sky. It’s a sunset sky, but the channels are non-blown:
Linear rendition:
HSB(1)= 62°, 43%, 97%
HSB(2)= 44°, 45%, 91%
Brightness 50 & Contrast 25:
HSB(1)= 62°, 22%, 99%
HSB(2)= 43°, 27%, 98%
Brightness 50 & Contrast 25 and Recovery 50:
HSB(1)= 47°, 15%, 99%
HSB(2)= 41°, 24%, 90%
The S-curve which is de facto introduced by Brightness 50 & Contrast 25 leads to a drop of color saturation in the highlights, beneath the upper shoulder of the curve. It’s the typical behavior of RGB curves, even though the ACR implementation was reported to include a hue-lock. Once such damage of highlight colors is done, it can’t be undone in general and in particular not by the Recovery slider. It’s too late. Even worse, Recovery further reduces saturation and also twists the hue.
Alternatively, in Photoshop we could start with the "linear rendition" again and add a tone curve to be equipped with an inverted Luminosity mask, thus preventing this effect ab initio. Or, another option is to restore the information by means of "true recovery / HDR blending (http://luminous-landscape.com/forum/index.php?showtopic=35412)".
Many workarounds for something that could/should work in the Raw converter based on one single Raw file.
Peter
--
-
Fill light as well as the Recovery slider seem to be selection based.
I don't think they're selection based. Their effects are certainly non-linear across the tonal range; but as you use higher amounts the effect gets much broader. Slide the Recovery slider to 100% and it's pretty clear it's not just affecting the highlights.
-
Mark:
The problem is easy to see, but let’s support it by some numbers. Starting point is a "linear rendition", means all relevant controls zero in ACR, no curve applied. Then, Brightness & Contrast as well as Recovery are raised. The file was handled as Smart Object in Photoshop in order to have HSB readings (referring to ProPhoto RGB). Two sampling points were placed in the sky. It’s a sunset sky, but the channels are non-blown:
Linear rendition:
HSB(1)= 62°, 43%, 97%
HSB(2)= 44°, 45%, 91%
Brightness 50 & Contrast 25:
HSB(1)= 62°, 22%, 99%
HSB(2)= 43°, 27%, 98%
Brightness 50 & Contrast 25 and Recovery 50:
HSB(1)= 47°, 15%, 99%
HSB(2)= 41°, 24%, 90%
The S-curve which is de facto introduced by Brightness 50 & Contrast 25 leads to a drop of color saturation in the highlights, beneath the upper shoulder of the curve. It’s the typical behavior of RGB curves, even though the ACR implementation was reported to include a hue-lock. Once such damage of highlight colors is done, it can’t be undone in general and in particular not by the Recovery slider. It’s too late. Even worse, Recovery further reduces saturation and also twists the hue.
Alternatively, in Photoshop we could start with the "linear rendition" again and add a tone curve to be equipped with an inverted Luminosity mask, thus preventing this effect ab initio. Or, another option is to restore the information by means of "true recovery / HDR blending (http://luminous-landscape.com/forum/index.php?showtopic=35412)".
Many workarounds for something that could/should work in the Raw converter based on one single Raw file.
Peter
--
Peter,
Reverting to your original points about the difficulties with Recovery, suppose you don't introduce an S-Curve with Brightness and Contrast (leave them at 0) and use Recovery in one iternation, no recovery in the other? It's not obvious that the demonstration you provide here proves your original points, or perhaps I'm a bit dense this morning.
-
Slide the Recovery slider to 100% and it's pretty clear it's not just affecting the highlights.
This is correct depending on how many tonal levels you include in "highlights", and most likely by design. I'm not sure exactly what is meant by "selection-based" in either your post or Peter's or whether it matters to the discussion. For any of these controls to work as intended, somehow or other they must be "selection-based" in the sense that they operate on select sets of tonal values in the image, with overlaps to provide for smooth tonal transitions between lighter and darker tones. So it seems to me that your observation quoted here means that as the Recovery is increased, the number of levels it impacts increases. This would correspond with one's observation of what occurs as we implement it. While there are other techniques in Photoshop which have different and perhaps stronger impacts in respect of revealing information from within the file, none of this would seem to support a contention that the Recovery function in LR/ACR has the kind of fatal negative effects being discussed in this thread.
-
I don't think they're selection based. Their effects are certainly non-linear across the tonal range; but as you use higher amounts the effect gets much broader. Slide the Recovery slider to 100% and it's pretty clear it's not just affecting the highlights.
That is called a changing selection threshold
-
I don't think they're selection based. Their effects are certainly non-linear across the tonal range; but as you use higher amounts the effect gets much broader. Slide the Recovery slider to 100% and it's pretty clear it's not just affecting the highlights.
So when we reverse-engineer Fill Light or Recovery, every change per pixel can expressed by one single global equation? Let’s take something like a polygon approximation to cover complex functions including any non-linearity. I’ve yet to see this. Please show me.
Slightly philosophical: colors are 3D and so are blending operations. For reference, if we look at a Curve + inverted Luminosity mask, or the S/H tool in Photoshop, the basic three dimensions of freedom now are:
amount (strength) – opacity of blending.
tonal width – contrast of the mask.
and radius – blur applied on the mask.
Interestingly, it’s the Radius which definitively brings us out of the range of any A to B equation per pixel (as long as we don’t include an analysis of environmental pixel).
Assuming that Fill Light and Recovery are just (bundled to) one single slider, respectively, can not mean that the other parameter don’t exist. It’s just a lack of controls.
Peter
--
-
Peter,
Reverting to your original points about the difficulties with Recovery, suppose you don't introduce an S-Curve with Brightness and Contrast (leave them at 0) and use Recovery in one iternation, no recovery in the other? It's not obvious that the demonstration you provide here proves your original points, or perhaps I'm a bit dense this morning.
To continue with above example, if we just raise Recovery while leaving Brightness and Contrast at zero, highlight saturation is reduced again compared to the initial "linear rendition". Also, note the hue twist which is more than 10° with the first measuring point.
Recovery 50:
HSB(1)= 48°, 33%, 94%
HSB(2)= 42°, 35%, 73%
Means, the approach to keep Brightness and Contrast low while raising Recovery + Exposure does not really help to preserve highlight color.
Mark, - at the end of the day there are many images, as far as I can tell, which profit from "HDR blending" from one single Raw file (see your own tutorial) and I’d still blame this to the insufficient mechanics of the Recovery slider. It’s both, a lack of controls (missing a Tonal Width slider) and the algorithm itself which does not preserve or restore the original color information.
Peter
--
-
Peter,
The logic of the Recovery function suggests to me that one should NOT expect hue neutrality here, because it is building information for the missing (clipped) channels. Likewise for saturation, where the change you note is small in any case, and of course B should be reduced, as your measurement shows.
In my tutorial I used a combination of Recovery, Highlight adjustment in the LR Tone Curve and then further tweaking in PS. I do fully agree that luminosity blending in PS can produce value-added here, but I don't see how this somehow means there is something inherently wrong with the Recovery function in LR/ACR in respect of what it was designed to do.
Mark
-
I was just learning more about the Shadows/Highlights adjustment in Photoshop and it appears this has local control.
The manual states: "The Shadow/Highlight command does not simply lighten or darken an image; it lightens or darkens based on the surrounding pixels (local neighborhood) in the shadows or highlights. For this reason, there are separate controls of the shadows and the highlights."
For those who described their manual procedure to create HDR (for lack of better term) from one photo in Photoshop, did you know this adjustment tool existed!?
-
For those who described their manual procedure to create HDR (for lack of better term) from one photo in Photoshop, did you know this adjustment tool existed!?
Yes, from the day it was put on the market.
-
Yes, from the day it was put on the market.
Forgive my generalized question.. I'm sure you guys know the tool. It was very briefly mentioned as S/H in a reply earlier in this thread.
But since this thread has spent so much time discussing either doing HDR fully manually in photoshop or automated in other HDR programs like Photomatix, I was puzzled why no one mentioned using the Shadow/Highlights adjustment perimeter in Photoshop.
I know Photoshop does automated merge to HDR which in my limited experience produces far worse images than Photomatix (though on the more realistic side the results IMO are still worse). But it seems based on responses here and your own tutorial, Mark, that people don't generally use Photoshop's automated tool. Of course the automated tool is for true bracketed exposure HDR and not the pseudo-HDR I began this thread about--however one still could process multiple exposures from one RAW file and use it as such.
But for true pseudo-HRD (from one image) I would love to hear how you feel about Photoshop's Shadows/Highlights tool since it does appear to do the local adjustments you all claim are necessary for good HDR-- and there are perimeters to fine tune it.
I sincerely want to understand this better. I am a very confident shooter, it's what comes naturally to me. I know exactly what I want and how to get it. Now with digital processing I've finally realized I'm limiting myself as a photographer by not understanding these techniques. However, the endless possibilities and countless ways to go about performing the exact same tasks (even within one program!), understanding the differences has been very frustrating, confusing, and ultimately has taken a big toll on my confidence. I need to spend more time shooting and less learning all these things I never worried myself about before. But there seems to be so many opinions and misdirection as to which way to go----I just want to know the right way to do it.
What is the downside of using the Shadows/Highlights tool (i.e. In discussing HDR and Photoshop, why hasn't anyone mentioned it being a useful tool if everyone knows about it?)
-
Now with digital processing I've finally realized I'm limiting myself as a photographer by not understanding these techniques. However, the endless possibilities and countless ways to go about performing the exact same tasks (even within one program!), understanding the differences has been very frustrating, confusing, and ultimately has taken a big toll on my confidence.
----I just want to know the right way to do it.
I think it isn't worth getting frustrated for not understanding how a particular tool works, but just if it serves you well. After all they are black boxes that only the developers know exactly how operate. You are allowed to try them, master them looking at the results they produce, and use them if you feel they fit your needs. More and more tools and programs will always be, and there is not a right way to do things, there are many.
I will give you my approach in case it is useful to you. It's clear that to tone map the shadows and highlights of a high dynamic range scene we need to do local and global adjustments: global to fit the entire dynamic range of the original scene in a low dynamic range image, and local to enhance contrast locally so that images don't look dull.
To achieve that you can use black boxes (tools made by someone else that give you control through sliders):
- Photomatix
- Photoshop HDR Merge
- Photoshop Shadows/Highlights
- LR/ACR Shadow and Highlight recovery
...
and you can either use tools that allow you to see exactly what you are doing:
- Manual merging layers in PS with 2 different exposure RAW developments
- Local level adjustments with mask layers using the Levels/Curves tool
...
I chose the second option because even if it could _seem_ that some of the black boxes yield a good and quick result, you never know how it worked. So if one day they produce a bad result you will not know how to fix it.
In particular I start from a single neutral RAW development (MarkDS called that 'linear rendition'), and use curves with optional mask layers to lift the shadows while preserving the highlights. Don't need more that adjustment layers to go from the dull and underexposed original image to the final image without leaving PS:
(http://img18.imageshack.us/img18/8696/samplesy.jpg)
You can download these 2 images with the appropiate layers from: capasauditorio.tif (http://www.guillermoluijk.com/misc/capasauditorio.tif), capas2.tif (http://www.guillermoluijk.com/misc/capas2.tif). This is just my way, probably not the best, not the only one.
BR
-
This is just my way, probably not the best, not the only one.
BR
Thank you for your help! Those samples are amazing. Opening them in photoshop I see the layers and exactly what you did. It really helps me understand the amount and kind of work you do to your images. To be honest your skills are well beyond mine (as much as you humbly say "probably not the best"), and so I still have a long way to get to your level. Maybe I'm in over-my-head seeking advanced level advice from experts with these questions, but I don't know where else to start---I want to learn these tools and techniques. Do you recommend something that could get me up to speed with your level of understanding?
I am a little out of my league in what I'm trying to learn, but so often I find it's best to ask the pros and avoid learning bad habits.
I've been reading up about your Zero Noise program. But I have a Mac and it seems all the programs you have are for PCs. It looks amazing though!
-
Thank you for your help! Those samples are amazing. Opening them in photoshop I see the layers and exactly what you did. It really helps me understand the amount and kind of work you do to your images.
As you can see I only use curves. Curves are a very powerful and flexible tool (in fact many of the 'black boxes' found on imaging programs are just hiding curves under their sliders) and I find them very intuitive, but for many people aren't. Consider them just another of the many possibilities to try.
I have no idea about English literature for image processing, but there must be a lot stuff. Perhaps someone here could recommend a good book oriented to teach film photographers about digital imaging, showing how they are to process their images to achieve the same things that previosly were done in the wet darkroom. That's why I wouldn't recommend a Photoshop manual, but a book specifically written for photographers.
-
Perhaps someone here could recommend a good book oriented to teach film photographers about digital imaging, showing how they are to process their images to achieve the same things that previosly were done in the wet darkroom. That's why I wouldn't recommend a Photoshop manual, but a book specifically written for photographers.
Yes, I would highly recommend two books: Ben Willmore's Photoshop Studio Techniques, and Katrin Eismann's Creative Digital Darkroom.
-
Forgive my generalized question.. I'm sure you guys know the tool. It was very briefly mentioned as S/H in a reply earlier in this thread.
But since this thread has spent so much time discussing either doing HDR fully manually in photoshop or automated in other HDR programs like Photomatix, I was puzzled why no one mentioned using the Shadow/Highlights adjustment perimeter in Photoshop.
I know Photoshop does automated merge to HDR which in my limited experience produces far worse images than Photomatix (though on the more realistic side the results IMO are still worse). But it seems based on responses here and your own tutorial, Mark, that people don't generally use Photoshop's automated tool. Of course the automated tool is for true bracketed exposure HDR and not the pseudo-HDR I began this thread about--however one still could process multiple exposures from one RAW file and use it as such.
But for true pseudo-HRD (from one image) I would love to hear how you feel about Photoshop's Shadows/Highlights tool since it does appear to do the local adjustments you all claim are necessary for good HDR-- and there are perimeters to fine tune it.
I sincerely want to understand this better. I am a very confident shooter, it's what comes naturally to me. I know exactly what I want and how to get it. Now with digital processing I've finally realized I'm limiting myself as a photographer by not understanding these techniques. However, the endless possibilities and countless ways to go about performing the exact same tasks (even within one program!), understanding the differences has been very frustrating, confusing, and ultimately has taken a big toll on my confidence. I need to spend more time shooting and less learning all these things I never worried myself about before. But there seems to be so many opinions and misdirection as to which way to go----I just want to know the right way to do it.
What is the downside of using the Shadows/Highlights tool (i.e. In discussing HDR and Photoshop, why hasn't anyone mentioned it being a useful tool if everyone knows about it?)
I almost never succeed in getting the effect I want out of Photoshop's Shadow/Highlight function, even with all the bells and whistles it comes with. It has a tendancy to turn tonal values into grey mush, and it seems less effective than good use of Curves and Blend Modes for extracting detail, increasing shadow detail and taming highlights, while maintaining good contrast in those areas at the same time. There are of course images on which it may suffice and work very well. This isn't a business where one size fits all, as you have most likely come to appreciate; so I would not dismiss S/H out of hand - like all these functions, it has its place, but for much of what I do I find it less satisfactory than other techniques.
Yes, you are right, there are endless possibilities and countless ways of achieving essentially the same objectives, all producing either similar or differing results by a little of a lot. Photoshop/ACR/LR are very deep programs. But this isn't a reason to be frustrated or to lose confidence. On the contrary, treat it as a challenge and a learning experience. No-one serious about using these applications ever stops learning.
-
I know LR can do curve adjustments, what do you mean by LR's brush do tone curve adjustments?
It's because in LR, tone curve adjustments are global. As said here, this can be less than adequate for some tasks involving a good deal of tone mapping.
I personally feel that the ability to apply locally, with the brush, a curve adjustment, even if it wouldn't have all the power of PS's blending mode, could already do much of a good job in these diffults cases we're talking about.
Also I'm having trouble understanding your last sentence. "...one exposure at base ISA.." Do you mean process the RAW file bringing down exposure for the sky? and how would that effect shadows? or do you mean simply take the photo originally at base ISO unless you're worried about bringing up the shadows? Could you please explain that? (I didn't think you're talking about blending an HDR because we're talking about one exposure)
When facing a high dynamic scene, I generally take one exposure exposed for the highlight (that's what Guillermo Luijk did in the two examples above, if I'm not mistaken).
With such an exposure, I don't need much Recovery as the HL are already where they should (or not far). That's important because recovery is a problematic task, often involving some kind of guessing from the processing algorithm - better not to need it, imho.
As the shadow are very dark (as in Guillermo's examples), I need to bring them up - for that task, I find that LR/ACR's Fill Light slider (combined with the Blacks slider to restore contrast) is already not that bad. Of course, the PS methods used by Guillermo are better (because saturation of shadows is better controlled among other things) but it needs quite a bit of time, compared to pulling 2 sliders in LR. Compromises...
I take the shot at base ISO (100ISO with my goodol'Rebel) to ensure that there is not too much noise in the shadows after bringing them up.
-
As the shadow are very dark (as in Guillermo's examples), I need to bring them up - for that task, I find that LR/ACR's Fill Light slider (combined with the Blacks slider to restore contrast) is already not that bad. Of course, the PS methods used by Guillermo are better (because saturation of shadows is better controlled among other things)
To lift the shadows, a straight curve like this applied in Normal RGB blending mode:
(http://www.guillermoluijk.com/tutorial/hdr/curva_brillo.gif)
will always keep colour and saturation. This simple curve just scales the 3 RGB values of the deep shadows by some >1 factor. As long as the relative ratios between R, G and B are preserved, hue and saturation are preserved too since we are just changing exposure (this is true even when using non-linear gamma profiles, except sRGB for which it will be aproximate). That's why I find this simple curve magical and very adequate to lift the shadows after HDR blendings.
In the end of the shadows to be lifted it is recommended a toe towards (255,255) so that if some part gets too overexposed, we will never clip a channel avoiding any partial saturation. The effect of the toe is softly desaturating the highlights up to pure white, which is also a desired behaviour.
Typical values of the control points of the curve to achieve different +EV overexposures are:
(http://img44.imageshack.us/img44/2829/curves.gif)
Regards.
-
Guillermo, what Blend Mode are you using here - is it "Normal" or "Luminosity", because with "Normal" Blend Mode in RGB for curve movements this large one would at least expect saturation shifts.
-
This was posted by another forum member in the Adobe Camera RAW forum here. I've found this to be a very useful technique for making the best use of all of the range captured in a single frame.
The tutorial addresses what you are looking for: managing recovery of tones in a way just a bit more specific than ACR's Recovery tool can manage. Give it a try.
http://imagingpro.wordpress.com/2008/12/03...ingle-raw-file/ (http://imagingpro.wordpress.com/2008/12/03...ingle-raw-file/)
-
Here is a link that works.
SingleRawFile (http://imagingpro.wordpress.com/2008/12/03/expanding-the-dynamic-range-of-a-single-raw-file/)
Thierry
-
Guillermo, what Blend Mode are you using here - is it "Normal" or "Luminosity", because with "Normal" Blend Mode in RGB for curve movements this large one would at least expect saturation shifts.
Normal blending mode. There are no saturation shifts since the curve is a straight line beginning in (0,0), i.e. it's a linear scaling.
Do you experience saturation shifts on your camera when you shoot during 2s instead of shooting during 1s? or when you adjust exposure in the RAW developer? it's the same here, that straight curve just changes exposure so it's perfect for shadow-lifting without altering colour parameters.
Saturation occurs with >1 slope straight curves not passing by (0,0), like in a typical 'S' contrast curve.
Regards.
-
Normal blending mode. There are no saturation shifts since the curve is a straight line beginning in (0,0), i.e. it's a linear scaling.
Do you experience saturation shifts on your camera when you shoot during 2s instead of shooting during 1s? or when you adjust exposure in the RAW developer? it's the same here, that straight curve just changes exposure so it's perfect for shadow-lifting without altering colour parameters.
Saturation occurs with >1 slope straight curves not passing by (0,0), like in a typical 'S' contrast curve.
Regards.
Over a certain range this is correct as far as I've seen it in PS. Take an image with a skin tone, place a marker on it. Create a Curves Adj Layer in Normal Blend Mode, leave the 0'0 point intact and shift the 255'255 point directly leftward accross the top of the Curves dialog box. If you get what I get, you'll see that H and S remain quite constant over part of the range, but disconnect as you get into way more radical shifts which you probably wouldn't use for that part of the image anyhow. So "yes and no and OK"!
In ACR and LR I cannot measure S independently of H and B because there is only an RGB read-out in those applications - something I wish Adobe would improve upon. However, we have heard from them that they purposely include some saturation change consistently with contrast change because their research demonstrates that most people would not find the result of a pure luminosity change of contrast to look "natural". (Not a big deal as one can dial-in or dial-back any reasonable amount of saturation change per colour group one wants after shifting the tone curve - but that's another topic.)
-
However, we have heard from them that they purposely include some saturation change consistently with contrast change because their research demonstrates that most people would not find the result of a pure luminosity change of contrast to look "natural".
The true is that a change in exposure, i.e. a linear scaling of the RGB values that can for example be achieved with a curve passing by (0,0) as described, is the only possible pure luminosity change.
Anything else: HSB/HSV colour models, Lab, blending modes, etc... are just perceptual models to help us deal with the concepts of luminosity and colour (typically split in these models into hue and saturation, or the ab channels, etc...). E.g. if you push L a lot in Lab mode, the proportions between R, G anB will change to finally end in pure white. Speaking in physical terms, that is not only pushing luminosity but also desaturating, i.e. changing colour. Changing the luminosity of a red lamp consists of producing more red light, not turn the red lamp into a pink lamp and finally white.
So paradoxically, changing luminosity using those tools or modes typically intended to allow a control of luminosity without altering colour, means changing not only luminosity but also colour (of course in a way perceptually as pleasant and natural as possible).
Regards.
-
It's a question of which software tool or method is more adequate to ease the task. The information contained in the RAW file (or files) is the same for all of them:
- A RAW developer with shadows/highlight and so forth sliders and options
- Two (or more versions) of the same RAW developed at different exposure values and manually blended in PS layers
- A specific tone mapping tool designed to enhance local contrast while reducing global contrast like Photomatix
...
Using 2 (or more) versions of the same RAW file developed at different exposures allows to easily obtain the desired result, and meanwhile there is no other clear winner option this will be one of the methods to use.
Let me show you a test I did some days ago to find out how well Photomatix can blend RAW files with a big exposure gap (4EV): it showed to be a total dissapointment!. A simple algorithm (implemented both in Zero Noise (http://www.guillermoluijk.com/tutorial/zeronoise/index.htm) or through this simple tutorial (http://jtrujillo.net/qpix/) in PS) consisting in selecting the pixels with the highest exposure performs much better than Photomatix's information blending:
(http://www.guillermoluijk.com/tutorial/hdr/resultadolite10.jpg)
(http://img19.imageshack.us/img19/8554/comp1g.jpg)
(http://img19.imageshack.us/img19/4519/comp2o.jpg)
(http://img190.imageshack.us/img190/4452/comp3p.jpg)
If not even a specialized HDR blending software manages to achieve better results than a proper PS blending, why should we forget manual solutions and think a software tool should always be preferred?
Regards
I am striving to get good result with two versions of the same RAW and I installed on my PC your interesting Zero Noise program.
Sorry for this question, which may seems funny; my question is: how can I feed Zero Noise with the files to be blended, as there is no Browse tool to select the files into the different folders of the PC.
Sorry for this, but I could find the way from the page which is coming upon opening the program Zero Noise.
Many thanks for your help and best regards