Skip navigation
This discussion is archived
Noel Carboni 23,514 posts
Dec 23, 2006
Currently Being Moderated

How Camera Raw and Various Profiles Handle Transition Into Overexposure

Oct 3, 2011 1:07 PM

With all the attention I've been paying to profiles lately, I have been opening a lot of images with bright blue colors (hence my recent post on trying to get better blue sky color). 

 

One of the things I've noticed is that the transition into overexposure appears to be handled VERY differently depending on the profile chosen.

 

Of particular interest is an image I shot recently toward a stage where some bright blue lights were shining on the crowd and into my Canon EOS-40D's lens.  I was surprised at how differently the transition into overexposure is made depending on the profile chosen.

 

Notably not one of the Adobe-supplied profiles handles this transition smoothly at all, showing some sharp transitions - which helps explain why I had been having such trouble with sky color.

 

Some examples follow (these have all been converted to sRGB for display via browser).  Specifically note the area around the bright blue light in the upper-right.

 

Adobe Standard profile, as supplied with Camera Raw by Adobe:

AdobeStandardProfileBlueOverexposure.jpg

 

Camera Standard profile, as supplied with Camera Raw by Adobe:

CameraStandardProfileBlueOverexposure.jpg

 

Camera Landscape profile, as supplied with Camera Raw by Adobe:

CameraLandscapeProfileBlueOverexposure.jpg

 

 

The very good 40D "DPP Standard sRGB" profile Vit Novak provided me a couple of days ago, based on his measurements of how Canon DPP handles color:

VitNovak40DProfileBlueOverexposure.jpg

 

 

The in-camera JPEG:

InCameraJPEGBlueOverexposure.jpg

 

-Noel

 
Replies
  • Currently Being Moderated
    Oct 3, 2011 1:37 PM   in reply to Noel Carboni

    Noel, can you post the cr2

     

    I want to try a code to smoothen up the highlights. I've been testing it on blown sky, but I don't have a raw like this, as 99% of my photos are daylight. I can make a transition smoother, although with existing ACR workflow I think it's not possible to make the same degree of matching with in-camera jpeg as in non-blown parts of the picture.

     

    Older Adobe Camera profiles perform badly in this area because of several reasons - baseline exposure tag (causing unnecessary clipping), matrix color space transformation and clipping into Photo Pro (especially in blue color, as blue virtual primary position is outside Photo Pro gamut for most cameras) etc ... Things get worse if you use exposure slider and many times even recovery slider, because profile is calculated on clipped data ... etc ...

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 4, 2011 11:00 AM   in reply to Noel Carboni

    Thanks.

     

    Actually, at default exposure setting (-0.24), my profile also produces abrupt banding around the strongest blue light. However, it's gone by decreasing exposure to about -0.4, or decreasing saturation by 5-6%, or applying recovery 4-5. This banding already exist on the raw - it's where the green channel gets clipped. It looks like the camera / DPP has a sort of automatic recovery procedure or some other mechanism to deal with this, because this part of the picture is rendered differently than the same colors on the test image, so I think I'm not able to emulate this with a dng profile, although I can probably reduce banding at this position in the profile

     

    However, while camera deals better with banding than my profile or Camera Standard , hue is even more wrong - it drops to around 160 (in sRGB), while the original hue of the light is 220 or more. Restoring highlights is always guessing, as information is missing

     

    I hope somewhere in the future we'll get sensors with better dynamic range ...

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 5, 2011 4:34 AM   in reply to Noel Carboni

    Here is another version of Standard profile, where I did some interpolations in highlight range. It behaves better with above picture at default setting of exposure (-0.24) and above. Also, I corrected strange hue shift of near-blown sky present in jpegs from camera. However, it's not possible to eliminate banding around the light present in raw and there will still be some photos rendered better by the camera - can't do better with my current knowledge

     

    It's not possible to eliminate negative exposure correction until Adobe implements this option in dng standard and ACR, unless I do the same as Adobe in profiles for 40D, which results in highlight clipping and issues connected with that

     

    http://www.megaupload.com/?d=2NYD4UWD

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 5, 2011 11:57 PM   in reply to Noel Carboni

    Ok Noel, I'm glad this worked fine for you

     

    Several months ago, in one post I stated that I wouldn't underestimate knowledge of companies like Nikon and Canon. Jeff didn't quite agree with me about that, claiming that Adobe was much longer in color business. However, 5 years (or so) ago, when I bought my 400D, it had quite good color management and various things like dealing with highlight clipping were sorted quite good. At that time, PS didn't have profiles at all, and in some aspects even newest profiles don't do the job as good

     

    I made this mostly out of curiosity and because my Canon compacts were lacking ACR support. And to stimulate Adobe to improve color management. Of course, it was a big waste of time. I have no intention to sell these things - I didn't invent anything new, I just tried to extract color profiles from the camera / DPP so they work almost as well as in camera

     

    Also, it would be good if somewhere in the future Adobe provides better profiles for older cameras

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 6, 2011 3:03 AM   in reply to Noel Carboni

    I haven't been following this too closely, but isn't the way highlights are badly rendered partly to do with baseline exposure in ACR? Looking at the example above, I'd guess an exposure offset of half a stop would fix the strange blue hue and banding. My own defaults include -0.5EV exposure compensation, and, as I try to expose to the right, this means my high DR shots are often underexposed intentionally to preserve highlights. This is usually compensated by a mixture of small amounts of highlight Recovery, a boost to Brightness, and a suitable amount of Fill, un-flattened by a moderate amount of Clarity.

     

    One thing I don't do is adjust Exposure for the midtones, which is how I'd expect it to work, as this tends to provide washed-out highlights in ACR.

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 6, 2011 8:19 AM   in reply to Noel Carboni

    I get what you're saying now.

     

    It's similar to an epiphany I had a few months ago, when dealing with a small number of photos with very wide gamut colours.

     

    I always (wrongly) assumed that "rendering intent" was very clever in the way that images were converted from one colour space to another. I understand the semantics of the different intents (e.g. gamut clipping as opposed to gamut compression), but in practice this didn't match reality.

     

    Some may remember my incredibly red tulips being converted from ProPhoto to AdobeRGB to sRGB. I understood that conversion of the image to a smaller space using relative colormetric intent would lose some of extreme reds and turn sections of the image to solid red, but what confused me was how perceptual intent produced equally useless conversions. I'd expected a hue shift in order to accomodate the extreme reds, but the results were more or less the same as before. It was explained to me that I would have to manipulate HSL before attempting any colour space conversion. So, I still ask myself what the point of perceptual conversion is.

     

    You're quite right in expecting ACR to deal with highlight areas more satisfactorily. My suggestion was in effect a workaround, which shouldn't be required. Of course you should adjust for midtones, then deal with highlights and shadows later. And you shouldn't have to come up with workarounds to return proper saturation to skies and skin, just because they are very bright.

     

    Like Vit said earlier, it might be cool to see some better camera profiles. I'm not sure how that would help me though, as I generate my own using Adobe's profile editor software and a ColorChecker. Maybe I need to buy Vit a beer and get him to make me one too. What's your poison, Vit?

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 6, 2011 2:16 PM   in reply to Yammer

    About rendering intents ... icc profiles also evolved during timeline. sRGB.icc on my computer is only 3 kB in size. It's a matrix profile and no matter what rendering intent I select in PS, result is the same (can't be different I suppose). However, on icc site, you can download sRGB v4 profile, which is about 60 kB in size and has matrix for colorimetric rendering and lookup table for perceptual rendering, which means compression of bigger color space into sRGB and various adjustments of hue/saturation/brightness.

     

    Rendering intent in ACR is colorimetric - out of gamut colors are clipped by clamping R, G and B into range 0..1

     

    Rendering intent in Canon cameras and DPP is a kind of perceptual - out of gamut colors are compressed into sRGB or AdobeRGB (at least partially - on top of the gamut it looks like they are also clipped, hence cyan tone of near-blown sky in sRGB mode)

     
    |
    Mark as:
  • Currently Being Moderated
    Oct 6, 2011 10:59 PM   in reply to Noel Carboni

    This hue shift in case of blue/cyan is intentional, also because blown blue colors are rendered around sRGB hue 190 deg, so transition has to be smooth. But in case of sRGB profile, it is abrupt somewhere around half of LUT (depending on saturation), where sRGB nonlinear values reach 1, while in Adobe RGB it's much smoother. It looks like it is clipped into sRGB

     

    Problem about highlights and possible occurence of hard ring is that Canon seems to has built-in LUT that extends above brightness value 1, to cover sensor output after whitebalancing, where R and B can be above 1 (because R and B channels are less sensitive and they are multiplied with WB R and B multipliers). Dng profiles don't offer this possibility - everything above 1 is clipped before LUT calculation, so I adjusted calibrating procedure to this difference, but it can't be handled properly. So whatever I do, it won't be the same in this region (+ there can be some additional logic in the camera). But ACR has recovery slider, which compresses upper region to fit into LUT

     

    This subject is quite complex, really

     
    |
    Mark as:

More Like This

  • Retrieving data ...

Bookmarked By (1)

Answers + Points = Status

  • 10 points awarded for Correct Answers
  • 5 points awarded for Helpful Answers
  • 10,000+ points
  • 1,001-10,000 points
  • 501-1,000 points
  • 5-500 points