This content has been marked as final. Show 94 replies
Post an example raw file that demonstrates your point. (Do not post a JPEG, TIFF, or otherwise rendered file.)
try this image:
look at yellows and blues. terrible....
I downloaded your image, into ACR 5.2 set to ProPhoto space. Backed down the exposure and all highlights look ok, good yellows and blues. Two things to remember:
1. Your only chance of "containing" the entire tonal range of all three channels with a high dynamic range shot like yours is to be in ProPhoto.
2. Don't trust your monitor, check the rgb values with the eye dropper, the colors are subject to being clipped when rendered to your monitor since its gamut capability is much less than the scene.
prophoto does not change anything in the yellows and blues.
i recognized that also when setting the exposure lower, this color shift happens.
here´s the result with 50% highlight repair:
maybe a canon problem???
I took a look at your image in Rawnalyze and very few pixels appear to be clipped at the raw level so that doesn't seem to be a factor.
Here are some of the things that I think are contributing to the color shifts you are seeing:
1. This image has a very wide dynamic range as Richard mentioned.
2. The ACR defaults for this image make the neon lights very over-exposed.
3. If I understand Eric's posts from the past, ACR changes to Exposure, Brightness, and Contrast are not pure luminosity changes. They DO make adjustments to Saturation to try to produce a pleasing effect.
4. The combination of 2 and 3 is producing some color shifts. Bringing the exposure back down by about 1.5 removes the shifts and brings just about everything within gamut (ProPhoto anyway).
You may also be seeing something that has been a bit controversial in forum discussions. Adobe applies a baseline exposure offset behind the scenes for some cameras. For some cameras, this offset can be a half-stop or greater and can create the initial impression that the image is more over- or under-exposed than it really is. In this case it looks to me like that MAY be a factor since the ACR defaults imply heavy over-exposure when in reality few pixels are clipped in the raw file. The good news/mitigation to all of this is to just learn to bring the exposure back down manually in ACR.
> maybe a canon problem???
It is a combination of errors, none of them of Canon's (they do maintain their fare share, but not in this case).
1. The shot is overexposed: see the raw histogram:
The overexposure affects much of the colored fluorescent tubes over the shop, but it does NOT affect the "fetons ...whatever" under "REPLAY" between the two doors.
2. ACR interprets the 5D2 raw data incorrectly. It sets the saturation level at 15600, while it is in fact 15760. This increases the apparent clipping, though the direct effect is limited to the sky.
3. ACR adds 0.4 EV to the exposure without telling it to you.
The above effects *together* push the "fetons whatever" tubes into clipping. This needs to be corrected by -0.6 EV "exposure", not by "Recovery".
However, all this won't help on the problem of color gamut: some of the resulting colors are simply not in sRGB. You need to reduce the brightness even more if you want to keep the colors inside sRGB.
Important note: if you switch in ACR to ProPhoto, you seem to have caught the proper colors. This is eye-wash. As your monitor is certainly incapable of displaying ProPhoto colors, what you see may be more pleasing than what you get with sRGB, but that is NOT the color, which can not be reproduced. You have to go back and take another look in order to see the "true" colors, for no printer and no monitor can reproduce that. Your only alternative is reducing the brightness.
I was working on my post for a while and have not seen yours before sending it away, thus there is much overlap.
thanks guys for your imput.
bringing down EV to -0,6 or even -1,5 makes the color shift even worse in my case.
actually EV + 0,3 displays the colors best i found out.
from what i experienced with the 5d mkII, highlights IN GENERAL have clipped red channels.
think that is the reason why EV and Repair sliders make a mess out of it trying to "save" the reddish highlights.
even with moderate "lighter" images there can occur trouble: light faces can get red, etc...
it seems to be an overal characteristic of this camera with ACR 5.2.
Is there any way to let Adobe know about that issue?
Maybe there is some ACR update possible.
G: No problem on the overlap. Most of what I stated I learned from you and a few other guys here anyway. :)
Nikolaus - One thing I forgot to ask is what profile you are using on the Calibration Tab. You might want to experiment with the 5DMkII-specific profiles (Camera Neutral, Camera Standard, etc.) if you haven't already.
>Is there any way to let Adobe know about that issue?
You just did, by posting here. :)
The Camera Raw team is very active in this one particular forum.
But you can always use the Contact button at the top of this page.
Dennis, I use "Camera Standard". Works best for me. Adobe Standard is second best for my taste.
This is an interesting example of highlight recovery, and I would like to add a few comments to those made by Dennis and Gabor. Gabor's raw histogram demonstrates that the green channel is clipped; the blue channel is near saturation but does not appear to be clipped and the red channel has over 2/3rds f/stop of headroom. In this situation, I would expect ACR's highlight recovery to work well.
As Dennis and Gabor have pointed out, the use of ProPhotoRGB is necessary to avoid saturation clipping. When we open the image in ACR, we can use the alt key with the exposure slider to demonstrate the clipping:
This maneuver demonstrates that the clipping is almost entirely in the sky, which has no important image detail.
The baselilne offset of +0.4 EV causes the image to appear more overexposed than it actually is; one could use a negative exposure value of -0.4 EV to counteract this. Since the foreground is not overexposed, I would leave the exposure control at 0, and use the recovery slider to eliminate the highlight clipping as shown below. A value of 28 eliminates most of the clipping.
The saturated colors of the neon lights may not reproduce on screen or in the print, but are captured in the rendering and I see nothing wrong with the highlight recovery.
Until now I worked on a couple more images I have taken with the Canon 5D II.
Still Highlight repair does odd things.
See how the colors pink and blue change.
That definitely should be no job for Highlight Repair.
Adobe, what´s going on here? Can it be fixed in the next ACR?
Have you tried ACR 5.3 RC?
>See how the colors pink and blue change.
>That definitely should be no job for Highlight Repair.
What you're seeing there is the fact that these colors are outside of your monitor's gamut. The blue tires and parts of the pink tires are outside of sRGB even. Many laptop displays have primaries different from sRGB. I see no change in hue at all when doing recovery on that on my wide gamut display where almost none of the colors are outside gamut. On my laptop (a mac Book Pro) some of the pink tires are washed out because it cannot display those colors. The colors appear to change upon highlight recovery in those washed out areas but that is not real but a consequence of a limited display.
> See how the colors pink and blue change
I do see that someone had much paint and time on hand, but I don't see any problem here.
If you are converting the image in sRGB and the clipping indication is turned on, you can see the red clipping indication on the top of some pink and blue tires. This is a color space issue (they used the wrong paint for sRGB).
I suggest you to do following experiment:
1. switch to ProPhoto color space in ACR,
2. turn on the highlight clipping indication,
3. increase the brightness by either "exposure" or "brightness", until the red clipping indication appears on some spot,
4. pick a point on that with the color sampler; the blue and/or the red will be 255,
5. reduce now the brightness, just so much that the clipping indication vanishes; the RGB values on the sampled point go slightly under 255,
6. switch to sRGB: one or more of the RGB values is 255, and the clipping indication shows the clipping.
If this is not the problem you are complaining about, then I am sorry, pls give a better description than "the colors pink and blue change".
you beat me; I verified this in ACR with different versions and in the meantime you posted already.
Nikolaus, <br /> <br /> Just so we are on the same page, can you tell me if these colors are close to what you see? <br /> <br />c <a href="http://www.pixentral.com/show.php?picture=18tEgqwBYNl0Riun7evAyilfQmhkd" /></a> <img alt="Picture hosted by Pixentral" src="http://www.pixentral.com/hosted/18tEgqwBYNl0Riun7evAyilfQmhkd_thumb.jpg" border="0" />
I have ACR 5.3 beta now with no cure.
Ramon: Exactly. Pink and blue change. Same happens in pictures with yellow getting orange. There is a recent post from a guy having the same problem with the camera:
G Sch; There is NO clipping in the tires. I emphasized that the colour change occurs in NON highlight areas. Prophoto does make no difference.
Jao: The colors of the tyres are displayed correctly and perfectly nice - in gammut - with standard settings. As soon as you turn up highlight recovery it goes wrong...
People, I develop RAW files with Olympus E300 and Nikon D200 also. I never got issues like that with any camera. The pictures I am talking about are well exposed, but exposure correction and highlight repair make a mess of some colours.
Open my RAW files and play with it. What happens here is a bug and should be corrected soon by Adobe.
>Jao: The colors of the tyres are displayed correctly and perfectly nice - in gammut - with standard settings. As soon as you turn up highlight recovery it goes wrong...
As I said, not on my display. The color stays the same when I increase highlight recovery they just desaturate a little which is what you expect for HR. The colors are out of your display's gamut and what you see is a color management artefact! You can test this by softproofing to your monitor profile after import (don't use the highlight recovery and do use ppRGB and make sure to NOT check "preserve RGB numbers") When you then turn on the gamut warning in Photoshop's softproof, you'll see the tyres are all outside the gamut of your display (many are even outside of sRGB BTW). On my wide gamut display they are not and I don't see a "shift". On my Mac Book Pro, which is smaller gamut than sRGB, the tyres are almost completely out of gamut and I see a shift upon highlight recovery. That's not real however but just my MBP's display not being good enough for this image.
Jao, can you send me a jpeg with recovery at let´s say 80% ?
I just don´t understand this since the other cameras i mentioned
do not do anything like that. And my pictures are nothing new
concerning exposure ore colors. Just the camera is new.....
your Canon file as well as the ACR conversion is ok; I don't see any color shift with Recovery. Pls try this version:
It is converted by the DNG converter V184.108.40.206
>Jao, can you send me a jpeg with recovery at let´s say 80% ?
where do you want me to send it? nmachek at mac.com?
Nikolaus <br />>can you send me a jpeg with recovery at let´s say 80% ? <br /> <br /> That image does not need such a brutal recovery setting. A recovery setting of 27 eliminates all highlight clipping. <br /> <br />c <a href="http://www.pixentral.com/show.php?picture=18tEgqwBYNl0Riun7evAyilfQmhkd" /></a> <img alt="Picture hosted by Pixentral" src="http://www.pixentral.com/hosted/18tEgqwBYNl0Riun7evAyilfQmhkd_thumb.jpg" border="0" /> <br /> <br />c Adobe Standard rendering, As Shot, Recovery setting of 27, no other adjustment. <br /> <br />ACR 5.3 RC
Same image as #24 above, Recovery setting of 80, all other settings the same as above: <br /> <br />c <a href="http://www.pixentral.com/show.php?picture=1HkGu7vfA2dSW7Ol4rN3MMb5i7eEq1" /></a> <img alt="Picture hosted by Pixentral" src="http://www.pixentral.com/hosted/1HkGu7vfA2dSW7Ol4rN3MMb5i7eEq1_thumb.jpg" border="0" /> <br /> <br />ACR 5.3
That looks like my rendering (obviously since Ramon and me are using the same code). No change in the blue or pink tint just a negligible loss of saturation.
Jao, please use firstname.lastname@example.org
ok, look here what happens. try it yourself.
the loss of saturation is not neglible to me. neither is yellow turning orange:
I sent you two jpegs in sRGB. I don't see the difference you see in that screenshot on my screen, but you might, even with my jpegs, on yours. The tires stay pink for me. This is confirmed by using the HSB indicator. The hue is the same regardless of recovery, just the saturation and brightness change. So I think indeed this is coming from your display profile. If I take your png screenshot above and "taste" the tires in Photoshop, in the recovery 0 version, the green channel is zero on the pink tires, which means it is blown out of the display gamut. While when I take the recovery 50 version, the green channel is non-zero. This means that recovery brings the tires into your screen's gamut and this appears as a tint shift. Probably if the display could display the original tint, it would not show as a difference in tint, just a desaturation.
What's still puzzling to me is why you insist on using such massive Recovery settings. Your images don't need that kind of brutal correction by a long shot.
When you darken yellow successively, you go from yellow through orange to brown, so I'm not surprised you wrecked the yellow tones in your highlights with a Recovery setting of 80.
these images are just an example. i often used recovery with different cameras.
it does a great job! even at 100%.
but now hl recovery and even exposure create unwanted color shifts.
this guy has the same problem.
with olympus e300 or nikon d200 yellows never gets orange or brown. neither do blue or other colors.
OK. I usually don't have to go above a setting of 30 for the highlight Recovery. In those extremely rare occasions where more is needed, I'm prepared to do other corrections as well.
What was the result with the DNG file I uploaded for you?
G Sch, thanks, I tried the DNG file.
Actually it just shows the same desaturation I talk about.
For my camera i increased dynamics and did some other little corrections, so the difference could be stronger visible.
what do yo see? We are not talking about highlights changing, but colors changing in not-highlight areas.
For yellow changing to orange also check:
I was looking for the problem on the wrong spots (literally). I agree with Nikolaus: something is seriously wrong with the "recovery" function.
Example, on the shot #1504, the fancy "golden" strings on the shoulder of the woman; the other shot, with the colorful tires exhibits the same effect.
I put a sampler on a spot, which obviously changed the color. (Note: the color is out of the sRGB gamut.)
The sampler with Recovery=0 is (255,198,2) in sRGB, (251,197,32) in aRGB, (244,161,65) in ProPhoto RGB.
With Recovery=50, this changes to (255,163,52) in sRGB, (244,161,65) in aRGB and (212,160,68) in ProPhoto RGB.
This means, the "recovery" function decreased the red and green components, but it
the blue one.
This is IMO an error in ACR.
Yeah I see it too. The sequins on the woman's shoulders go from nice yellow to orange. Recovery should NOT change the tint of anything just brightness and potentially saturation. If you take the rendering at recovery 0 in ppRGB and save it as a tiff and open that tiff in ACR, the recovery slider gives the normal effect. Just darker yellow instead of orange when you start with the raw. Clearly there is a bug in ACR for recovery with this camera.
The color shift you see is a result of the (intentional) design of the color profile, not the Recovery or Exposure sliders.
The new color profiles, such as Adobe Standard, Camera Standard, etc. contain 3D tables that involve lightness-dependent hue twists. Again, this is by design and was very commonly (though indirectly) requested by many users. Since Recovery and Exposure run before the application of the 3D table, this means that adjusting these controls can affect the resulting hues and/or saturation.
To summarize: Recovery and Exposure are working fine. The color shift is coming from the profile, not the Recovery/Exposure sliders.
My practical recommendation is: If you don't like the resulting colors, change them to what you do like, either by (1) using the HSL sliders -- the TAT tool is quite convenient here, or (2) using the DNG Profile Editor to make your own profile.
(BTW, you would not have seen this hue twist in the past, because earlier Camera Raw profiles did not include a 3D table.)
Could you perhaps expand a bit on what the underlying reason for the lightness-dependent hue twists is? Frankly, when I first saw one, I thought it was totally bizarre(!) And it does risk strange effects like the one described here. I guess what I'm asking is what is the big advantage that this delivers versus a 2D table?