Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Luminance vs saturation clipping  (Read 15660 times)
Gurglamei
Guest
« on: September 09, 2009, 02:35:32 AM »
ReplyReply

I just finished looking at the video tutorial on ACR, and a couple of places Jeff Schewe mentions the difference between luminance and saturation clipping. However, I didnt really understand how to tell the difference or the importance about the difference or how to deal with it.  

Jeff, could you or somebody else for that matter, please help me out and explain in som more detail?


Christopher
« Last Edit: September 09, 2009, 02:40:51 AM by Gurglamei » Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5426


WWW
« Reply #1 on: September 09, 2009, 11:54:35 AM »
ReplyReply

Saturation clipping occurs in one color (although the color may be made from two RGB colors such as yellow or magenta) because the shot color exceeds the gamut of the color space at a given exposure. Often you can help get rid of saturation clipping by using Pro Photo RGB instead of Adobe RGB (or, heaven forbid, sRGB).

Luminance clipping is pure over exposure at those given pixels...of all three RGB color at pinned to 255, that's luminance clipping.

The advantage of suffering only saturation clipping vs luminance clipping is that highlight recovery can often get back detail even if there's only one channel of image data. It may be off color, but you can deal with that. HOwever, if all three channels are blown, there ain't nothing left to recover.

And if you engage in ETTR, it's your job to know exactly what point the increase in exposure will just fall short of luminance clipping. The image itself may dictate whether or not saturation clipping is a problem.
Logged
evonzz
Jr. Member
**
Offline Offline

Posts: 79


« Reply #2 on: September 09, 2009, 07:03:11 PM »
ReplyReply

Quote from: Schewe
And if you engage in ETTR, it's your job to know exactly what point the increase in exposure will just fall short of luminance clipping. The image itself may dictate whether or not saturation clipping is a problem.

Whats ETTR?
Logged

www.centralstudios.cn
- a large shooting space in Central Shanghai

Phase One P40+/P65+/IQ180
and lots of other stuff
Schewe
Sr. Member
****
Offline Offline

Posts: 5426


WWW
« Reply #3 on: September 09, 2009, 10:25:59 PM »
ReplyReply

Quote from: evonzz
Whats ETTR?


Expose to the Right?

Read this.
Logged
Gurglamei
Guest
« Reply #4 on: September 10, 2009, 03:03:25 AM »
ReplyReply

Quote from: Schewe
Saturation clipping occurs in one color (although the color may be made from two RGB colors such as yellow or magenta) because the shot color exceeds the gamut of the color space at a given exposure. Often you can help get rid of saturation clipping by using Pro Photo RGB instead of Adobe RGB (or, heaven forbid, sRGB).

Luminance clipping is pure over exposure at those given pixels...of all three RGB color at pinned to 255, that's luminance clipping.

The advantage of suffering only saturation clipping vs luminance clipping is that highlight recovery can often get back detail even if there's only one channel of image data. It may be off color, but you can deal with that. HOwever, if all three channels are blown, there ain't nothing left to recover.

And if you engage in ETTR, it's your job to know exactly what point the increase in exposure will just fall short of luminance clipping. The image itself may dictate whether or not saturation clipping is a problem.

Thank you, that was very clarifying.


Christopher
« Last Edit: September 10, 2009, 03:03:57 AM by Gurglamei » Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #5 on: September 10, 2009, 06:06:52 AM »
ReplyReply

Quote from: Schewe
Expose to the Right?

Read this.

and this
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #6 on: September 10, 2009, 01:13:02 PM »
ReplyReply

Something interesting about partial saturation (the way I call saturation clipping), although obvious is seldom mentioned: partial saturation in a digital image can strongly affect colour images since any clipped channel produces wrong colours (see for example typical yellowish circle around the sun). In addition to this, if the RAW was already clipped in some channel and highlight recovery is used in the RAW developer, textures can be recovered but in a monochrome fashion. Colour is totally lost and this is evident looking at the image.



One of the versions displays partial saturation in the lamps (see the awful yellow colour), the second version doesn't and has a smooth white luminance transition from the lamps.

On the contrary, with B&W images in mind partial saturation is never a problem. As long as some channel remains unclipped in any RAW channel (typically the B channel), we can recover texture there without even noticing some channel was blown.

The conclusion for this is that B&W images are easier to expose: we can maximize ETTR, and we can make mistakes with less consequences than in colour images. If there is usually a gap of around 1.5EV from the B to the G channel in common RAW files (depending on highlights subject), that gives B&W images 1.5EV of extra headroom from partial saturation to total texture loss. This is good news for B&W images since the absence of colour in the deep shadows makes less noise even more desirable than in colour images, and this can be achieved by exposing more.

Regards
« Last Edit: September 10, 2009, 01:24:19 PM by GLuijk » Logged

Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #7 on: September 10, 2009, 01:54:40 PM »
ReplyReply

Quote from: GLuijk
... partial saturation in a digital image can strongly affect colour images since any clipped channel produces wrong colours (see for example typical yellowish circle around the sun). In addition to this, if the RAW was already clipped in some channel and highlight recovery is used in the RAW developer, textures can be recovered but in a monochrome fashion. Colour is totally lost and this is evident looking at the image.
...
On the contrary, with B&W images in mind partial saturation is never a problem. As long as some channel remains unclipped in any RAW channel (typically the B channel), we can recover texture there without even noticing some channel was blown.
Whenever I revisit classic film, it seems to hold highlight colors much better, even and in particular when luminosity differences are already clipped (see e.g. sunset shots).

Im wondering if anyone ever compared the DR along vivid color scales.

Peter

--
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #8 on: September 10, 2009, 02:16:56 PM »
ReplyReply

Quote from: DPL
Whenever I revisit classic film, it seems to hold highlight colors much better, even and in particular when luminosity differences are already clipped (see e.g. sunset shots).

Im wondering if anyone ever compared the DR along vivid color scales.
Since film is all about analogue chemicals, clipping on it is a smooth transition for any colour from a given exposure to pure white. Sometimes it's easy to recognize a digital landscape just by looking at the wrong colours surrounding the sun, try it. This is one of the disadvantages of digital.



I think we cannot call this a problem of DR, but of how the highlights reach saturation: progressively in film, suddendly on each RGB channel in digital.

BR
« Last Edit: September 10, 2009, 02:20:12 PM by GLuijk » Logged

Schewe
Sr. Member
****
Offline Offline

Posts: 5426


WWW
« Reply #9 on: September 10, 2009, 10:28:45 PM »
ReplyReply

Quote from: DPL
Whenever I revisit classic film, it seems to hold highlight colors much better, even and in particular when luminosity differences are already clipped (see e.g. sunset shots).

Doode, you seriously need to shoot some digital side-by-side with file. Film sucks in terns of recoverable scene contrast range relative to what film or digital can capture. Seriously, you are thinking in a non-realistic "nostalgic through the rose-color lens" sorta of frame of mind. I've shot enough sunrise/sunsets with both film and digital to tell you that film seriously and completely sucks compared to digital.

The _ONLT_ dick-measuring allowed is how good you are at post-processing the scene.... Me? I'm pretty friggin' good at post, so the question is kinda moot...the bog question remains, how long is yours?

:~)
Logged
Rob C
Sr. Member
****
Offline Offline

Posts: 12215


« Reply #10 on: September 11, 2009, 02:24:05 AM »
ReplyReply

Indigestion seems to be a common ailment.

Rob C
Logged

NikoJorj
Sr. Member
****
Offline Offline

Posts: 1063


WWW
« Reply #11 on: September 11, 2009, 04:17:07 AM »
ReplyReply

Quote from: DPL
Whenever I revisit classic film, it seems to hold highlight colors much better, even and in particular when luminosity differences are already clipped (see e.g. sunset shots).
I respectfully disagree : it's not that it holds highlight colors better, it's just that it distorts the color of saturated areas less than with mainstream raw developers (the way dcraw handles partial saturation seems often much more elegant, from what I've seen here and notably from Guillermo).

But film definitely desaturated HL more : don't you remember slightly underexposing slides to keep saturation at an optimal level?
« Last Edit: September 11, 2009, 04:18:42 AM by NikoJorj » Logged

Nicolas from Grenoble
A small gallery
ckimmerle
Sr. Member
****
Offline Offline

Posts: 442



WWW
« Reply #12 on: September 11, 2009, 09:48:04 AM »
ReplyReply

Quote from: GLuijk
On the contrary, with B&W images in mind partial saturation is never a problem. As long as some channel remains unclipped in any RAW channel (typically the B channel), we can recover texture there without even noticing some channel was blown.


I've got to take issue with the word "never". Yes...most texture can be recovered from a single clipped channel when working in grayscale, but depending on the area of the clipping, it may not be enough. For instance, in very bright and flat highlight areas, that missing information my be quite valuable to maintain the desired level of detail. Same for the deep shadows, which never have enough information even when exposed perfectly.

In my workflow, the first adjustment layer I open in Photoshop is always LEVELS (simply for the large detailed histogram) so that I can look for channel clipping, which usually happens (at least for me) in the blue channel. If that is indeed the case, my next adjustment layer is CHANNEL MIXER, where a simple blue channel mix of 90B 10G is usually enough to bring the offending channel back into gamut. Occasionally, I'll have a clipped blue channel in the highlights, and a clipped red in the shadows. I'll try to repair both. Only then will I do a grayscale conversion and set black and white points. Of course, every photog/camera combination will have different problem areas.

The changes may not be dramatic but, in the deep shadow and bright highlight areas, can be invaluable.

addendum: it's not always possible to bring all colors into gamut in ACR while still maintaining a decent color balance, so the technique above is a very good alternative even if working in color.
« Last Edit: September 11, 2009, 10:23:48 AM by ckimmerle » Logged

"The real voyage of discovery consists not in seeing new landscapes, but in having new eyes." Marcel Proust

Chuck Kimmerle
WWW.CHUCKKIMMERLE.COM
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #13 on: September 11, 2009, 10:21:13 AM »
ReplyReply

Quote from: GLuijk
Since film is all about analogue chemicals, clipping on it is a smooth transition for any colour from a given exposure to pure white. Sometimes it's easy to recognize a digital landscape just by looking at the wrong colours surrounding the sun, try it. This is one of the disadvantages of digital.

I think we cannot call this a problem of DR, but of how the highlights reach saturation: progressively in film, suddendly on each RGB channel in digital.

BR

Guillermo's post raises some interesting aspects of digital imaging and white balance.

The spectral response of a sensor can be determined with scientific equipment, as shown in a post from Stanford University, reproduced there under fair use copyright. The study is for the Nikon D70 and shows that the sensor is most sensitive to green, followed by blue and then red.

[attachment=16517:D70_spec...response.png]

If you take a picture of a white reflector with the D70, the RGB response will be determined by the amounts of red, blue, and green in the illumination and the spectral response of the sensor. Daylight has relatively equal amounts of red, blue and green as shown in this spectral analysis on the General Electric Lighting Web Site.

[attachment=16518:daylightSpectrum.png]

To determine the camera response, one must convolve the spectral response of the camera with the spectral composition of the daylight. For example, with green one would divide the camera response of the green channel into small ranges, perhaps at 10 nm intervals (bands). One would do the same with the daylight spectrum and then multiply the radiance in each band by the response of the camera.

This image shows a shot of a MacBeth color checker photographed in daylight with the D70 without any white balance. The histogram shows relative exposures of the channels. To apply white balance, the red and blue channels are multiplied by white balance coefficients such that the exposures are equalized. In the raw file, green is blown first, followed by blue and then red.

[attachment=16519:D70_raw_daylight.png]

If you photograph the target under tungsten light, you get a quite different response since tungsten is quite deficient in blue as shown in the SPD of tungsten illumination. With tungsten and the D70, green blows first, followed by red and rather distantly by blue.

[attachment=16520:D70_raw_tungsten.png]

[attachment=16521:TungstenSpectrum.png]

Now to Guillermo's picture. The disc of the sun is completely blown and all pixel values will be white, RGB = 255, 255, 255

As one moves away from the disc, the radiance decreases and color shifts are introduced as shown in the histogram below. At the indicated sampling point, the green and red are near saturation and the blue is below clipping. The result is yellow.

[attachment=16522:blownSunset.png]

When channels are blown with a digital sensor, there will be color shifts due to unequal sensitivities of the sensor to red, blue, and green as shown.

If the spectral responses of the sensor were equal for all colors at the given illuminant, there would be no color shift. Since transparency film is color balanced for the illuminant, it does not suffer this shift. One could put a filter over the camera lens with a digital camera so as to equalize the channels and avoid the shifts. Due to nonlinearity with clipping, this would be difficult to do in post, just as color balancing an image taken under tungsten illumination with daylight Kodachrome is difficult.


Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #14 on: September 11, 2009, 11:38:48 AM »
ReplyReply

Quote from: ckimmerle
I've got to take issue with the word "never". Yes...most texture can be recovered from a single clipped channel when working in grayscale, but depending on the area of the clipping, it may not be enough. For instance, in very bright and flat highlight areas, that missing information my be quite valuable to maintain the desired level of detail. Same for the deep shadows, which never have enough information even when exposed perfectly.

In my workflow, the first adjustment layer I open in Photoshop is always LEVELS (simply for the large detailed histogram) so that I can look for channel clipping, which usually happens (at least for me) in the blue channel.

Do you have some RAW file where with one unclipped channel you were not satisfied by the amount of information you managed to recover for B&W imaging? it would be interesting to study it in depth.

BTW the blue channel very rarely (I don't want to say never because someone will come to demonstrate the opposite hehe) is the first in clipping in the RAW file (see Bill's explanations and histogram). Your experience is due to the fact that after white balance, the B channel gets overexposed with respect to its captured values possibly reaching higher values than the other 2 channels in the highlights (this is always true in the sky, since skies are blue!). That is why you see that channel clipping first in most of your images once in the postprocessing stage. But in the camera, the B channel is the safest of the three.

Regards.
« Last Edit: September 11, 2009, 11:42:49 AM by GLuijk » Logged

Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #15 on: September 11, 2009, 02:01:16 PM »
ReplyReply

Quote from: GLuijk
... Sometimes it's easy to recognize a digital landscape just by looking at the wrong colours surrounding the sun, try it. This is one of the disadvantages of digital.



I think we cannot call this a problem of DR, but of how the highlights reach saturation: progressively in film, suddendly on each RGB channel in digital.

BR
I do have no problem at all to admit when a subject-matter is going over my head - like this one here.
(actually I find it more problematic to understand, as a non-native speaker,  if "doode" is only negatively meant or has a slight positive touch.)

What I can say for sure is that I have a similar image of a sunset scene printed, framed and hanging in our staircase (family likes it).
Shot about 20 years ago when Provia film was newly available.

Shadows are darker though. But "surprisingly", the circle of the sun is intensively yellow to orange,
rather than blown white. It does of course not show any sunspots, but color is still in place.

Peter

--
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #16 on: September 11, 2009, 05:53:54 PM »
ReplyReply

Quote from: DPL
I do have no problem at all to admit when a subject-matter is going over my head - like this one here.
(actually I find it more problematic to understand, as a non-native speaker,  if "doode" is only negatively meant or has a slight positive touch.)
Peter,
Since you are one of the most scientifically advanced posters to these forums, I rather doubt that the subject matter is going over your head. Something presented here may be new to you, but once you have studied whatever that might be, I'm sure you could understand the subject matter. Jeff Schewe does not always use the King's English, but since American English is my native language and since I am familiar with Jeff's posting style, I would not regard that he used "doode" in a positive sense.

Quote from: DPL
What I can say for sure is that I have a similar image of a sunset scene printed, framed and hanging in our staircase (family likes it).
Shot about 20 years ago when Provia film was newly available.

Shadows are darker though. But "surprisingly", the circle of the sun is intensively yellow to orange,
rather than blown white. It does of course not show any sunspots, but color is still in place.

Film does saturate, but less abruptly than digital due to the shoulder, which can be rather long with negative film. Roger Clark has compared the dynamic range of his Canon ID MII to film, both negative and positive. Of course, the digital is superior to transparency film (Fuji Velvia, which is rather contrasty and perhaps not the best choice for DR), but (contrary to popular belief) he also finds it superior to negative film. He exposed the Kodacolor film to saturation as seen by his Polaroid Sprintscan, which uses a bit depth of 12. Perhaps he missed a bit of the shoulder. Anyway, I would think that digital would be better than film for high contrast sunset images.

Was the sun's disc yellow on the transparency also? If so, the yellow layer of the emulsion must not have been fully saturated or perhaps solarisation took place. Perhaps your image was not overexposed to the extent that Guillerom's was. I Think my explanation of the yellow halo is correct, but can't explain the appearance of your image.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #17 on: September 12, 2009, 08:31:04 AM »
ReplyReply

Quote from: GLuijk
BTW the blue channel very rarely (I don't want to say never because someone will come to demonstrate the opposite hehe) is the first in clipping in the RAW file (see Bill's explanations and histogram). Your experience is due to the fact that after white balance, the B channel gets overexposed with respect to its captured values possibly reaching higher values than the other 2 channels in the highlights (this is always true in the sky, since skies are blue!). That is why you see that channel clipping first in most of your images once in the postprocessing stage. But in the camera, the B channel is the safest of the three.

Regards.

If the image contains a lot of blue and few whites, the blue can easily saturate first, but what Guillermo says is generally true. I am posting this image not because of its artistic content, but because it demonstrates several points raised in this thread. It is in sRGB for web display.

[attachment=16527:0004_small.jpg]

The raw histogram shows that I didn't do a very good of exposing to the right. The blue channel is closest to saturation, but it is one stop below clipping. Since the blue multiplier for white balance is 1.36, the camera histogram would look more reasonable. With overexposure, the blue raw channel would blow first.

[attachment=16528:04_rawHistogram.png]

If one renders this image into sRGB using the Adobe Standard profile and default settings (except for black), there is strong saturation clipping of the blues, some saturation clipping of the reds, and slight luminance clipping as shown on the preview exposure adjustment with the alt key pressed.

[attachment=16529:sRGB_clip.png]

Changing the space to ProPhotoRGB nearly eliminates the blue clipping, but there is still slight luminance clipping as expected, since the luminance gamut of the two spaces is the same. Some red and green saturation clipping is also present.

[attachment=16530:ProPhoto_clip.png]

One might think that there should be no saturation clipping with such a wide space as ProPhotoRGB, but overexposure of the rendered image explains the clipping. One characteristic of RGB spaces is they have maximum saturation at mid luminances, but as one approaches a luminance of 100, the gamut narrows. This is shown in this 3D gamut plot of sRGB:

[attachment=16531:sRGB_gamut.png]

Examination of the preview reveals that the image is overexposed and washed out.

[attachment=16534:ProPhoto_default.png]

The rendered image is overexposed because ACR uses a baseline exposure offset of +0.5 EV for this camera. If one uses negative exposure to eliminate the luminance clipping, the image is in the ProPhotoRGB gamut.

[attachment=16532:ProPhoto_noClip.png]

Take home points are if you really want to expose to the right, you should have UniWB to get a better idea of the status of the raw file channels. Blue can clip first, if the subject contains a predominance of blue and few whites. Saturation clipping can be removed or lessened by exposure adjustment but this may not be a good idea, since the image will become too dark. In the current case, use of proper exposure eliminated most of the saturation clipping in sRGB, and any difference between sRGB and ProPhoRGB would not be seen with most monitors which can hardly exceed the sRGB gamut.

Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #18 on: September 12, 2009, 06:28:35 PM »
ReplyReply

Quote from: bjanes
If the image contains a lot of blue and few whites, the blue can easily saturate first, but what Guillermo says is generally true.

Agreed Bill, but remember Nikon cameras perform a clandestine (and apparently stupid) pre-white balance, where the R and B channels get slightly overexposed prior to saving the final RAW data. We should substract this effect to find out more about the sensor's real response.

In a Canon camera it's _very_ difficult for the B channel to clip the first. This scene is rich in blues (sky) and reds (ground):




However G still wins in the RAW data:




And if we restrict the RAW histogram to the upper half of the image (sky) we still get the same story:



Regards.
« Last Edit: September 12, 2009, 06:29:44 PM by GLuijk » Logged

ckimmerle
Sr. Member
****
Offline Offline

Posts: 442



WWW
« Reply #19 on: September 12, 2009, 11:16:03 PM »
ReplyReply

This is all too much thinking for my brain but, if it helps the discusison any, the blue channel usually clips on the shadow end. Not a lot, but enough that I feel it needs to brought back into gamut. This happens primarily on outdoor shots and, if I remember correctly, was the same on the D2x.

I looked for good examples of before and after, but everything is too subtle to really be of use. That may mean my channel mixing for gamut isn't all that necessary, but I swear it helps.

Logged

"The real voyage of discovery consists not in seeing new landscapes, but in having new eyes." Marcel Proust

Chuck Kimmerle
WWW.CHUCKKIMMERLE.COM
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad