One Way of Fixing an Out-of-Gamut Flower Shot

[xpatUSA]

This post is for those who are irritated by color clipping after converting to sRGB from a wide color space, such as ProPhoto, and who are more into viewing on a color monitor or posting on the Web.

I've been battling with a sunflower shot for 11 months and today got quite a good output, gamut-wise. I was so pleased, I posted the method on my website here:

http://kronometric.org/phot/convRAW/...compFlows.html

Warning - color management terminology used . . ;-)

[Cantab]

Ted, thanks for posting this. I read the post on your website with interest albeit not with 100% understanding. Is the simple workflow to use Dry Creek's profile converter program without embedding the profile?

[xpatUSA]

Thanks Bruce,

For me, workflow depends on the intended target audience and, to an extent, on the captured image itself. For snaps made for illustrative purposes, i.e. not artistic merit, I go straight from raw to 8-bit sRGB JPEG in one shot with minimal processing (crop, resize, sharpen) and no worries about gamut clipping and such. For keeper images: watches get more attention to the watch-grail of sharpness. For flower shots, more care about everything. Landscapes less care with processing, more with composition. All my my output is JPEG sRGB, rarely printed, if ever.

By not embedding the V4 profile, the Profile Converter utility changed the image data less severely than otherwise (it seemed to me but it doesn't make much sense) and not all applications do a good job on images with an embedded V4 profile, I'm told. Embedding a V2 without changing the image data seemed to work much better.

From your kit, what you already have should do the job almost all of the time, by the looks of it. Much of my complication (fun) stems from refusing to play the "constant upgrade" game - to the puzzlement of many. A recent post on LL about the shortcomings of XP's Windows Explorer thumbnails (only partially color-managed) brought some pretty contemptuous responses ;-)

[Glenn NK]

Ted:

I saw the post - there are some pretty blunt and (obnoxious) people on LL.

Glenn

[DanK]

Ted,

I found your post really interesting (and would like to play with the color viewing program), but I think there may be easier options. I do a lot of flower photography and have encountered problems like this, although more often with reds than yellows. However, I leave images in raw format as long as I can and edit in LR first, which has both perceptual and relative colorometric conversions. It also allows you to pull down or increase individual color channels. That usually seems sufficient, although out of gamut colors require that you lose some information regardless in converting to a smaller color space.

Is there reason to think that the Dry Creek converter will do a better job of reducing the colors to the sRGB space than Adobe's perceptual rendering?

Dan

[xpatUSA]

Ted,

I found your post really interesting (and would like to play with the color viewing program), but I think there may be easier options.

Is there reason to think that the Dry Creek converter will do a better job of reducing the colors to the sRGB space than Adobe's perceptual rendering?

Dan

Thanks for sharing your workflow, Dan. I should have emphasized that I don't have the latest in CS, PS, LR preferring to stick with simple PSE6 + sundry utilities - being an interested dabbler, rather than a Pro ;-)

Your question raises a point that I've read on other forums: although Adobe shows an option of Perpetual rendering intent in their applications, you actually get Relative Colorimetric, at least for non-printing output files. So, from that viewpoint, Dry Creek Photo's converter would do a better job.

Something to do ICC V2 versus V4? Or perhaps simple (small) matrix-based profiles versus those with look-up tables? Probably doesn't apply to print output profiles but, there again, I don't print anything.

Perhaps in LR4 you could test the one intent against the other? I would be happy to view the result in ColorThink if you post the images or give us a link to them.

If you'd like to play with a sunflower TIFF here's one:

http://kronometric.org/phot/convRAW/...7/sfPP8bit.tif

It's in 8-bit ProPhoto space, small size 576x378px.

[DanK]

Ted,

I just downloaded your image and started to play with it in LR and photoshop. I had forgotten one important detail (very important for your workflow), which is that LR seems to give you the option of changing between relative and perceptual only in the print module, when you soft proof. I'll have to explore when I have more time whether it is there more generally. Photoshop offers percepual, relative, absolute, and something called "saturation" that I don't understand.

Seems to me that a better test would be with a 16-bit image, no?

Dan

[xpatUSA]

Ted,

I just downloaded your image and started to play with it in LR and photoshop. I had forgotten one important detail (very important for your workflow), which is that LR seems to give you the option of changing between relative and perceptual only in the print module, when you soft proof. I'll have to explore when I have more time whether it is there more generally. Photoshop offers percepual, relative, absolute, and something called "saturation" that I don't understand.

Never used saturation intent but I've read that it tries to maintain color saturation at the expense of color accuracy. Good for clip art, for example. It would probably not work well when going from a well-saturated wide color space to a smaller one.

Seems to me that a better test would be with a 16-bit image, no?

I can't see any difference in ColorThink because the gamut graphs have much fewer than 16.7M colors in them anyway.

[xpatUSA]

. . . although Adobe shows an option of Perpetual rendering intent in their applications, you actually get Relative Colorimetric, at least for non-printing output files. So, from that viewpoint, Dry Creek Photo's converter would do a better job.

Something to do ICC V2 versus V4? Or perhaps simple (small) matrix-based profiles versus those with look-up tables?

After poking around in various profiles and testing various conversions of a flower shot from ProPhoto to sRGB, I believe I have found the difference in profiles that causes the Perceptual intent to fail. The ICC sRGB V4 profile has tags for some big tables called A2B and B2A which V2 does not have.

Here's some gamut plots comparing ProPhoto to V2 and V4, both selected to 'Perceptual' intent.

From the side of the sRGB color space:



The ProPhoto (yellow) is well outside of the sRGB gamut. The V2 image gamut (blue) is scrunched up against the side of the sRGB gamut, typical of Relative Colorimetric even though Perceptual was selected. The V4 image gamut (red) is well inside the sRGB gamut, almost too much, but fatter indicating that the conversion was indeed Perpetual as selected.

The view from the top (2D) is less cozy:



Same plot colors as above. The hue in theProPhoto original image is a nice golden color. However, both conversions show a color shift counter-clockwise, especially the V4. This is fairly obvious in the following composite screen shot from PSE6:



Notice also the heavily clipped blues in the V2 conversion at left!

Almost makes you want to work in sRGB right away and forget about wide color spaces . .