Ad
Ad
Ad
Pages: « 1 [2]   Bottom of Page
Print
Author Topic: Looking for a good color space converter - found, thank you.  (Read 6626 times)
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #20 on: April 07, 2013, 01:56:22 PM »
ReplyReply

Depends on the profile, how it was built and where you can use it. Generally speaking, V4 profiles bring nothing useful to the party, they are half baked "V2 profiles in V4 sheep's clothing". I don't know what profile package today supports what is called the PRMG which is a key to V4 profiles being (more) useful.

Until PRMG is fully supported, we're back to the part of the conversion whereby the CMM gets Lab data and has no idea about the source (was it big like ProPhoto or small like sRGB?). Each package makes it's own assumptions which is, well an assumption. It is useful with the V4 sRGB profile to have a perceptual table, but by the time that profile get's it's data, the Perceptual Rendering Gamut is no more defined than if you were using a V2 profile.

Shown is my sunflower image that was in ProPhotoRGB and converted to sRGB with massive clipping (left). Converting to sRGB perceptual v4 with perceptual rendering helps somewhat, but colors are still clipped (right). As the Dog states, the CMM has no idea of what color was in the original, and the colors are merely compressed by a predetermined amount.

Bill

Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5416


WWW
« Reply #21 on: April 07, 2013, 02:01:44 PM »
ReplyReply

As the Dog states, the CMM has no idea of what color was in the original, and the colors are merely compressed by a predetermined amount.

But, it's an improvement...and with soft proofing you could tweak the colors even more.

So, zooming in, do you see any examples of where relcol produces blotches that perceptual doesn't?
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #22 on: April 07, 2013, 04:30:54 PM »
ReplyReply

But, it's an improvement...and with soft proofing you could tweak the colors even more.

So, zooming in, do you see any examples of where relcol produces blotches that perceptual doesn't?

No, I don't see any blotches with either conversion, but I would imagine that blotches would depend on the subject matter.

Bill
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #23 on: April 08, 2013, 12:24:57 AM »
ReplyReply

I'm back, after a some more research, some experimenting with the sunflower and re-thinking the reasons behind the OP.

For me, my sunflower problem posted earlier occurred with ACR 5.4's conversion from X3F into it's working space. In that version of ACR what you see (all sliders 0) is what you got, assuming WB to be as good as it can be. That is to say, the conversion itself is fixed by Adobe. If there is a clipping problem caused by an image occupying a large gamut, that can not be fixed in ACR. That is to say, the image appears on the monitor clipped or blotched and, only when the sliders are set to extreme levels downward do the blotches go away - at which point the image looks stupid, washed out, a total failure. It may well be that the in-camera exposure was high for the shot (it was a one-off snap) but the RAW data itself was not clipped - I checked and re-checked.

On the other hand, Dave Coffin's 'dcraw' allows all sorts of adjustments to the raw data during the conversion, among them being saturation, brightness and gamma - not to mention a choice of output color spaces: raw, sRGB, aRGB, aWIDE, PP and XYZ. Not that I recommend dcraw particularly, it has a clunky command line interface and a somewhat protracted work cycle.

So it is that, armed with a better understanding of color spaces and profiles gained from this thread, I've managed to produce a blotch-free sunflower (again) but this time with some knowledge of how that was possible.



(Slightly sharpened, no other processing).

I no longer regard 'perceptual' as a kind of Grail, especially for one who only produces sRGB images and never prints anything. I will continue with what applications I have for now. Any 'difficult' capture will get either dcraw'ed or trashed; ACR->PSE for the rest . . ;-)

Perhaps I should mention that my serious cameras are RAW only (early Sigmas).
« Last Edit: April 08, 2013, 07:56:52 AM by xpatUSA » Logged

best regards,

Ted
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #24 on: April 08, 2013, 09:38:41 AM »
ReplyReply

I'm back, after a some more research, some experimenting with the sunflower and re-thinking the reasons behind the OP.

For me, my sunflower problem posted earlier occurred with ACR 5.4's conversion from X3F into it's working space. In that version of ACR what you see (all sliders 0) is what you got, assuming WB to be as good as it can be. That is to say, the conversion itself is fixed by Adobe. If there is a clipping problem caused by an image occupying a large gamut, that can not be fixed in ACR. That is to say, the image appears on the monitor clipped or blotched and, only when the sliders are set to extreme levels downward do the blotches go away - at which point the image looks stupid, washed out, a total failure. It may well be that the in-camera exposure was high for the shot (it was a one-off snap) but the RAW data itself was not clipped - I checked and re-checked.

On the other hand, Dave Coffin's 'dcraw' allows all sorts of adjustments to the raw data during the conversion, among them being saturation, brightness and gamma - not to mention a choice of output color spaces: raw, sRGB, aRGB, aWIDE, PP and XYZ. Not that I recommend dcraw particularly, it has a clunky command line interface and a somewhat protracted work cycle.

So it is that, armed with a better understanding of color spaces and profiles gained from this thread, I've managed to produce a blotch-free sunflower (again) but this time with some knowledge of how that was possible.

I no longer regard 'perceptual' as a kind of Grail, especially for one who only produces sRGB images and never prints anything. I will continue with what applications I have for now. Any 'difficult' capture will get either dcraw'ed or trashed; ACR->PSE for the rest . . ;-)

Perhaps I should mention that my serious cameras are RAW only (early Sigmas).

Ted,

Your revised image may be free from blotches, but IMHO the midtones are too dark and the three quarter tones in the center of the flower are severely lacking in detail. Furthermore, the reds are clipped in sRGB. You may underestimate the power of ACR and overestimate the advantages of dcraw. According to Eric Chan an image in ProPhotoRGB may be converted to whatever space you want with no loss of quality as long as you are in 16 bits. Gamma 2.2 is the preferred editing space, since it is approximately perceptually uniform.

It would be interesting if you would make your raw file available for others to try their hand with ACR and compare the results.

Regards,

Bill
Logged
elied
Sr. Member
****
Offline Offline

Posts: 253


« Reply #25 on: April 08, 2013, 11:40:28 AM »
ReplyReply

d.d. et alii,
A theoretical question: Would it be possible to work around the impossibility of Perceptual rendering with a matrix working space by converting first to a device space that is close in gamut to sRGB (my monitor's space, for instance, is 96% of sRGB) using Perceptual R.I. and then a second conversion to sRGB, or would the inaccuracies, quantization errors, etc. make it not worthwhile?
« Last Edit: April 08, 2013, 11:48:38 AM by elied » Logged

Roll over Ed Weston,
Tell Ansel Adams the news
Schewe
Sr. Member
****
Offline Offline

Posts: 5416


WWW
« Reply #26 on: April 08, 2013, 11:59:47 AM »
ReplyReply

A theoretical question: Would it be possible to work around the impossibility of Perceptual rendering with a matrix working space by converting first to a device space that is close in gamut to sRGB (my monitor's space, for instance, is 96% of sRGB) using Perceptual R.I. and then a second conversion to sRGB, or would the inaccuracies, quantization errors, etc. make it not worthwhile?

Nope...you would have to convert to an output profile space that contained perceptual rendering lookup tables (LUTs)...your display space is just another matrix based color space and only has Relative Colorimetric rendering...and what you would loose converting to some sort of real output profile whose gamut is described by the profile would make that trip pretty nasty for color.
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #27 on: April 08, 2013, 12:06:12 PM »
ReplyReply

Your revised image may be free from blotches, but IMHO the midtones are too dark and the three quarter tones in the center of the flower are severely lacking in detail. Furthermore, the reds are clipped in sRGB.

Thanks, Bill,

Yes, the image had no post-processing other than sharpening for posting. It was a quick test; one of several I had already made to check the embedded ICC profiles (or lack thereof). Blotch removal was the only point really. As to detail, the original raw is a LO res 0.8MP image because I was on my way to shoot a wrist watch on the bench. I do have a few more processed images; here's one I did a few minutes ago:



Quote
You may underestimate the power of ACR and overestimate the advantages of dcraw.

I must not have made myself clear, I wasn't pushing dcraw over ACR. I was trying to emphasize that ACR 5.4 has no options for the conversion per se from raw data to ACR's working space, whereas dcraw allows several.

Quote
It would be interesting if you would make your raw file available for others to try their hand with ACR and compare the results.

The raw file is attached for anyone to download and have a go.

P.S. I'm stuck at ACR 5.4 because that's the highest possible download for PSE6. I'm sticking with PSE6 because it's fast, it suits my purpose and I have it on a CD. So, although interested as always, it won't mean too much for me to be told about later versions, or ACR as it opens with more tabs in CS, or LR, etc.
« Last Edit: April 08, 2013, 01:27:33 PM by xpatUSA » Logged

best regards,

Ted
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #28 on: April 08, 2013, 04:43:09 PM »
ReplyReply

Yes, the image had no post-processing other than sharpening for posting. It was a quick test; one of several I had already made to check the embedded ICC profiles (or lack thereof). Blotch removal was the only point really. As to detail, the original raw is a LO res 0.8MP image because I was on my way to shoot a wrist watch on the bench. I do have a few more processed images; here's one I did a few minutes ago:

I must not have made myself clear, I wasn't pushing dcraw over ACR. I was trying to emphasize that ACR 5.4 has no options for the conversion per se from raw data to ACR's working space, whereas dcraw allows several.

The raw file is attached for anyone to download and have a go.

P.S. I'm stuck at ACR 5.4 because that's the highest possible download for PSE6. I'm sticking with PSE6 because it's fast, it suits my purpose and I have it on a CD. So, although interested as always, it won't mean too much for me to be told about later versions, or ACR as it opens with more tabs in CS, or LR, etc.

Ted,

Thanks for the raw file. The latest ACR is of little use here as the blotches appear when exposure is reduced. Even though the blotchy areas do not appear completely blown, I think that ACR highlight recovery is coming into play and it recovers to neutral, so the yellow is lost. DCraw has a better highlight recovery that looks at adjacent pixels and tries to maintain the highlight color. Look at Guillermo Luijk's essay on highlight recovery in DCraw (fig 9). Rawtherapee also has an option for this type of recovery.

I don't know much about Foveon sensors and hope some one else can clear up this behavior.

Regards,

Bill
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #29 on: April 08, 2013, 08:14:57 PM »
ReplyReply

Ted,

Thanks for the raw file. The latest ACR is of little use here as the blotches appear when exposure is reduced. Even though the blotchy areas do not appear completely blown, I think that ACR highlight recovery is coming into play and it recovers to neutral, so the yellow is lost. DCraw has a better highlight recovery that looks at adjacent pixels and tries to maintain the highlight color. Look at Guillermo Luijk's essay on highlight recovery in DCraw (fig 9). Rawtherapee also has an option for this type of recovery.

I don't know much about Foveon sensors and hope some one else can clear up this behavior.

Thanks for the link which I had downloaded some time back. But, re-reading, I did notice that I've been using BT 709 gamma (DCraw's default) and not the sRGB, so there's another thing to try on those dark bits in the middle of sunflower . .

dcraw -v -a -b 0.75 -o 1 -g 2.4 12.95 -q 3 -6 -T sf.X3F

The highlights in the top left petals, while not blown in the raw data, may well be in the less linear part of the F7 Foveon dynamic range. The camera outputs "saturation" metadata for each of the 3 channels, somewhere around 6,000 but, if you really over-expose, values up around 10,000 can be found in the raw data.

Interestingly, ExifToolGUI reveals that DCraw actually tacks a V2.1 perceptual profile on to it's output file, not that anything on my computer seems to do anything with it Wink.

TTFN,
« Last Edit: April 08, 2013, 11:28:28 PM by xpatUSA » Logged

best regards,

Ted
Ernst Dinkla
Sr. Member
****
Offline Offline

Posts: 2804


« Reply #30 on: April 09, 2013, 02:43:38 AM »
ReplyReply


On the other hand, Dave Coffin's 'dcraw' allows all sorts of adjustments to the raw data during the conversion, among them being saturation, brightness and gamma - not to mention a choice of output color spaces: raw, sRGB, aRGB, aWIDE, PP and XYZ. Not that I recommend dcraw particularly, it has a clunky command line interface and a somewhat protracted work cycle.

So it is that, armed with a better understanding of color spaces and profiles gained from this thread, I've managed to produce a blotch-free sunflower (again) but this time with some knowledge of how that was possible.



(Slightly sharpened, no other processing).

I no longer regard 'perceptual' as a kind of Grail, especially for one who only produces sRGB images and never prints anything. I will continue with what applications I have for now. Any 'difficult' capture will get either dcraw'ed or trashed; ACR->PSE for the rest . . ;-)

Perhaps I should mention that my serious cameras are RAW only (early Sigmas).

Qimage Ultimate uses dcraw for its RAW conversions. It may not have all dcraw cords to pull and the interface is not what most like but at least is not a command line program. Whether sRGB V4 is also used I did not check but it supports the ICC profile formats up to the latest v4.2 ICC standards. The extrapolation and sharpening tools are excellent too. Its control on anti-aliasing in downsampling may be of interest for web use.

The developer, Mike Chaney, has/had some early Sigma cameras and wrote articles on them that you probably know.

--
Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
December 2012, 500+ inkjet media white spectral plots.
Logged
jbrembat
Full Member
***
Offline Offline

Posts: 177


« Reply #31 on: April 09, 2013, 09:07:33 AM »
ReplyReply

Quote
Interestingly, ExifToolGUI reveals that DCraw actually tacks a V2.1 perceptual profile on to it's output file

dcraw include the HP sRGB profile (no perceptual intent). The tag you see is the tag of sRGB profile.


You get a conversion problem not a gamut problem.
Working with dcraw to get a ProPhoto image, the out of gamut colors are limited (dEmax=5.5, only 83 colors have  dE>=4).
Performing a gamut mapping from ProPhoto to sRGB, instead of clipping (that is using the sRGB profile), you have a slight difference not visible on my monitor.

Jacopo
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #32 on: April 09, 2013, 10:19:11 AM »
ReplyReply

dcraw include the HP sRGB profile (no perceptual intent). The tag you see is the tag of sRGB profile.

Thanks Jacopo,

I said that dcraw gives 'perceptual' based on ExifToolGui information. Here is a screen shot:



You can see that it says 'perpetual' and mentions 'auto-generated by dcraw' copyright. The profile shown is embedded in a TIFF made by dcraw V 9.02 and the source image file was of type X3F.

On the other hand, I see that you are showing us a screen shot of a non-embedded profile in system32\..\color\ folder. Please explain why, to help me to understand your post.

Quote
You get a conversion problem not a gamut problem.
Working with dcraw to get a ProPhoto image, the out of gamut colors are limited (dEmax=5.5, only 83 colors have  dE>=4).

Your measurement of delta-E's: what image were they from (I have post several) and what application was used?

Quote
Performing a gamut mapping from ProPhoto to sRGB, instead of clipping (that is using the sRGB profile), you have a slight difference not visible on my monitor.

What application was used to perform the gamut mapping? I would be interested in that.
« Last Edit: April 09, 2013, 10:28:03 AM by xpatUSA » Logged

best regards,

Ted
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #33 on: April 09, 2013, 10:53:58 AM »
ReplyReply

Qimage Ultimate uses dcraw for its RAW conversions. It may not have all dcraw cords to pull and the interface is not what most like but at least is not a command line program. Whether sRGB V4 is also used I did not check but it supports the ICC profile formats up to the latest v4.2 ICC standards. The extrapolation and sharpening tools are excellent too. Its control on anti-aliasing in downsampling may be of interest for web use.

Thanks for the tip, Ernst,

I don't print, so I had ignored Qimage. I've used UFraw in the past which is also a GUI that uses dcraw but without all the options.

You might be interested in this discussion elsewhere, where Mike Chaney gets a bit of a bashing!

http://www.dpreview.com/forums/thread/2840961#forum-post-35878372

I use dcraw mainly for research. For example, to get linear, high-gamut image files of lamp spectral emissions.

Otherwise my few images go straight to ACR and I accept what comes out (except for blotchy sunflowers!).

« Last Edit: April 09, 2013, 06:25:43 PM by xpatUSA » Logged

best regards,

Ted
jbrembat
Full Member
***
Offline Offline

Posts: 177


« Reply #34 on: April 09, 2013, 11:44:51 AM »
ReplyReply

Quote
On the other hand, I see that you are showing us a screen shot of a non-embedded profile in system32\..\color\ folder. Please explain why, to help me to understand your post.
On ICC ProfileInspector display you can read: "Intent: Perceptual".
This is from "Intent" tag of ICC profile.
This tag may not have any meaning until the profile is used in some context, e.g in a DeviceLink.
But the tag exists and a zero value is the encoding for "perceptual". You can find the same zero value into many profiles (Adobe1998, ProPhoto,....).
Quote
Your measurement of delta-E's: what image were they from (I have post several) and what application was used?
From your raw, converted using dcraw to get a ProPhoto image.
The application used is PhotoResampling. But the tool is not realased currently.
Quote
What application was used to perform the gamut mapping? I would be interested in that
.
Same answer as above.

Jacopo

Logged
Ernst Dinkla
Sr. Member
****
Offline Offline

Posts: 2804


« Reply #35 on: April 09, 2013, 12:30:57 PM »
ReplyReply

Thanks for the tip, Ernst,

I don't print, so I had ignored Qimage. I've used UFraw in the past which is also a GUI that uses dcraw but with out all the options. You might be interested in this discussion elsewhere, where Mike Chaney gets a bit of a bashing!

http://www.dpreview.com/forums/thread/2840961#forum-post-35878372

I use dcraw mainly for research. For example, to get linear, high-gamut image files of lamp spectral emissions.

Otherwise my few images go straight to ACR and I accept what comes out (except for blotchy sunflowers!).



People complaining about a financial solution that changed their update fee from zero for lifetime (Mike's? User's?) to 20 dollar a year 8-)
One wonders what they think about the Adobe update policy changes over the last years. Anyway Mike is a tough sparring partner in any discussion .......

Adding the sRGB V4 profile to the list for Qimage is all it takes.

Ernst, op de lei getypt
Logged
xpatUSA
Sr. Member
****
Offline Offline

Posts: 293



WWW
« Reply #36 on: April 26, 2013, 07:43:37 AM »
ReplyReply

Found one . . .

http://www.drycreekphoto.com/tools/profile_converter/
Logged

best regards,

Ted
Pages: « 1 [2]   Top of Page
Print
Jump to:  

Ad
Ad
Ad