Raw Conversion - Color Space Issues
Please note that this post was originally drafted as a response to ChrisM's contribution to the recent Pentax K-3 debate (now locked). The issue was the relative strengths and weaknesses of Cap1 and DxO, particularly in relation to color integrity and color spaces.
Hello Chris
Well, your recent post has given my little grey cells a bit of a workout! Thank you for taking the trouble. Bearing in mind that clearly you know more about this topic than I will ever do, perhaps you could comment on my simple-minded summary below.
1. DSLR sensors are designed so that each pixel has an equal response to the range of energy of visible light. And the same range applies regardless of the in-camera white balance setting and the color space setting (sRGB or Adobe RGB). Perhaps the sensor ‘space’ can be thought of as a 2-dimensional spreadsheet.
2. The color energy information stored in each pixel will vary according to the image. Perhaps this information can be visualised as a third dimension rising from the pixel spreadsheet surface. But with one difference – the pixel space is finite and fixed, while the energy information is open-ended. All this information is stored in an untreated raw file.
3. Raw conversion software has to make sense of the color energy information. There are two constraints. The first is a decision where to draw the line with the energy information so that the mathematical treatment is in a closed space rather than an open-ended one. So the software creates some kind of energy/color envelope within which it can operate. Everything outside this envelope is dumped. We call this clipping. One application of clipping is to deliberately make the envelope smaller so that raw processing is quicker.
4. In addition to this deliberate dumping/clipping for mathematical purposes software can impose clipping in anticipation of the way the output file (tif, etc.) will be processed. For example cheaper monitors and printers will have a smaller color envelope (gamut) than more expensive ones. So some of the raw processing envelope may be wasted on them. Hence a deliberate decision to impose arbitrary clipping. It is this, I think, that you say DxO is doing to the detriment of color in tif outputs. Whereas Cap1 imposes little of no clipping of this kind.
You identified yellows as being the main problem in Adobe space. This is interesting because, although we think of fertile rural landscapes as being green, my experience is that the dominant color is in the yellows. So perhaps I have been correct in avoiding Adobe RGB? But, I confess, I am an habitual user of sRGB.
However, thanks to your post and a perusal of the Joseph Holmes articles I will now switch to Prophoto and see what difference it makes to some archived images of mine. I will also look into the mechanics of downloading and installing DCAM 4, which you recommend. (Any further advice?)
All of which sheds new light on how I regard DxO and, more importantly, Cap1. It is ironic that DSLR and sensor manufacturers go to such lengths to convert electromagnetic radiation into a visible representation of the color range we can see, only for one software company (DxO) to simply dump some of it – if I have understood the situation correctly. How on earth have they kept this quiet for so long? Perhaps because many of us (I include myself) have not spent £1000+ on monitors and printers? Perhaps because many people (I exclude myself) are absorbed in the current trend for high-contrast/HDR-like images? Perhaps because many fine-art photographers (I hesitantly include myself) work in black and white, where color tones can be tweaked at will?
I will shortly be posting a Cap1 article on integrating Cap1 user styles into a color-oriented workflow. Hopefully you will find it of interest.
Best wishes.
Peter
Hello Chris
Well, your recent post has given my little grey cells a bit of a workout! Thank you for taking the trouble. Bearing in mind that clearly you know more about this topic than I will ever do, perhaps you could comment on my simple-minded summary below.
1. DSLR sensors are designed so that each pixel has an equal response to the range of energy of visible light. And the same range applies regardless of the in-camera white balance setting and the color space setting (sRGB or Adobe RGB). Perhaps the sensor ‘space’ can be thought of as a 2-dimensional spreadsheet.
2. The color energy information stored in each pixel will vary according to the image. Perhaps this information can be visualised as a third dimension rising from the pixel spreadsheet surface. But with one difference – the pixel space is finite and fixed, while the energy information is open-ended. All this information is stored in an untreated raw file.
3. Raw conversion software has to make sense of the color energy information. There are two constraints. The first is a decision where to draw the line with the energy information so that the mathematical treatment is in a closed space rather than an open-ended one. So the software creates some kind of energy/color envelope within which it can operate. Everything outside this envelope is dumped. We call this clipping. One application of clipping is to deliberately make the envelope smaller so that raw processing is quicker.
4. In addition to this deliberate dumping/clipping for mathematical purposes software can impose clipping in anticipation of the way the output file (tif, etc.) will be processed. For example cheaper monitors and printers will have a smaller color envelope (gamut) than more expensive ones. So some of the raw processing envelope may be wasted on them. Hence a deliberate decision to impose arbitrary clipping. It is this, I think, that you say DxO is doing to the detriment of color in tif outputs. Whereas Cap1 imposes little of no clipping of this kind.
You identified yellows as being the main problem in Adobe space. This is interesting because, although we think of fertile rural landscapes as being green, my experience is that the dominant color is in the yellows. So perhaps I have been correct in avoiding Adobe RGB? But, I confess, I am an habitual user of sRGB.
However, thanks to your post and a perusal of the Joseph Holmes articles I will now switch to Prophoto and see what difference it makes to some archived images of mine. I will also look into the mechanics of downloading and installing DCAM 4, which you recommend. (Any further advice?)
All of which sheds new light on how I regard DxO and, more importantly, Cap1. It is ironic that DSLR and sensor manufacturers go to such lengths to convert electromagnetic radiation into a visible representation of the color range we can see, only for one software company (DxO) to simply dump some of it – if I have understood the situation correctly. How on earth have they kept this quiet for so long? Perhaps because many of us (I include myself) have not spent £1000+ on monitors and printers? Perhaps because many people (I exclude myself) are absorbed in the current trend for high-contrast/HDR-like images? Perhaps because many fine-art photographers (I hesitantly include myself) work in black and white, where color tones can be tweaked at will?
I will shortly be posting a Cap1 article on integrating Cap1 user styles into a color-oriented workflow. Hopefully you will find it of interest.
Best wishes.
Peter
0
-
Some interesting reading on DxO's decision here:
http://www.luminous-landscape.com/forum ... ic=71979.0
More here:
http://www.luminous-landscape.com/forum ... ic=54284.0
I don't think it's a matter of DxO "keeping this quiet" so much as (with all due respect to Chris for whom this is clearly a problem) not being much of an issue for most users - rightly or wrongly.
Even so, I don't see the logic of not using a wide gamut internal colour space like ProPhoto.0 -
Keith - this has been an interesting read. Thanks.
I may have to experiment a bit more myself.
That said, most of my images are shown or shared on sRGB monitors. Even some of the commercial printers ask for sRGB space.
This has bothered me at times, but it bothers me a bit more now.
Jim0 -
Peter wrote:
Please note that this post was originally drafted as a response to ChrisM's contribution to the recent Pentax K-3 debate (now locked). The issue was the relative strengths and weaknesses of Cap1 and DxO, particularly in relation to color integrity and color spaces.
Hello Chris
Well, your recent post has given my little grey cells a bit of a workout! Thank you for taking the trouble. Bearing in mind that clearly you know more about this topic than I will ever do, perhaps you could comment on my simple-minded summary below.
1. DSLR sensors are designed so that each pixel has an equal response to the range of energy of visible light. And the same range applies regardless of the in-camera white balance setting and the color space setting (sRGB or Adobe RGB). Perhaps the sensor ‘space’ can be thought of as a 2-dimensional spreadsheet.
2. The color energy information stored in each pixel will vary according to the image. Perhaps this information can be visualised as a third dimension rising from the pixel spreadsheet surface. But with one difference – the pixel space is finite and fixed, while the energy information is open-ended. All this information is stored in an untreated raw file.
3. Raw conversion software has to make sense of the color energy information. There are two constraints. The first is a decision where to draw the line with the energy information so that the mathematical treatment is in a closed space rather than an open-ended one. So the software creates some kind of energy/color envelope within which it can operate. Everything outside this envelope is dumped. We call this clipping. One application of clipping is to deliberately make the envelope smaller so that raw processing is quicker.
4. In addition to this deliberate dumping/clipping for mathematical purposes software can impose clipping in anticipation of the way the output file (tif, etc.) will be processed. For example cheaper monitors and printers will have a smaller color envelope (gamut) than more expensive ones. So some of the raw processing envelope may be wasted on them. Hence a deliberate decision to impose arbitrary clipping. It is this, I think, that you say DxO is doing to the detriment of color in tif outputs. Whereas Cap1 imposes little of no clipping of this kind.
You identified yellows as being the main problem in Adobe space. This is interesting because, although we think of fertile rural landscapes as being green, my experience is that the dominant color is in the yellows. So perhaps I have been correct in avoiding Adobe RGB? But, I confess, I am an habitual user of sRGB.
However, thanks to your post and a perusal of the Joseph Holmes articles I will now switch to Prophoto and see what difference it makes to some archived images of mine. I will also look into the mechanics of downloading and installing DCAM 4, which you recommend. (Any further advice?)
All of which sheds new light on how I regard DxO and, more importantly, Cap1. It is ironic that DSLR and sensor manufacturers go to such lengths to convert electromagnetic radiation into a visible representation of the color range we can see, only for one software company (DxO) to simply dump some of it – if I have understood the situation correctly. How on earth have they kept this quiet for so long? Perhaps because many of us (I include myself) have not spent £1000+ on monitors and printers? Perhaps because many people (I exclude myself) are absorbed in the current trend for high-contrast/HDR-like images? Perhaps because many fine-art photographers (I hesitantly include myself) work in black and white, where color tones can be tweaked at will?
I will shortly be posting a Cap1 article on integrating Cap1 user styles into a color-oriented workflow. Hopefully you will find it of interest.
Best wishes.
Peter
Hello Peter,
I also noticed the post on Pentax K3 support was locked, so I placed a new post in the windows section of CO1, but Drew rightly directed me to your post here.
I've been in contact with DxO support, and they did go through the trouble to look into this issue. The situation seems a little bit different than I assumed. Shortly, with raw image processing you go from input profile > working space color gamut > output profile.
Now, imaging software does not necessarily convert the input data to the working space gamut permanently. Photoshop offers you the option to convert to the working space or keep the existing profile. If you convert to the working space gamut, than any clipping is irreversible. I could not get my head around why DxO would convert to a (much) smaller Adobe rgb-like working space, when the digital camera color space is generically so much larger. This is apparently not what DxO is doing however. Simply put: they calibrate all their supplied input profiles (camera default, but also "neutral color, neutral tonality"), to the Adobe rgb color space. That is why no image rendered by DxO will ever contain data outside of the Adobe color space, when using their supplied input profiles. This is intended design, and they don't keep quiet about this, but there is some ill logic behind this in my opinion nonetheless. They feel that, as most of their users edit and process their images in either the sRGB or Adobe rgb color space, they should supply their input profiles optimized for these color spaces. the ill-part of the logic, is that this would only hold true for 8 bit (jpeg) images, which is not what raw conversion is about. As long as you keep an image in 16 bit, there will never be image artifacts caused by conversion from a very large camera color space to whatever other color space. So the decision to start off small has no advantages whatsoever.
DxO does support another option though: the one under "icc profile" in the input tab. If you want to work with larger color spaces when processing your images, you will have to supply your own input .icc camera profile. I tested this by importing the Pentax K5IIs profile from CO1 into DxO, and yes, this will allow use of the large camera native color space (as Phase one supplies it) and convert to a large output space like DCAM 4. But the CO1 profile doesn't look optimal at all in DxO, and creating your own camera profile requires more work and investment in high end software/hardware than I care for.
Regarding your post: the visible part of the color spectrum has been defined in the so called spectrum locus, a two dimensional shape, where three primaries are plotted to define the boundaries of a color space: red, green and blue (rgb). The blue and red primaries that the Adobe color space uses, are the same as the primaries used in the sRGB color space. Only the green primary was shifted outwards in the spectrum locus to allow for more vivid greens, as well as for finer gradients in the displaying and printing of (mainly) green colors. Apart from the positioning of the primaries, color spaces also employ a tone curve: black and white images will be displayed in the three dimensional graphics of a program like "colorthink" (that allows you to view a graphic representation of an image with overlay of a color space), as an almost vertical line from the white point in the spectrum locus upwards to black. The reason I chose DCAM 4, is that it is a similar to ProPhoto colorspace, but doesn't use the 1.8 gamma tone curve that always gave me problems with shadows in printing. The tone curve is much better defined in Joseph Holmes DCAM 4 color space. By the way: Holmes also offers the DCAM 1,2,3 and 5 color spaces, but 5 is not really a practical space, and as long as you stay in 16bit, there is no benefit to using the smaller spaces, other than that the tone curve will be slightly more refined for images that have small gamuts.
As to how important larger color spaces are? It depends on your printer and display of course, but what is often overlooked, is that Color spaces are three dimensional, and especially the dedicated A3 or larger photo printers can print quite a few colors that are outside Adobe rgb, let alone sRGB. On top of that, the coming generations of LED backlit (wide gamut) screens are capable of showing many colors outside of the standard color spaces, even though they will be designed to conform to either of these two dominant color spaces.
This will certainly not be of paramount importance with most images, but there is just no real reason to place the color space primaries away from the outer (most vivid) edges at an early stage in raw processing.
Joseph Holmes has a lot of sensible things to say on color spaces by the way.
He feels that the use of properly designed color spaces has not evolved the way it should have, and it is hard not to agree with him. sRGB, the "web standard" was designed to create a standard for displaying digital images on CRT monitors. Now that we all use LCD monitors, that have totally different characteristics, this sRGB standard is still in place. Adobe RGB was also a rather arbitrary invention by Adobe, keeping the red primary of sRGB, and consequently seriously confining the very important (to nature photography) yellow region of the color spectrum. My calibrated Macbook pro (led) screen can show yellows well outside of both sRGB and Adobe RGB.
Capture one handles the raw workflow exactly as it should be, I feel myself it is a pity that DxO does not follow the same logic workflow, although it leaves a small backdoor open to use your own custom color spaces (input and output).
Even if the adopted standards like sRGB and Adobe RGB (also used in designing high end screens, as a target for coverage like 98% Adobe rgb) will never allow us to see wider color spectrums, especially with Raw image data I like to keep the color spaces large.
regards
Chris0 -
ChrisM wrote:
This is apparently not what DxO is doing however.
Or then again, perhaps it is:
http://www.luminous-landscape.com/forum ... #msg571280
Or was, anyway...0 -
Keith Reeder wrote:
ChrisM wrote:
This is apparently not what DxO is doing however.
Or then again, perhaps it is:
http://www.luminous-landscape.com/forum ... #msg571280
Or was, anyway...
Hello Keith,
Thanks for the link, very informative.
I don't believe it's the working color space that is crucial here. You can use working spaces without them influencing the actual output, as long as the source data is not converted to the working space profile on processing for output. Think of Photoshop in this regard: you can set it up to maintain the source profile and only convert for proper displaying, without affecting the output.
What I don't read anything about in the luminous landscape topic, is how DxO supplies and handles the source data, that you can select on the input tab. It turns out, according to DxO support, that all these profiles are supplied or handled in the Adobe rgb color space. This, again according to support, does not apply to possible custom .icc profiles that you have to supply yourself and then can select in the input tab. Conversion should be straightforward, i.e. directly to the output profile. Whether or not DxO still uses Adobe rgb as a working space in this case, is irrelevant to me: I won't ever supply my own camera profile. This all renders DxO pretty much useless for me, unless I know for certain the color gamut of (a series of) images does not extend beyond Adobe rgb.
Regards
Chris0 -
Hello again
Well, my little grey cells are getting greyer! So thank you all for these inputs.
I have to admit that the more I look into the way color spaces have evolved and the arbitrary choice in workflow that they present the more it seems to be a combination of pragmatic suck it and see and quasi-science by people who have a vested interest. Sleight of hand may sum it up! But I admit that much of the detail is over my head. All I know (I think!) is that I should be using ProPhoto in future. In which case, why are sRGB and the Adobe version still being imposed by photo software? (I have read that ProPhoto is not very good with 8-bit files - posterisation problems? But I always use a 16-bit, uncompressed tif workflow.)
It does seem that if you don't have a high end monitor and printer one colour space is as good as another - as long as you stick to it throughout the workflow. I have done it all - from camera calibration to monitor to printer. I don't think I achieved any major improvements (apart from printer output color) except in one regard. Cap1 processing of raw RAW files seems to produce greater tonal depth and robustness to tweaking than other raw converters (e.g. ACR, DxO and Silkypix) and all raw converter tif outputs benefit from a quick review and tweak in LR5.
The main conundrum for me is that even if I had the perfect color-consistent workflow I would still want to make my own adjustments to get the 'look' I want. How self-defeating can I get? However if anyone wants to have a go at calibrating their camera and lens I can suggest QPCard 503, which is relatively inexpensive - and a useful learning experience.
I recently came across a website 'Cambridge in Colour'. It is pitched at a level that encourages me to think that I can understand it.
Thanks again.
Peter0
Please sign in to leave a comment.
Comments
6 comments