On Tue, 04 Jun 2013 13:18:48 -0400 Iliah Borg wrote:
3. Use dcpTool. (http://dcptool.sourceforge.net/Introduction.html) This command-line driven software can decompile binary format DNG profiles into XMP, compile XMP files into new compiled DNG profiles, make a profile "invariate" (an invariate profile won’t cause changes in tint when you make adjustments to exposure settings), and untwist a profile. (An invariate profile won’t change color when you adjust exposure settings, but still has hue twists embedded within it. Untwisting a profile removes all hue twists completely. See http://dcptool.sourceforge.net/Hue Twists.html)
With dcpTool you can reduce the "DNG profile" (DNG Camera Profile, DCP) to a simple matrix (ColorMatrix), however in ACR/LR white balance is still applied using ColorMatrix interpolation or extrapolation. Unless ad hock dcp is calculated for the light in the scene the result of interpolation/extrapolation is unpredictable in terms of colour.
Iliah, Of course, DNG profiles can be created from a single illuminant with both the ColorChecker Passport software and the DNG Profile Editor, where both matrices end up identical. dcpTool can do a lot more than just reduce a DNG profile to a simple color matrix. TheHueSatDelta tables in DNG profiles have third dimension, value, that in the past was almost always set to 1, but that Adobe now uses to implement the “hue twists” in their new generation profiles. There is also a LookTable, which is exactly the same as a HueSatDelta Table, but is intended to be applied later in the processing pipe, “after any exposure compensation and/or fill light stages, but before any tone curve stage” that can also apply hue twists. dcpTool can make a profile Invariate. Making a profile ivariate shifts any hue twists to before the basic exposure adjustments stage, so they are applied immediately. Therefore an invariate profile won’t cause changes in tint when you make adjustments to exposure settings. This is a way to modify the standard Adobe profiles if you want to keep the twists, but just prevent colors from changing when you adjust Carmera Raw or Lightroom’s exposure controls. Making a profile ivariate moves any LookTable to the HueSatDelta table. Specifically, if a LookTable exists, and no HueSatDelta tables exist, dcpTool simply copies the LookTable to the HueSatDelta tables, and the removes the original LookTable. If both a LookTable and HueSatDelta tables exist, the tables are mathematically combined, and the combination becomes the new HueSatDelta table. Again, the original LookTable is then removed. If no LookTable exists, then the profile is already invariate. dcpTool can also Untwist a profile entirely. Untwisting a profile removes all hue twists completely. An untwisted profile will always keep colors true, regardless of original in-camera exposure, or changes to exposure controls in Camera Raw or Lightroom. Untwisting a profile removes all intensity (V in HSV) dependencies in both the LookTable and any HueSatDelta tables. Effectively, the twist at a single point in the twist table is chosen, and applied to all intensities. The twist chosen is that corresponding to a skin tone that is accurately exposed, thus preserving the profile's skin tone. I'm particularly looking for feedback from people who have been using this tool, as appropriately modified DNG profiles would appear to effectively deal with many of the complaints about LR/ACR and colorimetric accuracy. Thanks, --Rich Wagner
On Jun 5, 2013, at 11:06 AM, Rich Wagner wrote:
On Tue, 04 Jun 2013 13:18:48 -0400 Iliah Borg wrote:
3. Use dcpTool. (http://dcptool.sourceforge.net/Introduction.html) This command-line driven software can decompile binary format DNG profiles into XMP, compile XMP files into new compiled DNG profiles, make a profile "invariate" (an invariate profile won’t cause changes in tint when you make adjustments to exposure settings), and untwist a profile. (An invariate profile won’t change color when you adjust exposure settings, but still has hue twists embedded within it. Untwisting a profile removes all hue twists completely. See http://dcptool.sourceforge.net/Hue Twists.html)
With dcpTool you can reduce the "DNG profile" (DNG Camera Profile, DCP) to a simple matrix (ColorMatrix), however in ACR/LR white balance is still applied using ColorMatrix interpolation or extrapolation. Unless ad hock dcp is calculated for the light in the scene the result of interpolation/extrapolation is unpredictable in terms of colour.
Iliah,
Of course, DNG profiles can be created from a single illuminant with both the ColorChecker Passport software and the DNG Profile Editor, where both matrices end up identical.
dcpTool can do a lot more than just reduce a DNG profile to a simple color matrix.
I know. I just do not see it solving colorimetric renedering.
I'm particularly looking for feedback from people who have been using this tool
I do.
, as appropriately modified DNG profiles would appear to effectively deal with many of the complaints about LR/ACR and colorimetric accuracy.
Addressed it. Apart from being far from colorimetrically accurate dcp profiles with extra parameters introduce additional noise and artifacts (smoothing of LUTs leave a lot to be desired). CC24 is too small for any colorimetric accuracy anyway. -- Best regards, Iliah Borg
On Jun 5, 2013, at 8:15 AM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
CC24 is too small for any colorimetric accuracy anyway.
Not only that, but there's another elephant in the room. Two of them, in fact. Essential to creating a good profile is being able to get to the unprofiled data, including before white balancing and gamma and exposure correction and what-not as well as after but before anything else has been done. Until you can get to that kind of data (which you easily can with any of the ``alternative'' raw developers), the whole issue is rather moot. ...aaaaannnnnddd...I'm not aware of any profile-building tools that know about DNG. That is, anything that can extract RGB values from a photo of a chart, compare them with expected absolute color values, and output the matrix / LUT / whatever for the DNG. That's the heart and soul of profiling, when it comes right down to it. If you don't have that, whatever you're doing isn't color profiling. (Of course, there's the tool that ships with the ColorChecker Passport, but it only uses the 24 patches -- and the same deal with Adobe's DNG Profile Editor.) Sure, you can kinda fake it with DNG by using the eyedropper on your chart, but just for the classic ColorChecker that's an order of magnitude more work than is involved in even creating a printer profile from a chart with several thousand patches. Can you imagine the work involved in attempting to manually create a DNG profile just with all of a ColorChecker Passport's 50 patches? Cheers, b&
Essential to creating a good profile is being able to get to the unprofiled data, including before white balancing and gamma and exposure correction and what-not as well as after but before anything else has been done. Until you can get to that kind of data (which you easily can with any of the ``alternative'' raw developers), the whole issue is rather moot.
...aaaaannnnnddd...I'm not aware of any profile-building tools that know about DNG. That is, anything that can extract RGB values from a photo of a chart, compare them with expected absolute color values, and output the matrix / LUT / whatever for the DNG. That's the heart and soul of profiling, when it comes right down to it. If you don't have that, whatever you're doing isn't color profiling. (Of course, there's the tool that ships with the ColorChecker Passport, but it only uses the 24 patches -- and the same deal with Adobe's DNG Profile Editor.)
I made such a tool for myself, it does all the above and even works over SG or spectral measurements. I've spent some serious time optimizing it. So I know the internal limitations of the dcp implementation. -- Best regards, Iliah Borg
On Jun 5, 2013, at 8:31 AM, Ben Goren wrote:
Essential to creating a good profile is being able to get to the unprofiled data, including before white balancing and gamma and exposure correction and what-not as well as after but before anything else has been done. Until you can get to that kind of data (which you easily can with any of the ``alternative'' raw developers), the whole issue is rather moot.
Ben, Are you implying that a developer of DNG profiling software cannot get access to the "raw data" using the DNG SDK? If so, I don't think that's accurate. While it's true that the end user might not have access to this data, developers certainly should. How would you propose that the CC Passport software works to build de novo profiles? Have you personally used dcpTool to untwist DNG profiles and rigorously test their accuracy to compare with your "other methods?" If you have, I would be interested in your results. There are a lot of reports of photographers getting excellent results with this method when looking to accurately reproduce color in studio settings. As Eric Chan wrote roughly 4 years ago, "If accurate scene colorimetry estimates were all that photographers wanted, a 3x3 matrix (or set of 3x3 matrices) would suffice, possibly with a 2D table appended. Experience, however, has shown that photographers generally don't want that (witness the popularity of certain film stocks which also twist significantly)." http://forums.adobe.com/message/1209891 I'm personally not willing to give up the ease of use of ACR/LR right now for any other RAW processor, because whatever advantages they may offer in so-called "colorimetric accuracy" are far offset by workflow issues. In addition, in my experience with CC Passport built DNG profiles, they are more than adequate for most uses. Art repro is a different matter and an interesting topic, but that's not what most of us do. --Rich
participants (3)
-
Ben Goren
-
Iliah Borg
-
Rich Wagner