Since the latest versions of the well-known raw file analysis tool RawDigger developed by LibRAW LLC, the possibility of extracting colorimetric values from a more or less processed raw image has been incorporated. These values can be exported to a CGATS format document and used by tools such as ArgyllCMS.
Another improvement that gives rise to this workflow is the possibility of exporting TIFF images, or in other words, more or less processed raw images from RawDigger.
Based on this possibility, the photographer and developer Iliah Borg proposed a workflow compatible with his tool MakeInputICC.
MakeInputICC (Win and Mac) is a GUI for certain ArgyllCMS tools such as iccdump, profcheck, xicclu, txt2ti3 and of course colprof. The concept of this tool is very similar to my RoughProfiler, but MakeInputICC does not include scanin, so it cannot extract colorimetric values from images by itself; for that, it requires RawDigger.
Below I will detail step by step the workflow proposed by Iliah, as I found it to have some potential, but personally I do not fully agree with it, since from my point of view it is not especially consistent, an aspect I have already discussed with Iliah himself; however, perhaps some reader may find it useful.
Extraction of colorimetric values and development with RawDigger.
First, we go to the Selection menu and then Selection Grid. As its name indicates, it allows generating a grid to automate the extraction of color samples from a color patch pattern of a given color chart. If we work with Colorchecker, we can leave a default Grid Size of 4×6, and then adjust the ROI Relative Size to take more or less surface from each color patch.
Once the grid is located and adjusted over our chart, we activate Selection To Samples to load the color samples into a table. From here, the Save Samples option opens a new dialog where we must choose the CGATS format, StdDevValues, Average Green Channels, and in Data Scaling, problems arise. On one hand, we can select the gamma, the usual 2.2 or 1 if we want to work linearly; however, in Maximum Value, we face a certain paradox: the problem is that the RGB values provided by RawDigger are in 12 bits, and we need to scale them to 8 bits for ArgyllCMS to understand them. During this process, we must set the maximum RGB value corresponding to the maximum white of our color chart.
Obviously, this poses a certain dilemma since the RGB values depend on the color spaces, an aspect that at this point in our process does not exist. Therefore, we must test values, for example 242 for a supposed ProPhoto, or 245 for a supposed sRGB. This generates a certain inconsistency in our values that I will explain later.
Once our CGATS file is generated, we can proceed to export a TIFF image from our raw. For this, we go to the File menu, Export Tiff. From the export-development options panel, we choose at least Raw Composite, 3-Channel (RGB Output), and Half Size. Options like gamma 2.2 will depend on the value set when generating our CGATS file.
This generates a TIFF file with an sRGB color space. This is another point that does not convince me, as I suspect that this space is not merely “assigned” but constitutes the destination space of a certain process. This implies that our TIFF image is no longer scene-referred but rather referred to our sRGB space.
Working with MakeInputICC
The use of this tool is quite simple: On one hand, we load the reference file in CGATS format with Open Reference. The author himself has included several reference files for the Colorchecker Classic, Passport, SG, QPCard 201-202, and SpyderCheckr.
Next, we need to load the device data file, also in CGATS, generated with RawDigger. After the usual operations of choosing the algorithm, we can generate our input ICC profile by clicking Go!.
Finishing the process.
As we see, the process is quite simple; it is enough to assign our ICC profile to the image developed with RawDigger and evaluate the results.


After some attempts with this workflow modifying the Maximum Value, etc., I cannot get below a DE < 4, mainly encountering certain problems in tonal reproduction, as seen in the test graphs.
Conclusions
Although this workflow has been proposed to design input profiles for development tools that support ICC profiles (CaptureOne, DarkTable, etc.) as a means of device characterization, it seems to me hardly feasible to apply it precisely compared to other possible workflows, although in some way we can achieve not insignificant results.
Personally, the problem comes from the concept of “device data” as a set of values “relative to a device” for calculating our input ICC profile. It is worth remembering that ICC color management is of the Output Referred type, which roughly speaking, we could say is designed to work with processed images (TIFF, JPEG, etc.) and not raw files. To extract certain RGB values from a raw to form a CGATS-type document, a certain level of processing of our raw is necessary. This creates “device” values not only dependent on the device (e.g., camera) but also dependent on the algorithms and processes that generated those RGB values. Therefore, when we talk about ICC profiles for raw development processes, we are talking about profiles that characterize device and processing, or at least part of that processing. Certainly, in ICC colorimetry, an input profile characterizes a device, for example a scanner, since it produces output-referred images, whereas a camera working in raw produces input-referred images.
In other words, if we generate our document with the device reference colorimetry from a tool that performs raw development or interpretation processes, and use them to build a profile that will be used with another tool that has its own development processes, we face a certain inconsistency between reference values.
