One of the aspects of color management during RAW development that has triggered the most tutorials in recent years has perhaps been Adobe’s commitment to its well-known camera profiles or Adobe Camera Profile (DCP) created through the combined use of the Colorchecker chart with tools such as the Adobe DNG Profile Editor, or X-Rite’s own tool intended for this process. Recently, similarly through the QPCard 202 charts and the QPCalibration tool, it is possible to create good camera profiles.
However, of all that is usually discussed about these profiles, little has been said about the correction mechanisms they apply to the image, and the truth is that, since they are the result of a standard that is open yet proprietary, it is quite unknown where exactly they are interleaved during the RAW development process.
From what little has been inferred through rumors on the web, Adobe’s famous DCPs would work on color mapping between the RIMM and ROMM spaces, applying the appropriate transformations via 3×3 transformation matrices to each “hue” in the step from reality to image, to put it in a simplified way. If the process is indeed like this, certainly the concept, in my opinion, is elegant enough to show a certain respect towards this strategy.
In the previous paragraph, I highlighted the term “hue” as the literal translation of the English word “hue,” and not to be confused with the tonal reproduction curve that I will discuss later, since this is an important process to familiarize ourselves with and master this type of transformation or camera profiles.
DCP camera profiles essentially work on the chromatic aspect of color, that is, Hue, and have no effect on achromatic aspects such as Luminosity or Chromaticity. Hence, when applying a camera profile to a development, the changes are often minimal compared to working with ICC color profiles, where color transformations, or at least the look of our image, tend to be much more radical.
Therefore, camera profiles do not encompass all the potential necessary for good color correction in an image by themselves, requiring adjustments related to tonal mapping through the design of an appropriate tonal curve. Tonal curves, often simply called “curves” in the Adobe world, allow us to place each hue or group of hues in their proper place, for example, highlights, midtones, and shadows in their correct positions, so that our image comprehends all these situations coherently, meaning highlights are not midtones and midtones are not shadows, etc.
For tonal control of our scene, it is necessary to use grayscale scales, grayscale wedges, or tonal scales, that is, a scale where we have a sequence of neutral patches such as the grayscale of our Colorchecker or Kodak Q13, etc.
From a tonal scale, knowing the luminosity values of each patch, we can have a reference to approximate our curves during development, and to evaluate this process, we need to perform a tonal reproduction evaluation with a tool that allows us to extract an OECF curve from our scale and compare it with a reference. The concept and work with OECF analysis were previously addressed in the article “Evaluation of tonal reproduction with OECF measurements“.
To calculate the reference RGB values of our chart for a given color space, we must undergo a process as described in the article “ICCLU: Help with exposure and temperature control“ so that from CIELAB values obtained with a spectrophotometer, we can deduce RGB values through a specific color space. This is important because depending on the color space used in ACR, whether ProPhoto, AdobeRGB, or sRGB, these values will vary significantly. And if we want to deduce the luminosity value of each patch, we only need to apply the well-known formula:
Y = 0.2126*R + 0.7152*G + 0.0722*B
In this way, we obtain a single reference value instead of three. However, our ACR works with the three RGB colorants in its color samples, so it is necessary to perform a prior white balance to equalize the three channels and then, “by eye,” adjust the tonal curve values so that our color samples approximate the reference value. Evidently, this is a method with a small margin of error, but after evaluating the OECF of this process, we will see that the accuracy is very high.
To save the reader some calculations, I have deduced the reference values for the following color spaces for the grayscale of the Colorchecker:
| AdobeRGB | Prophoto | ||||||
| R | G | B | Y | R | G | B | Y |
| 244 | 244 | 238 | 244 | 240 | 241 | 235 | 241 |
| 199 | 201 | 200 | 200 | 188 | 189 | 189 | 190 |
| 159 | 160 | 160 | 160 | 143 | 144 | 144 | 144 |
| 117 | 118 | 118 | 118 | 98 | 99 | 99 | 99 |
| 84 | 85 | 85 | 84 | 65 | 66 | 66 | 66 |
| 54 | 53 | 54 | 53 | 37 | 37 | 38 | 38 |
If we wanted to be more rigorous and use a Kodak Q13 grayscale wedge, the values would be as follows:
| Prophoto | AdobeRGB | |||||||
| R | G | B | Y | R | G | B | Y | |
| 1 | 233,448801 | 234,237263 | 233,549896 | 234 | 237,082609 | 238,023869 | 237,243706 | 238 |
| 2 | 208,705156 | 208,95369 | 206,301248 | 209 | 216,829651 | 216,68822 | 214,121238 | 217 |
| 3 | 186,004424 | 186,34672 | 184,117787 | 186 | 197,253097 | 197,318688 | 195,092063 | 197 |
| 4 | 161,053139 | 162,310121 | 163,179164 | 162 | 174,424584 | 176,407863 | 177,032158 | 176 |
| 5 | 141,213078 | 141,567095 | 142,19882 | 142 | 156,918183 | 157,568793 | 158,140799 | 157 |
| 6 | 123,784745 | 124,503917 | 125,467356 | 124 | 140,638578 | 141,924808 | 142,783654 | 142 |
| 7 | 108,872879 | 109,696232 | 110,497866 | 110 | 126,544601 | 127,988204 | 128,674286 | 128 |
| 8 | 95,663527 | 96,456741 | 97,663728 | 96 | 113,690193 | 115,208185 | 116,357481 | 115 |
| 9 | 84,49379 | 85,161644 | 85,744728 | 85 | 102,821419 | 104,032574 | 104,540511 | 104 |
| 10 | 72,907346 | 73,440096 | 73,676297 | 73 | 91,20329 | 92,143832 | 92,298025 | 92 |
| 11 | 64,11421 | 64,495639 | 65,00828 | 64 | 82,057509 | 82,826422 | 83,341537 | 83 |
| 12 | 56,175588 | 56,573665 | 56,78516 | 57 | 73,669739 | 74,414828 | 74,577585 | 74 |
| 13 | 48,853793 | 49,183789 | 49,658454 | 49 | 65,644729 | 66,351153 | 66,857579 | 66 |
| 14 | 43,229955 | 43,477021 | 44,105162 | 43 | 59,356727 | 59,965616 | 60,695814 | 60 |
| 15 | 37,893433 | 38,113115 | 38,4978 | 38 | 53,327992 | 53,837949 | 54,279378 | 54 |
| 16 | 33,123443 | 33,323713 | 33,85773 | 33 | 47,708766 | 48,233328 | 48,887112 | 48 |
| 17 | 30,421876 | 30,480615 | 31,68513 | 31 | 44,362825 | 44,800069 | 46,393983 | 45 |
| 18 | 26,648271 | 26,375279 | 27,087586 | 26 | 40,0579 | 39,704864 | 40,754847 | 40 |
| 19 | 22,197355 | 21,910105 | 22,642953 | 22 | 34,480616 | 34,091827 | 35,210078 | 34 |
| 21 | 21,41453 | 21,094121 | 21,7581 | 21 | 33,515766 | 33,035702 | 34,072667 | 33 |
Thus, the values in the Y column will be what we should use as reference or, in Anglo-Saxon terms, “aimpoints.”
If after tonal adjustment, we evaluate our process colorimetrically and tonally, we can observe significant precision indices at the Delta-e or tonal mapping level.
On the left, we can see a tonal reproduction evaluation report reviewing the OECF. Although barely visible in the capture, each graph consists of lines, one of measured values and another of reference; when these two curves overlap, it indicates that the tonal reproduction is exact. Conversely, if deviations appear in any area of the curve, it is a sign of deviation in highlights, midtones, or shadows that we must compensate with the tonal curve.
In this second graph, we see an assessment of the colorimetric accuracy of this development process using a DCP camera profile and fine tonal reproduction adjustment. The result is very low deltas, around ΔE* 2 for CIE73 and ΔE* 1.3 for CIEDE00. This difference between the two formulas and the optimism of CIEDE00 indicates an error more related to luminosity or chromaticity than to hue; in fact, it can be seen that Δa and Δb are smaller than ΔL or ΔC. Even so, our process is colorimetrically within very high acceptance thresholds.
