Color Profiling Guide for Adobe Camera Raw 3

October 14, 2005

Update (July 2008)

For the time being I've decided to leave the original text of this article online.

Why profile?

Photographed with a Canon EOS 5D and 50 mm lens. The color of the foliage surrounding the waterfall improved noticeably after profiling.

I recently found an excellent online article by Bruce Fraser that describes the process of modifying ACR's built-in profiles. Essentially, the process involves photographing the 24-square GretagMacbeth ColorChecker Chart with a camera, then fiddling with color calibration controls in ACR to attempt to match the image values with known reference values.

A digital version of the ColorChecker created by Bruce Fraser.

I couldn't be more pleased with the results. I've made profiles for the conditions under which I shoot the most often: cloudy bright (about 6200 K) and indoor tungsten (about 2800 K). In both cases, I found that ACR's rendering of warm tones (reds, yellows) improved significantly. Improvements in other colors are less prominent, but the overall color balance is much better.

Toy example. Place your cursor over the image to see the result after profiling.

The comparison above illustrate these improvements. I photographed the original image with a Canon EOS 5D camera and a 50 mm lens. In the raw converter ACR, I set the white balance using the gray WhiBal cards seen in the left side of the images. I created the top image using ACR's built-in profile and found that the red jacket was very weak and lacked saturation, as did Tigger's face. Looking at the original scene and comparing it to what I saw on my calibrated monitor, I also noticed differences in the green pillow and the blue jacket, though these differences were relatively minor. After tweaking ACR's profile for the 5D using the procedure described below, I created the image on the bottom and found that it was far more accurate: the red jacket, blue jacket, green pillow, and orange Tigger were nearly spot-on. (Yes, the jacket is really that red!) The images you see above have been resized and converted to sRGB for web display.

Another advantage of going through this procedure is that my two cameras (a Canon EOS 5D and a Canon EOS 1D Mark II) now produce nearly identical results in color. No, not an exact match, but definitely close enough for my needs.

The purpose of this web page is to describe exactly the steps I used to perform the profile-tweaking in ACR; I've adapted Bruce Fraser's approach a bit. If accurate color reproduction is important to you (or if you have nothing else to do and feel that spending 10 minutes fiddling with calibration sliders in ACR sounds like fun), keep reading. I hope you find this guide helpful!

Profiling: step by step

The following steps explain how to modify the profiles using ACR 3.x; this version of Camera Raw is part of Photoshop CS 2. Note that this is a practical guide with little explanation of the theory and the principles. If you wish to learn more about color science and its relevance to photography, I suggest visiting Bruce Fraser's online articles and Bruce Lindbloom's web site.

You will need a GretagMacbeth ColorChecker Chart. This is a checkerboard containing 24 colored squares with known spectral properties. (No, it does NOT suffice to print out an image of the ColorChecker on your inkjet printer and use that instead!)

1. Expose the ColorChecker under your desired illumination conditions.
   • Be sure to expose in RAW mode.
   • Expose at a low ISO (e.g. 200).
   • Avoid nearby colored objects (e.g. grass, painted objects, etc.), because these may leave a color cast on the chart.
   • Avoid including shiny objects (or anything else brighter than the white square) in the frame.
   • Keep the illumination on the ColorChecker as even as possible.
   • Expose carefully to avoid clipping the white patch and the dark patch. Use your camera's histogram to assist you.
   • Stop down the lens (e.g. f/11 or smaller aperture) to minimize vignetting. It's reasonable to avoid completely filling the frame with the chart.

2. Open this reference image (which uses the ProPhoto RGB color space) in Photoshop. (This image is a digital version of the ColorChecker, prepared by Bruce Fraser by averaging LAB values for several physical ColorChecker charts and then converting the resulting file to ProPhoto RGB.)

   The reference image on your screen should appear similar to the image below. (I converted this image to sRGB for the purposes of displaying it on the web.)

   
3. Copy the RAW image of your photographed ColorChecker to the computer and open it in ACR.
   
4. Use the crop and align tools in ACR to exclude everything else except the 24 squares of the ColorChecker (i.e. eliminate all background material). The purpose of this step is to make the histogram accurately reflect the squares' tonal values. You may want to zoom in after performing this step to make the ColorChecker image easier to see.
   
5. Perform the following setup in ACR:
   • Choose "ProPhoto RGB" as the color space in ACR's workflow options.
   • In the Adjust tab, set all values except white balance (color and tint) to 0. That is, set the exposure, shadow, brightness, contrast, and saturation sliders to 0.
   • In the Detail tab, set sharpness to 0, luminance smoothing to 0, color noise reduction to a small value (e.g. 5).
   • Set all values in the Lens tab to 0.
   • In the Curve tab, choose "Linear" from the Tone Curve menu.
   • Set all values in the Calibrate tab to 0.
   

        

6. Look at the histogram and make sure that nothing is clipped, i.e. no overexposure and no underexposure. You can check the highlight and shadow warning boxex to make sure that the white square is not overexposed and that the dark square is not underexposed. If overexposure or underexposure has occurred, try another exposure of the ColorChecker with your camera.
   
7. Press S to get the sampler tool. Add "sampler points" to the middle of each square in the bottow row (the gray patches). Also add sampler points to the leftmost three patches in the third row (blue, green, red patches). Sampler points are a great new feature in ACR 3.
   
8. Set the white balance (use shift-click with the sampler tool, or press "I" to get the white balance tool) using the patch in row 4, column 2 (i.e. the gray patch to the right of the white square).
   

   Move the sampler tool over the surface of that gray patch (R4C2), keep an eye on the RGB readout values, and make sure that the RGB values are fairly neutral. They should not be more than one level apart (e.g. 189 190 190 is ok). If this is not the case, then there are two likely scenarios.

   First, you may need to enter a value manually into the Temperature box (see the right-most red outlined region in the figure above). Using the arrow keys moves the values in +50/-50 value increments. You may need to fine-tune the values by entering something in between (e.g. 2725 instead of 2700 or 2750). Another possibility is that there is a color cast, mixed illumination, or uneven illumination on the ColorChecker; in this case, it's probably best to start over and be more careful with the setup when photographing the ColorChecker.

9. Go to the Curve tab. Look at the sampler points that you added for the six gray patches. Ctrl-click on each one to add a corresponding point on the tone curve. There should now be six interior points on the curve, corresponding to the six gray patches.
   
10. While keeping an eye on the six gray patch sampler values near the top of the ACR window, adjust the six points on the tone curve so that the sampler values match the gray patches' reference values:
    241   190   145   104   67   37

    The easiest way to do this is to use ACR's keyboard shortucts. Use ctrl-tab (or shift-ctrl-tab) to select the next (or previous) control point, and use the up/down arrow keys to adjust the tone values. Hold down the shift-key to make increments of 10 instead of 1.

    If the patches' RGB values aren't perfectly neutral, that's ok; just match the green values.

    
11. Go to the Calibrate tab and check the dark patch (bottom-right, R4C6). If it's significantly non-neutral, adjust the Shadow Tint slider until it becomes neutral. Move your cursor (sampler) over the dark patch; the RGB values should not be more than 1 level apart (e.g. 37 37 38 is ok).
    
12. Now comes the tricky part: matching the blue, green, and red patches (R3C1, R3C2, R3C3). While an exact match is unlikely (I've never succeeded!), it's possible to get the color balance right, and that's what matters.

    Here is the iterative approach I used.

1. Start with the green patch and try to get the values to be proportional to the reference values. The references values for the green patch are (85, 123, 67). So, for instance, if the green sampler in the target image has a green value of 118, then aim for (82, 118, 64).

   Here is a table of green values to try to match (for greens in range 115 to 130):

   R	G	B
   79	115	63
   80	116	63
   81	117	64
   82	118	64
   82	119	65
   83	120	65
   84	121	66
   84	122	66
   85	123	67
   86	124	68
   86	125	68
   87	126	69
   88	127	69
   88	128	70
   89	129	70
   90	130	71

   Use the green saturation slider to adjust red and blue values relative to the green value. Use the green hue slider to adjust red and blue values relative to each other. Tweaking the hue slider will sometimes cause the green saturation value to change (e.g. from 118 to 117). That's ok. Just keep tweaking the hue and saturation sliders until the green sampler shows values that match one of the rows in the table above.

   For my example, here is what I got:

   

   Notice how my values for the green sampler (sampler 8) match one of the row (the sixth row) in the table above. For your own target image, the necessary green hue and saturation slider values may be entirely different from mine, so don't simply enter the slider values from my example!

2. Next, switch to the blue patch and its sampler. Repeat the process with the blue saturation and hue sliders. The reference blue values are (59, 48, 126). The following is the blue table for blues in the range of 95 to 135:

   R	G	B
   44	36	95
   45	37	96
   45	37	97
   46	37	98
   46	38	99
   47	38	100
   47	38	101
   48	39	102
   48	39	103
   49	40	104
   49	40	105
   50	40	106
   50	41	107
   51	41	108
   51	42	109
   52	42	110
   52	42	111
   52	43	112
   53	43	113
   53	43	114
   54	44	115
   54	44	116
   55	45	117
   55	45	118
   56	45	119
   56	46	120
   57	46	121
   57	46	122
   58	47	123
   58	47	124
   59	48	125
   59	48	126
   59	48	127
   60	49	128
   60	49	129
   61	50	130
   61	50	131
   62	50	132
   62	51	133
   63	51	134
   63	51	135

   Again, tweak the sliders until the blue sampler's values match one of the rows above. Here is what I got in my case:

   

   After this step, you'll probably find that the sampler values for the green patch have changed and may not be balanced properly. That's ok. Don't worry about this for now; we'll come back to it later.

3. Repeat the process for the red patch and the red saturation and hue sliders. The red patch's reference values are (122, 58, 46), and the table is:

   R	G	B
   115	55	43
   116	55	44
   117	56	44
   118	56	44
   119	57	45
   120	57	45
   121	58	46
   122	58	46
   123	58	46
   124	59	47
   125	59	47
   126	60	48
   127	60	48
   128	61	48
   129	61	49
   130	62	49
   131	62	49
   132	63	50
   133	63	50
   134	64	51
   135	64	51
   136	65	51
   137	65	52
   138	66	52
   139	66	52
   140	67	53
   141	67	53
   142	68	54
   143	68	54
   144	68	54
   145	69	55

   Here is what I got after tweaking the red hue and saturation sliders:

   

4. This completes one round of color slider tweaking in the Calibrate tab. As I mentioned above, the red sampler might now show a balanced set of values, but the green and blue sampler values are probably off. In my example above, you can see that this is indeed the case. They shouldn't be off by too much, though. By repeating the above three steps, we'll quickly converge on a balanced set of red, green, and blue patches.

   Fine-tune the sliders by repeating the above steps, starting with green, then blue, then red. In most cases you shouldn't need more than three or four iterations. Once you become comfortable with how the hue and saturation sliders work in the Calibrate tab, this iterative process should only take a few minutes.

   After two more iterations in my case, I converged to this:

   

   Notice how each of the red, blue, and green samplers match rows in the tables above. None of them match the reference chart exactly, though the blue and green patches come very close. The red values are a little high compared to the reference chart; oh well, we tried.

13. Save only the calibration settings by going to the right-triangle pop-up menu and choosing "Save settings subset ...".
    

    In the dialog box that comes up, choose "Calibration" from the Subset menu and click the "Save..." button.

    Give the target .xmp file a descriptive name, such as "Canon_5D_xxxxxx_profile_daylight_iso200.xmp", where "xxxxxx" is the serial number (or some other identifier) of your camera. (Having this identifier is handy in case you use two cameras of the same model.)

Evaluation

Digital sensors usually have noticeably different responses under daylight and tungsten illumination. If you shoot regularly under both of these conditions, I recommend creating separate profiles for daylight and tungsten illumination.

Here is an example of an image taken in daylight with the Canon 5D in daylight, before profiling. Place your cursor over the image to see the result after profiling.

Was it worth the effort? I think so. Place your cursor over the image to see the result after profiling.

Below you can see the reference chart (top image) and the physical chart photographed under a tungsten bulb with my 5D (bottom image), after profiling. I converted both of these images to sRGB.

Clearly there are some differences, most notably the light skin patch (R1C2), the blue sky patch (R1C3), the foliage patch (R1C4), and the bluish green patch (R1C6). Even so, this is a pretty good "visual" match.

My Canon 5D and 1D Mark II digital bodies now produce nearly identical color after I tweaked their profiles. Place your cursor over the bottom image above to see the chart photographed with my 1D Mark II, after profiling.