This content has been marked as final. Show 6 replies
What settings? How did you shoot the target? What lighting? Have you checked the chromoholics web site? There's a user forum handled by Google that Tom sometimes posts to...other than the script, have you tried doing a manual calibration by eye?
I shot the target in direct sunlight at midday. I shot it out in the back yard but not "on" the lawn. In other words, it was not sitting on the surface of the grass. Actually, I was able to prop it upright on the branch of an evergreen bush. But it was not facing any part of the bush and was about 4 feet off the grass. So it did not appear to be placed in a position where it was likely to pick up a cast from the surrounding vegetation.
The settings that the calibration scripts gave me were roughly as follows. Where I am showing a range of numbers it was as a result of running both the Chromoholics and the Rags scripts.
Red hue(-20). Red saturation ranged from +25 to +45. Green hue ranged from zero to (-13). Green saturation ranged from zero to (+21). Blue hue ranged from +4 to +11. Blue saturation ranged from zero to (-21).
The best single set of readings that I got were red hue (-12), red saturation +23, green hue (0), green saturation +21, blue hue +4, blue saturation (-4). But even this set of settings appeared to render the red skintones overly saturated.
I did try a manual calibration where I attemtped to match the ProPhoto target RGB numbers. After a couple of hours I was fairly close but not exact with the red, green and blue primaries. However, many of the secondary patches were quite a ways off. I did not try just visually attempting to get a match. Would you suggest that?
The one consistent thing that I noticed in all my attempts was overly saturated skin tones when I applied the calibration settings to my photos. So I am a bit perplexed as I have not read anything that would indicate that others are having this problem. I did check the Chromoholics site and did not find anything that helped shed light on the problem.
Thanks for your help on this, I would really like to get it right.
If you could provide access to the raw image I might be able to offer some advice. I have a 5D test image that yields the following settings:
RedHue: -14 RedSaturation: +33
GreenHue: 0 GreenSaturation: -3
BlueHue: -2 BlueSaturation: +1
For the record, with this image the worst patches were light skin and red. And this is confirmed in the visual display. Overall the numbers (DE2K 2.0) and visual reflect a good calibration. I have tested many images from many cameras and find that the differences are usually in the shooting conditions, not the cameras. The Canon 5D is a fine camera.
With two each Canon 1D and Nikon D200 cameras, using the same camera model with the same settings and the same target, shot at the same time, the colors correlate extremely well.
You mention you are using a Munsell mini checker. I realize that the CC vendor has changed several times, but could this be an old target with different or faded patches? I recently shot three different CC24 targets in a single image (one a mini) and verified that all patches yielded the same color values. Your target is probably just fine, but the question remains. Since there is controversy about what the CC24 values should be, my script offers Gretag, Lindbloom, Fraser, and ProfileMaker 5 target options.
Shooting at high noon is not ideal. It can lead to uneven lighting. The best is with the sun at about 45 degrees high. Shooting near a large field of green is not ideal. The best is against a large gray background. There should be no backlight in the image field of view.
You will never achieve perfection. If the color matching functions in the raw image processor are not perfect to begin with, all a calibration script can do is to move any errors around. Its a zero sum game. Have you tried calibrating with light skin as the red patch?
My scripts report the errors as DE2K. Due to typical image color noise and histogram rounding, the best that can ever be achieved is about DE2K of 0.3. The best overall that I have seen from any camera in my test cases is 1.6.
Just for grins and my own education, I tried profiling with some borrowed equipment. I was enlightened with how much difference there can be simply measuring the same target multiple times. The documentation indicated that anything under DE2K 3.0 should be considered as good.
In closing, correct skin tones are really subjective. They will vary even with the same family members. Is your monitor calibrated?
Cheers, Rags :-)
I'd be happy to provide you with the raw image. Thanks for the offer. I received your email earlier and will load it up to your ftp server as you requested. My calibration settings were very close to yours.
Red Hue -14 Red Saturation +35
Green Hue 0 Green Saturation +5
Blue Hue +3 Blue Saturation -4
I hear what you are saying about the shooting conditions. Maybe that's what I am seeing. The target is brand new. I just purchased it from Samy's Camera. I re-shot it and used a gray background. The sun was lower in the sky as well like you suggested. Are you suggesting that I also run the script using the other target values like Lindbloom, etc.?
I hadn't thought of that but maybe it would also be worth a try.
Thanks for your assistance with this. As I said, I will upload the images to your server.
How important is the point curve setting. I have been using linear. But does it matter whether I use that or medium contrast?
Stick with the linear curve for calibration (only).
Happy New Year
Cheers, Rags :-)