I have a color weakness. Is there any tool or tools that I can use by using a formula or setting to correct skin color in Photoshop?
-
1\$\begingroup\$ There are some guidelines in my answer to the following question that may help: photo.stackexchange.com/questions/41928/… \$\endgroup\$– MikeW ♦Jan 21, 2015 at 3:12
-
1\$\begingroup\$ Use LAB color mode. Dan Margulis' book says human skin will always be on the red side of the red-green axis, and gives broad guidelines for cast-free skin values. Certainly can match a known good picture by the numbers, especially in LAB. \$\endgroup\$– JDługoszJan 21, 2015 at 6:33
-
1\$\begingroup\$ What is the deficiency? I understand red-green is by far the most common and often affects men (not women). I ask because it can make a difference to specific ideas. \$\endgroup\$– JDługoszJan 21, 2015 at 6:36
-
3\$\begingroup\$ Maybe using some kind of ColorChecker will help more than formula :) \$\endgroup\$– Romeo NinovJan 21, 2015 at 6:53
-
\$\begingroup\$ worth looking at: photo.stackexchange.com/questions/3635/… \$\endgroup\$– chuquiMar 22, 2015 at 15:32
3 Answers
I think I found something interesting. It feels like a revelation to me!
There are some articles about using the eyedropper but on CMYK model.
I found a very, very simple method that I feel works well to adjust skin colors on RGB model. I am currently testing on Caucasic skins, but I feel it works for almost all skin tones.
Open your color panel and use RGB sliders.
Grab your eyedropper and take some measures of the skin.
The trick is that the 3 sliders need to form (more or less) a straight line. That is it!
Another thing to consider is that this 3 sliders should stay below this 45° angle with respect to the vertical if it is outside this the skin will look too saturated.
Some Initial tests
Here is a photo that has a very clear yellow-green tone:
Taking some random samples the RGB sliders are out of my (currently arbitrary) graph.
Instead of removing the green, I opted to slide up the blue channel, because I noticed the lack of this.
It is not perfect but a big improvement.
One thing I noticed. This relation is more or less stable. The problem with specific measures, instead of proportions is that it is difficult to know what part to measure, highlights, shadows, etc.
With this method, you can measure with some freedom. It only will move your sliders left and right, will make it more or less vertical (depending on the style of the photo, saturated or not very much saturation)
Also, this gives me the freedom to choose which channel to adjust. The obvious choice on the example was to reduce the green, but It is not that important and moving the blue worked.
You now can reduce the brightness, reduce the saturation or do some other adjustments. The proportion of the 3 channels is what is important.
I think this is a good starting point.
Photoshop is lacking an important tool
In video editing, one important thing is to keep consistent skin tones. Some professional programs have a vectorscope. It is a type of graph to view color values (think of it as another type of histogram). There is a line called the skin tone line. https://www.google.com/search?q=vectorscope+skin+tone
Let me simulate this using an oldie free plugin.
This is a mapping of the values of the pixels on the white square on a 3D cube viewed from above. This is similar to the vectorscope.
On the original image, we see the pixels a little closer to the yellow axis.
Now compare it to the corrected image, where the same zone is mapped around the middle ground between red and yellow: Orange.
It is a little closer to the center too because we added blue, but that is not important.
According to videographers, this line is independent of skin tone (race) it can be closer or further away from the center but stays on the same hue (polar) angle. This means the same of my findings. The proportions of the colors are what matters.
This is a tool that would be handy in Photoshop.
Mvs Color cube plugin. It only works on old versions of photoshop. Probably 32 bits. http://www.vicanek.de/imageprocessing/colorcube.htm
-
\$\begingroup\$ Affinity Photo has Vectorscope, RGB Parade, RGB Waveform and Intensity Waveform if this is something you find useful in still photography. \$\endgroup\$ Sep 20, 2016 at 7:44
My comments are becoming short answers, so I'll start an answer already. Originally I figured without the specific numbers but only a citation it was a comment.
Anyway, it is no accident that the axis of CIE LAB color spaces match the way our eyes perceive (oppenent perceptual space). I foind this treatise to be fascinating and very detailed, concerning how we "see". And "see" as in perceive because our raw sensors are essentially RGB but it gets converted to LAB for later perceiving workmin the brain. IIRC, yellow-blue channel is most primitive and allows us to adjust white balance for sun/shade/midday/morning etc. Skimming through the section, I don't see what I thout I remember about red/green as a useful second axis.
But my idea is: go to LAB and, if it is the red-green axis your eyes lack, select that channel to be visible only, as grey. Or, swap the A and B and match the false color mapping against other photos and learn over time what the right colors are there. (Then swap back) that's assuming that the skin tint to fix is quite close to the red/green axis, and the blue color cast is perceived seperately, especially in how we have come to learn how photos reveal a blue or orange tint of the lighting. A blue error in the face won't be seen as bad skin tone but in cool lighting condition. And the proper color to hit happens to be red, which is the opponent axis.
Adjust the A and B channels separately by using numbers, greyscale view of the channel alone, or false color in the colors you are most descriminating in. For skin tone in particular, if the AB axes are not sufficient, tools can be used to transform the color to put one axis exactly along the direction of needed fiddling.
I hope you get the gist of what I'm thinking.
(P.s. I've wondered from time to time what colors my parrot sees, as primates are color blind from his perspective)
If you want the color correct, I think the only answer is a color checker. You will find a wide variety of these small targets that contain a set of different color squares. You put the target in the scene or have somebody hold it and take one photo. You then take other photo's without the checker.
When processing your images, you correct the image that contains the color checker and apply those corrections to all the others. Several color checker kits contain tools specifically for workflows such as Photoshop or Lightroom.
The issue of correct color or skin tone issue doesn't relate to your color deficiency. I don't think many people (anyone?) have perfect color memory that would allow them to accurately reproduce the color of a scene. (Is the dress blue and black or white and gold?) That is why there is a market for these checkers. In theory you could be completely color blind and reproduce correct color using a checker.
I keep highlighting correct because you mentioned skin tones. Accurate may be true, but not always what people want to see in a picture. You may actually want appealing skin tones. If that is the case, you likely need the help of your subject or an assistant and a set of color chips (for example a Pantone skin tone set). Your subject or assistant would have to tell you "this picture should be number 123-3" They in post processing you could match the image to the color number.