1

Here is example of what I need to do: I created image of a pallet with squares of different colors, they have some RGB values, which I do not know, lets say for example (100,20,157), (12,58,236)...etc. I can see image on my monitor, so colors are already a little different than they really should be, then I took my phone and take photo of that monitor, so values changed like (187,31,112), (17,102,214), these colors are really different. I need to find out what were original RGB values of those colors.

Some help is allowed like: on the picture, I can have reference color, whose RGB values I know, so I will know how that changed, etc.

I read about transforming it into another color space, but not what color space and how it can helps me.

So my question is: How to find out what were original RGB values of that image? I don't have to get exact values, just get closer to them. Example

AbdelAziz AbdelLatef
  • 3,650
  • 6
  • 24
  • 52
Dominik21
  • 178
  • 1
  • 8
  • related post. https://stackoverflow.com/questions/18897730/how-i-make-color-calibration-in-opencv-using-a-colorchecker – pangyuteng Oct 13 '19 at 22:59
  • 2
    Closer by which factor? And nope there is no way to get the original values back. Yes the different color profiles between your original image, your monitor, your camera will all add to the entropy (and thus make it harder), but simply the calibration of your monitor (I guess you don't have a pro calibrator at hand), the settings of your camera (just take two pictures with auto on and off), and even before all that, the ambient lightt, switch on and off your room's light) and the angle at which you'll take the photograph. – Kaiido Oct 13 '19 at 23:08
  • 1
    Under controlled capture conditions and with a good colour management chain you could be very close to the original colours but your example is anything but controlled capture conditions. – Kel Solaar Oct 14 '19 at 07:06
  • I would add from @KelSolaar comment: and good monitor and good camera sensor. "Good" in a colour fidelity sense. – Giacomo Catenazzi Oct 14 '19 at 08:04
  • You could check https://www.argyllcms.com/ (open source), you have tools and libraries to do most of work: creating patches, reading patches, creating profiles, doing colour correction (so you see: there are 4 complex steps, which are more complex than one could imagine). – Giacomo Catenazzi Oct 14 '19 at 08:07
  • 1
    This is a very complex subject. Books have been written about this. That means a good answer won't wit in the space available on Stack Overflow. – MSalters Oct 14 '19 at 09:28
  • Looking at the enormous variation across the photo of the top-left reddish block in your photo when the original was a single shade of red tells me you are going to have difficulties. You could try adding black, white and grey patches and using the black, white and grey colour droppers in the Curves dialog of Photoshop. – Mark Setchell Oct 14 '19 at 09:32
  • The point is, the photo will be given to me, so i can not choose any device, or calibration,i do not need to get exactly same values, just to get closer to them , maybe the image will be printed, then photographed... Thank you for answers, gonna check it out – Dominik21 Oct 14 '19 at 10:28
  • The problem is that there is not a single fixed variable in your system: your colour management chain is unknown, the capture parameters are unknown, the captured image might have been processed with unknown filters. At this point, there is an infinite amount of colours that could have generated the ones you are looking at and given you don't know anything, their distribution could be very large. – Kel Solaar Oct 14 '19 at 18:06

0 Answers0