Reduction in "image quality" while performing colour correction.

[MadhuSN]

I have been working on a project involving colour correction using colour checker card (Spyder checkr 24).
When I use the colour science library for the colour correction part, there is a significant reduction of quality in the image.
I have noticed that this happens at the input itself.
Could anyone help me out here and let me know on how do I retain the quality of image upon colour correction ?

[tjdcs]

Can you post the before / after examples. "Image quality reduction" is very generic. But if you are thinking about image noise, color correction is known to increase image noise typically.

[KelSolaar]

There is no scaling occurring during any of those two steps!

[KelSolaar]

Hello,

I looked at your values, and they are seemingly flipped for the first 3 rows:

import colour
import numpy as np
name, data, illuminant, row, columns = colour.CCS_COLOURCHECKERS["ColorChecker24 - After November 2014"]
RGB_r = colour.XYZ_to_RGB(colour.xyY_to_XYZ(data.values()), colour.RGB_COLOURSPACES["sRGB"], illuminant)
RGB_t = np.array(([ 0.15769309, 0.39624593, 0.22701213], [ 0.22282112, 0.16478506, 0.29807267], [ 0.05401623, 0.0879925 , 0.00954172], [ 0.07344554, 0.12108786, 0.21063492], [ 0.4383969 , 0.22097264, 0.09905463], [ 0.09021939, 0.02745266, 0.01039085], [ 0.52391714, 0.14460321, 0.00567051], [ 0.04223775, 0.05309333, 0.2878889 ], [ 0.5799691 , 0.08710817, 0.04322479], [ 0.10819899, 0.02226758, 0.08288278], [ 0.36828655, 0.4591894 , 0.0040894 ], [ 0.5491122, 0.2307647, 0.0061786], [ 0.02189209, 0.15521337, 0.26983383], [ 0.53995943, 0.07772446, 0.21374913], [ 0.7366904 , 0.4929144 , 0.00546476], [ 0.52082956, 0.0307001 , 0.0116053 ], [ 0.05664061, 0.27818367, 0.02142191], [ 0.008285 , 0.00946014, 0.16152452], [ 0.80920285, 0.7283255 , 0.59500945], [ 0.54571736, 0.49706468, 0.39122885], [ 0.37800524, 0.34004018, 0.24603282], [ 0.186911 , 0.16779776, 0.10665054], [ 0.056522 , 0.0516582 , 0.03113838], [ 0.01381 , 0.01308794, 0.00981142]))
colour.plotting.plot_multi_colour_swatches(colour.cctf_encoding(np.stack((RGB_r, RGB_t), axis=1).reshape(-1, 3)), columns=6, direction="-y", compare_swatches="Stacked")

[KelSolaar]

Are you using a SpyderCHECKR24 by any chance (seems like so)?

[MadhuSN]

Are you using a SpyderCHECKR24 by any chance (seems like so)?

Hi, Yes I am using Spyder Checker 24

[MadhuSN]

I added a separate configuration for SpyderCHECKR24 at colour.CCS_COLOURCHECKERS.

[MadhuSN]

Hi, Could someone help me understand the mechanism behind the colour correction module?
I am having some difficulty in understanding it through the code.

[MadhuSN]

Thank you very much for this. This helps a lot

Also I have one other quick doubt. Would I be able to process the same colour correction for raw image files ?

[KelSolaar]

I would not only be possible but you would yield better results because the images should be in a scene-referred state and thus linear with respect to captured scene irradiance.

[MadhuSN]

Hi ! I have been getting this error off late
"AttributeError: module 'colour' has no attribute 'utilities'"

I can see a utilities folder present in the package though.

[MadhuSN]

and also I have been trying to read raw image files using the colour library but I am unable to do so. Is there way you could help me on reading the RAW image files

Apologies I am very new to this whole image processing space. Any help would be much appreciated

[KelSolaar]

could you do a import colour;print(colour.__path__) and verify that it is really colour at that path? There is a package on Pypi that has the same namespace, i.e. colour. Our Pypi package is colour-science and the namespace is colour:

• https://pypi.org/project/colour/ --> colour
• https://pypi.org/project/colour-science/ --> colour

As for the RAW files, with the OpenImageIO backend this might work, I actually never tried as I process them using Rawpy.