ColorCorrectionML

ColorCorrectionML is a Python package for color correction of sRGB images using machine learning. It uses ML regression methods (linear, least sqaure, and partial least squares regression) to learn the color correction function from a training image with a color checker. The learned function is then applied to correct the color of a test image.

Installation

pip install colorcorrectionML

Usage

from ColorCorrectionML import ColorCorrectionML
import cv2

img = cv2.imread('Images/img2.png')

cc = ColorCorrectionML(img, chart='Classic', illuminant='D50')

method = 'pls' # 'linear', 'lstsq', 'pls' 
# for linear regression, least square regression, and partial least square regression respectively
show = True

kwargs = {
    'method': method,
    'degree': 3, # degree of polynomial
    'interactions_only': False, # only interactions terms,
    'ncomp': 10, # number of components for PLS only
    'max_iter': 5000, # max iterations for PLS only
    'white_balance_mtd': 0 # 0: no white balance, 1: learningBasedWB, 2: simpleWB, 3: grayWorldWB,
    }

M, patch_size = cc.compute_correction(
    show=show,
    **kwargs
)
    

# resize img by 2
# img = cv2.resize(img, (0,0), fx=0.3, fy=0.3, interpolation=cv2.INTER_AREA)

img_corr = cc.correct_img(img, show=True)
# img_corr = cc.Parallel_correct_img(img, chunks_=50000, show=True)

Example results

Scatter plot of the original image and the corrected image color values

Color correction results

TODO

1. Add other reference color values (D55, D65, D70, D75)
2. Add other color charts (ColorChecker24, ColorCheckerSG, ColorCheckerDC)
3. Add other color spaces (CIELab, XYZ, etc.)
4. Add other regression methods (Ridge, Lasso, ElasticNet, etc.)
5. Refine the white balance methods