Abstract: In this paper, a gradient-preserving color transfer algorithm based on multi-channel adaptive matching is proposed to improve the classical Welsh algorithm and Reinhard algorithm. L, a and b channels are introduced as matching criteria, the processing of a single channel will not affect the other two channels because of the sequence irrelevance between channels, eliminating the drawbacks of single target region in RGB color mode that is vulnerable to multi-channel influence. For images with similar gray values, the channel with the largest difference in the mean chromaticity between the processing area and the background area is adaptively matched with the corresponding channel of the reference image, which solves the problem that Welsh algorithm can easily misjudge the grading of color source images with small gray differences or overlapping gray ranges. Then, gradient factor is introduced and weighted with the standard deviation ratio of reference image to source image. The result is taken as the scaling ratio coefficient of the improved transfer algorithm in this paper. Gradient-preserving color transfer algorithm is used to color each group of matched pixels to avoid the problems of loss of color detail information and unnatural transition between tones in Reinhard algorithm. Experiments show that the color effect of the improved algorithm is better than that of the classical Welsh algorithm and Reinhard algorithm, which greatly improves the flexibility and applicability of the color transfer algorithm.

Key words: image processing, sequence irrelevance, multi-channel adaptive matching algorithm, gradient-preserving color transfer algorithm, improved structural similarity index algorithm