A color transfer method for colorization of grayscale image based on region histogram statistics

A color transfer method for grayscale image colorization based on region histogram texture description is proposed in this article. First a suitable reference image is selected for the grayscale image to be colorized, and linear transformation was applied to the luminance channels of the two images in a de-correlated opponent color space. Then the pixels were matched between two images using pixel neighborhood texture features described by histogram statistics, and the color values of best matching reference image pixel were transferred to the corresponding chromatic channels of grayscale image pixel. Finally the color transferred image is transformed back to the RGB space to be displayed. The result shows that the proposed method has the advantages of high pixel matching accuracy and low computational complexity. The obtained colorized image has better natural colorful feeling than that from image using conventional color transfer methods.