Optimizing color transfer using color similarity measurement

Abstract

Color transfer algorithms alter the color appearance in the source image by borrowing colors from the target image. In this paper, we present a simple yet effective generalized color transfer algorithm, taking into consideration the influences contributed from both the source and target images. We introduce the Gaussian membership function as our first color similarity measurement. This function aims at balancing the degree of similarity of color distributions between source and resultant images and that between target and resultant images, respectively. With regard to the second color similarity measurement, we combine color histogram with the statistic concept of the correlation. Referring to different weights, the histogram correlation method derives correlations at each color channel between source and resultant images and those between target and resultant images. Color transfer is optimized by automatically stabilizing correlations before producing the final color transfer resultant image. Experimental results demonstrate that our proposed algorithm automatically takes into consideration the source and target images producing a visually plausible result. The statics comparison shows that our scheme outperforms current state-of-the-art methods.