색상공간 lαβ를 이용한 학습기반 흑백영상 채색화 기법

초록

This paper proposes an algorithm to transform a gray-scale image to a color image based on a reference color image. Firstly, RGB color space of the reference color image is converted to lαβ because of its decorrelation property. Secondly, the luminance distribution of the reference image is remapped to that of an input gray image. Finally, the chrominance information for the gray image is synthesized using the learned relationship between luminance and chrominance channels in the reference color image. Experimental result shows that the proposed algorithm can improve the colorization effect of grayscale images.