Colorization algorithm using probabilistic relaxation

摘要

This paper presents a method of colorizing a black and white imagery based on the probabilistic relaxation algorithm. Since the colorization is an ill-posed problem, a user specifies a suitable color on each isolated pixel of an image as prior information in this paper. Then other pixels in the image are colorized automatically. The colorizing process is done by assuming local Markov property on the images. By minimizing a total of RGB pixel-wise differences, the problem can be considered as a combinatorial optimization problem and it is solved by using the probabilistic relaxation. The proposed algorithm works very well when a few percent color pixels are known with confidence.