Colorization based on piecewise autoregressive model

Colorization is a technique to produce color components for grayscale images with color assigned information provided by the user. The conventional method colorizes via propagating the assignations with weights of neighborhood pixels. In this paper, we propose the colorization method based on the piecewise auto-regressive (AR) model. The proposed method calculates weights considering taking account of an image structure, and improves objective and perceptual qualities of colorized images compared with the conventional method.