Color texture reconstruction using a bidimensional Markov model

A model for the reconstruction of real color textures is presented. It is an extension of a bidimensionnal Markov model previously proposed for the synthesis of gray level textures. The main problem is to limit the number of colors to avoid computationnal complexity. This is solved by operating a suboptimal quantization of the color information included in the real texture: i) firstly by transforming the color components into a perceptive uniform space, based on a vision model; ii) then by optimally clustering this color space.