Adaptive colorization-based compression for stereoscopic images

Considerable research efforts have been devoted to stereo image compression. However, most of them focused on the luminance and ignored the chromatic information. Consequently, in this paper we propose a stereo color image coding method based on colorization. The main idea is to compress one view of the stereo pair using a standard coding method, while for the other view, only the luminance component is considered for compression. The chromatic information of this latter view are transmitted to the decoder for a few representative pixels (RPs) only. These RPs are defined using a novel proposed RP extraction method based on skeletonization. At decoder side, color values of all the remaining pixels are reconstructed by colorization methods. Experimental results showed that our coding method can achieve considerable bit-rate saving compared to conventional coding method.