Colorization based compression using fuzzy c-means and smooth L0 minimization

In this paper, a high rate compression of color images is achieved using colorization technique and smooth L0 (SL0) minimization. At encoder, the color image is converted into Ycbcr space. The luminance (Y) information of image is segmented using fuzzy-c-means to construct colorization matrix. This colorization matrix is used in choosing few important color representative pixels (RP) by applying SL0 algorithm. Since luminance part of image is transmitted with only few RPs so the size of image reduces to greater extent. At decoder, using similar colorization matrix and available RPs with luminance information, color image is reconstructed. The quality of reconstructed color image depends heavily on the method adopted in selection of RPs. The reconstructed image quality measurement parameters used are MSE, PSNR and SSIM. It is found that comparatively higher compression rate is achieved with good image quality.