Adaptive image compression using saliency and KAZE features

In this paper we present a novel framework for compressing images using saliency maps and KAZE features. The method involves adapting the quality factor in JPEG compression scheme for each block instead of using the same quality factor for the entire image. This is achieved by adapting JPEG quality parameter based on visual saliency and KAZE keypoints. Subsequently, a piecewise function is used to compress the least important image blocks with higher compression ratio while maintaining the overall perceptual quality and avoiding blocking artifacts. This work introduces use of KAZE keypoints for image compression for the first time in literature. We show that the proposed method outperforms the JPEG compression using PSNR and FSIMc evaluation measures especially at high compression ratios.