Content-aware Deep Perceptual Image Compression

Currently, multimedia data grow in a explosive manner, whereas the effective bandwidth is quite limited in the wireless communications, such as high definition image/video communication, satellite communication and so on. Generative adversarial network (GAN) based image compression goes beyond the traditional image encoding standards under a obviously low bitrate. However, the most existing methods do not consider the importance of different image contents, where the regions with complex and saliency structures generally are more essential to users' quality of experience (QoE). In this paper, we propose a content-aware deep perceptual image compression method (CDPIC). The proposed approach involves a saliency subnet to produce content saliency map for adaptive bit allocation. Moreover, using QoE as the ultimate goal, our approach designs different compression objectives for different image contents, where it tends to preserve more details for the regions with complex and saliency structures, and improves the perceptual quality of the background region. Our experiments show that the proposed approach outperforms the state-of-the-art methods, and produces advanced performance in preserving details and semantics.