Compression Artifacts Reduction for Depth Map by Deep Intensity Guidance

摘要

In this paper, we propose a intensity guided CNN (IG-Net) model, which learns an end-to-end mapping between the intensity image and distorted depth map to the uncompressed depth map. To eliminate the undesired blocking artifacts such as discontinuities around object boundary, two branches are designed to extract the high-frequency information from intensity image and depth map, respectively. Multi-scale feature fusion and enhancement layers are introduced in the main branch to strength the edge information of the restored depth map. Performance evaluation on JPEG compression artifacts shows the effectiveness and superiority of our proposed model compared with state-of-the-art methods.