Improving depth map compression using a 3-phase depth map correction approach

Autosteroscopic multiview displays are the future of 3D technology. At the same time, this technology requires transmission of multiple video streams and depth maps, increasing the necessity for efficient compression. In this work, we address the challenging issue of 3D and multiview compression by introducing a method that corrects depth maps to achieve high quality synthesized views while significantly reducing the required bitrate. Our proposed method uses a three-phase depth map correction, including eliminating anomalies, segmentation, depth value correction, and finally inter-frame and intra-frame filtering. Subjective performance evaluations show that our method yields high quality synthesized views while significantly reduces the required bandwidth for the corrected depth maps (up to 48%).