Methods for Restoration of Compressed Depth Maps: a Comparative Study

The 3D-scene representation format consisting of singlechannel video augmented with per-pixel depth has been considered as one of the promising candidates for 3D video delivery. It provides backward compatibility and is compression-effective as the gray-scale depth channel takes a relatively small fraction of the total bit budget. While compressed with standardized compression tools, such as H.264 encoders, the decompressed depth images suffer from heavy blocking artifacts, which impede the correct rendering of virtual views at the receiver side. In this contribution, we compare methods aimed at restoring the decompressed blocky depth maps. We aim to find whether and which kind of structural information from the color channel can help in restoring the depth maps and to specify good measures for evaluating their quality. We present experimental results comparing five methods and employing six different measures. The results show that the blocky depth maps can be effectively restored thus allowing for better use of the available bit budget. 1