Multi-Layered Coding of Depth for Virtual View Synthesis
| Citation: | Yea, S.; Vetro, A., "Multi-layered Coding of Depth for Virtual View Synthesis", Picture Coding Symposium (PCS), DOI 10.1109/PCS.2009.5167368, pp. 1-4, May 2009 (IEEE Xplore) |
| MERL Report: | TR2009-033 |
| MERL Contacts: | Sehoon Yea, Anthony Vetro |
Map of depth coding error (δ >= 4)
It is well-known that large depth-coding errors typically occuring around depth edge areas lead to distorted object boundaries in the synthesized texture images. This paper proposes a multi-layered coding approach for depth images as a complement to the popular edge-aware approaches such as those based on platelets. It is shown that guaranteeing a near-lossless bound on the depth values around the edges by adding extra enhancement layers is an effective way to improved the visual quality of the synthesized images.
 | Read the full technical report (PDF: 203.7 kB) |