Improving Depth Estimation of 3D Video from Object Edges
| Autor: | Chun-Chi Shao, 邵俊棋 |
|---|---|
| Rok vydání: | 2011 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 99 Recently, 3D video has become an emerging medium that provides more realistic and natural experiences to users. A popular 3D format is the video-plus-depth approach, which may synthesize the required views without huge data. However, there remain big challenges for this approach in occluded regions, where inaccurate depth values may occur around the borders of foreground objects. In this paper, we propose a novel edge-based algorithm to improve depth estimation. Edges of foreground objects are first identified from both the texture images and estimated edge maps. The left edges and right edges of foreground objects are distinguished, and then synthesized from the left view or right view, respectively. The width of edge compensation is determined from the difference in depth values of the foreground object and the background. Experimental results show that the proposed method not only improves the accuracy of the depth map, but also increases the quality of synthesized images both objectively and subjectively. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|