 |
Technical Program
Paper Detail
| Paper: | MA-P8.11 |
| Session: | Stereoscopic and 3-D Coding |
| Time: | Monday, October 9, 09:40 - 12:20 |
| Presentation: |
Poster
|
| Title: |
LOWERING THE COMPLEXITY OF MULTI-VIEW ENCODING THROUGH DYNAMIC SEGMENTATION AND REGISTRATION OF VIDEO OBJECT |
| Authors: |
Xiaohui Wei; University of Texas at Arlington | | | | | Mei Yi Chu; University of Texas at Arlington | | | | | Ishfaq Ahmad; University of Texas at Arlington | | | | | Yongfang Liang; University of Texas at Arlington | | |
| Abstract: |
Exploitation of redundancies for superior coding efficiency is one of the major research issues in multi-view video coding (MVC). However, this incurs extra computation, counteracting the benefit gained from coding efficiency. This paper exploits inter-view correlations among video objects and background to lower the prediction complexity while achieving high coding efficiency in MVC. We propose an effective object-based scheme that utilizes video object information obtained from the coded base view. This information assists to predict disparity vectors and motion vectors in enhancement views (EVs) by employing object registration and warping. This leads to high compression and low-complexity coding scheme for EVs. The proposed scheme provides higher coding efficiency when compared to the segmentation-free scheme, while achieving lower computational complexity. |
|
