Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 100 === Depth distortions caused by depth video coding which is required for storage and transmission considerations may cause geometry changes in view synthesis. However, for the depth value of each pixel, there exists a range within which no geometry changes will be observed. We propose an efficient 3D video coding which, for each pixel in the macroblock, modifies, without inducing any geometry change, the depth value to minimize the residual. Experiment results show that the proposed method substantially reduces the bitrate while preserving the quality of the rendered image.
|