基于深度信息的立体视频错误隐藏方法error-电子与信息学报.pdfVIP

第 34 卷第 7 期 电 子 与 信 息 学 报 Vol.34No.7 2012 年 7 月 Journal of Electronics Information Technology Jul. 2012 基于深度信息的立体视频错误隐藏方法 时 琳* 刘荣科 李君烨 (北京航空航天大学电子信息工程学院 北京 100191) 摘 要：针对双视加深度结构的立体视频的网络传输，该文提出了解码端的错误检测和错误隐藏技术。结合深度信 息对发生错误宏块的编码模式进行估计，在错误宏块周围相邻宏块的编码模式不一致时，选择邻近宏块中深度与待 重建的错误宏块深度最为接近的宏块的编码模式作为其编码模式。然后，根据编码模式选择用视点内或视点间的相 关性信息重建错误宏块。实验结果表明采用该方法重建的图像客观质量峰值信噪比(PSNR)较传统的边界匹配错误 隐藏方法提高了 0.8-6.0 dB，并且主观质量也有提升。 关键词：立体视频编码；H.264；错误隐藏；深度图 中图分类号：TN919.8 文献标识码： A 文章编号：1009-5896(2012)07-1678-07 DOI: 10.3724/SP.J.1146.2011.01107 Error Concealment Based on Depth Information for Stereoscopic Video Coding Shi Lin Liu Rong-ke Li Jun-ye (School of Electronics and Information Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100191, China) Abstract: An improved error detection method and a novel Error Concealment (EC) algorithm at the decoder for ‘color plus depth’ formatted stereoscopic video transmission are presented. The encoding mode of the corrupted macroblock is predicted by the adjacent macroblocks and depth information. In the case that the encoding modes of macroblocks around the corrupted one are inconsistent, the encoding mode of the erroneous macroblock is predicted by the adjacent macroblock whose depth value is closest to that of the corrupted one. According to the encoding mode, the corrupted macroblock is reconstructed by intra-view or inter-view correlation. The experimental results show that the proposed EC algorithm provides 0.8-6.0 dB quality improvement of PSNR against the conventional Boundary Matching Algorithm (BMA) and achiev