多视点视频编码中宏块复杂度的研究

Research on Macro Blocks Complexity in Multi-view Video Encoding

  • 摘要: 多视点视频带来了更真实生动的画面感,但同时也需要更大的数据量。本文对多视点视频编码模型JMVC中采用的模式选择技术编码计算量大的问题,通过分析和研究当前宏块的模式与相邻的已编码宏块及参考帧相应位置宏块模式间的关系,利用宏块模式间的相关性,对当前编码宏块的模式选择算法进行了优化,减少部分细小块的分割及搜索。实验结果表明,在其重建视频质量及码率开销基本不变的前提下,改进算法对运动较平缓的立体视频序列效果较好,较JMVC模型中的算法对全部测试序列的平均编码时间减少了约67%,而对运动较平缓序列的编码时间减少达73%。

     

    Abstract: Multi-view video brought lively and vivid picture scope, but also needed huge amount of data. Aimed at the issue of large amount of calculation of the mode selection technology in Joint Multi-view Video Coding model(JMVC), through the analysis and research on the correlation among the macro-block modes, such as the mode of current macro-block , the mode of encoded adjacent macro block and the mode of reference frame in the corresponding position, according to the connection between the macro-block modes, the optimized macro-block mode selection algorithm is proposed, which can effectively reduce segmentation and searching to small pieces. Experimental results show that under the premise of the reconstruction of video quality and bit-rate overhead basically remain unchanged; the improved algorithm had better performance. Compared with the original algorithm in JMVC, the improved algorithm can reduce the average encoding time by 67% to all test sequence, and especially can reduce the encoding time by 73% to gentle motion video sequence.

     

/

返回文章
返回