基于多块模式决策的边信息生成方法

Information Generating Based on Multi-Block Mode Decision

摘要:

分布式视频编码中边信息的质量决定了系统的率失真性能,边信息质量越高,则率失真性能越好。针对视频序列中对象运动的不均匀特性,结合MCTI技术,本文提出了一种新的边信息生成算法。其基本思想是在编码端利用多块模式算法对帧中宏块进行划分,将宏块分为运动缓慢块和运动剧烈块；在解码端,对运动缓慢块直接由MCTI算法生成边信息,而运动剧烈块的边信息要经过后处理进行优化得到。仿真实验表明与直接由MCTI生成边信息方法相比,本文算法可以使生成边信息的峰值信噪比(PSNR)比原有的算法提高0.8dB-1.2dB左右,有效提高了边信息的质量。

Abstract:

The quality of the Side Information in distributed video coding determines the RD performance of DVC codec. In this paper, we propose a SI generating algorithm with the combination of MCTI based on the uneven motion of objectives in the video sequences. First, we place a multi-block classification in the encoder which categorizes the block into the high speed motion and the low speed one. The SI of the low speed motion is directly generated by MCTI . But for the high speed motion block , we optimize the SI by the procedure proposed in the paper. Simulation results indicate that, using the presented algorithm instead of the MCTI one to generate the SI, the average PSNR of the DVC codec can be increased by 0.8 ~ 1.2dB.