摘要
在分布式视频编码系统中,针对非线性运动区域边信息生成误差大的问题,提出了基于视频序列运动特性的边信息改进方法。解码端利用双向运动补偿时域内插技术生成初始边信息,再根据前后运动矢量的和值将边信息划分为线性与非线性区域,并对非线性区域用矢量滤波器进行去伪非线性区域处理,对真正做非线性运动的块进行分层运动估计以得到更精确的运动信息。将得到的边信息与初始边信息相比,改进后的边信息平均PSNR值提高了1dB。
To solve the problem that the generated error of areas doing nonlinear movement is greater than other regions in side information （SI）, a method based on motion features to refine the SI is proposed. First in the decoding side, initial SI is produced by motion compensated temporal interpolation （MCTI）, then linear and non-linear motion areas in SI are distinguished by the sum of the forand back-ward motion vectors （MVs）. After correcting the false MVs by motion vector filter （MVF）, hierarchi cal motion estimation （HME） is utilized for the further processing of blocks belonging to non-linear area to obtain more accurate motion information. The quality comparison is made between new SI frames and the initial ones, which finally indicate that the average PSNR of new ones improve ldB.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第11期4250-4254,共5页
Computer Engineering and Design
