摘要：
在Diversity & Importance架构的基础上提出一种改进的空间降尺度视频转码算法，利用方差作为转码目标宏块的运动矢量测度因子，并通过自适应阈值的分类方法对不同视频序列选择合适的重编码模式。该算法比传统的运动矢量合成方法提高了约1dB的PSNR，在保持较高转码图像质量的同时有效地控制了编码长度，其转码后码流长度只相当于AMVR-DIM方法产生码流的40%~95%。
关键词:
运动矢量合成,
运动矢量精细化,
多样性,
重要性,
帧内刷新
Abstract: An improved algorithm for down-scaling video transcoding is proposed based on diversity & importance transcoding architecture. By using the variance measure and an adaptive classification method, it can choose the appropriate re-encoding mode of each block in various video sequences. The algorithm produces about 1dB PSNR higher than the traditional mv-composition algorithm. While preserving high transcoding quality, it produces only 40%~95% bits than AMVR-DIM method.
Key words:
MV composition,
MV refinement,
diversity,
importance,
intra-refresh
中图分类号:
张宏波;彭玉华;刘兆广;刘志胜. 一种改进的降尺寸视频转码算法[J]. 计算机工程, 2008, 34(16): 241-243.
ZHANG Hong-bo; PENG Yu-hua; LIU Zhao-guang; LIU Zhi-sheng. Improved Algorithm for Down-sizing Video Transcoding[J]. Computer Engineering, 2008, 34(16): 241-243.