摘要
提出了一种压缩域背景建模与运动对象分割框架:首先提取H.264视频压缩码流中的MB-bits与4×4块残差系数,基于MB-bits场进行压缩域Vibe背景建模分割出宏块级运动对象区域,然后结合最大熵模型提取4×4边缘块进行运动对象轮廓细化最终分割出压缩域运动对象。试验对比分析表明,提出的算法能快速、准确地提取压缩域运动对象,系统具有一定鲁棒性。
A compressed domain based scheme is proposed to model background and segment moving objects in this paper. First of all,the size in bits that a macroblok MB occupies（MB-bits） and 4 ×4 residual coefficients within such a MB are extracted from the compressed bitstream,based on MB-bits field in compressed domain Vibe background modeling is build to segment moving object in MB level. Then combining maximum entropy model to refine object contour by extracting edge of 4 × 4 block, and finally segment moving object in compressed domain. Experimental comparison and analysis show that the proposed algorithm can extract the compressed domain motion object rapidly and accurately with certain robustness.
出处
《电视技术》
北大核心
2014年第15期24-28,共5页
Video Engineering
关键词
压缩域
背景建模
运动对象分割
最大熵
compressed domain
background modeling
moving object segmentation
maximum entropy
