一种基于帧间差与图像分割的运动目标分割方法  

A method of moving object segmentation based on inter frame difference and Graph Cuts

作　　者：陈娟[1] 应骏[1] 王健[1] 曾爽[1] 

机构地区：[1]上海师范大学信息与机电工程学院数字社区与智能家居研究中心

出　　处：《上海师范大学学报（自然科学版）》2017年第2期242-246,共5页Journal of Shanghai Normal University(Natural Sciences)

摘　　要：视频图像中对运动目标进行分割是十分有意义的.传统的背景减除法和帧间差法在提取运动目标的时候各有自身难以克服的缺点.基于阈值、边缘、直方图的图像分割一般只适合于静态目标的分割.提出一种结合图像分割(Graph Cuts)和帧间差的新方法进行运动目标分割.帧间差法具有良好的稳健性,而Graph Cuts算法则弥补了帧间差无法获得完整运动信息的缺点.利用帧间差法得到不完整的运动信息对Graph Cuts进行分割约束可以准确、稳定地提取出运动目标.Motion segmentation is meaningful video processing. Traditional method like background subtraction and frame differ-ence shortcomings. Image segmentation based on threshold,EDGE histogram only used for static object segmentation. Frame differ-ence method has good robutness and Graph Cuts can get complete motion. Segmentation of Graph Cuts is restricted by incomplete motion information provided by frame moving object can be extract accurately and stably.

关 键 词：帧间差法 图像分割 运动分割 

分 类 号：TN919.8[电子电信—通信与信息系统]