基于图割和Sobel算子的视频分割算法  

Video segmentation algorithm based on Graph-Cut and sobel operator

作　　者：薛宾田 冯子亮[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《四川大学学报（自然科学版）》2015年第2期281-286,共6页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(61179072)

摘　　要：针对前景和背景交界处对比度较低的图像分割问题,提出了基于图割和Sobel算子的视频分割算法.首先建立包括颜色分量和对比度分量的能量函数;然后对当前帧和背景求梯度,使用背景的像素梯度来减弱当前帧中背景部分强边缘的梯度值,同时对由颜色模型建立的图像进行Sobel边缘检测,将得到的边缘检测的结果应用到颜色分量和对比度分量中来增加前景和背景对比度;最后,使用图割算法,对能量函数进行最小化求解最终得到分割结果.实验结果表明,当前景和背景交界处颜色相近时,本文分割算法能有效降低分割错误率.A video segmentation algorithm for lower contrast between foreground and background based on Graph-Cut and Sobel operator is proposed.Firstly the energy function containing color term and contrast term is constructed.Then the gradient of current frame and background image to adjust the gradient of the boundaries of current frame should be calculated,the color term and the contrast term of which can be imporoved by the result of sobel edge detection.Finally the energy function could be minimized by Graph-Cut method and the segmentation result of the video could be got.Experiments shows the result in weak contrast situation is better than others.

关 键 词：图割 梯度 SOBEL算子 能量最小化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]