一种GrabCut分割边缘的优化方法

针对GrabCut分割精度不够以及边缘不平滑的问题,提出一种目标边缘优化方法.首先使用GrabCut算法获取目标的α掩像,然后对α掩像进行形态学处理而获取目标边缘区域,再将目标边缘区域内的像素与已知的绝对背景和绝对前景的部分像素进行颜色信息比较,根据比较结果重新分割边缘区域像素,最后对重新分割的目标边缘进行平滑处理.实验结果表明,本文的方法对边缘分割的精准性和平滑度都优于GrabCut算法.

An Optimization Method for Edge of GrabCut Segmentation

To overcome the disadvantages of the segmentation accuracy and edge smoothness that those were not good enough,an optimization method of target edge has been presented in this paper.Firstly,the GrabCut algorithm was used to obtain the target''s alpha matte,then the morphological processing of the corrosion and expansion on the alpha matte that it''s used to obtain the edge region of the target.Secondly,the pixels of the target edge region compared with the average color information of the pixels in the local area of the absolute background and foreground,and according to the comparison results re segmented edge region pixels.And,finally,the edge of the target was smoothed.The experimental results show that the presented method was superior to the GrabCut algorithm for the segmentation precision and edge smoothness.