基于GCBAC的羽毛分割改进算法

Improved algorithm of feather image segmentation based on GCBAC

针对基于活动轮廓模型的图割(Graph cuts based active contours,GCBAC)算法在进行羽毛分割时所出现的过度分割和邻域重叠问题,利用羽毛杆左右端顶点作为约束种子使活动轮廓线必须经过该点以改善过度分割;将轮廓线两侧膨胀改为单向外侧膨胀以改善相邻轮廓线邻域重叠问题.结果表明,该算法交互简单,鲁棒性强,能有效完成羽毛杆分割提取.

In order to solve the overlap problem between adjacent contour neighborhoods and over segmentation with graph cuts based active contours（GCBAC） algorithm in feather quill segmentation,a method of image segmentation was presented,which utilized left and right vertex of feather quill as seeds to pass by active contour to avoid the over segmentation,then bidirection dilation were changed to oneside direction dilation to decrease the overlap of adjacent contour neighborhoods.Experiment results showed that the proposed method was easy interactive,robustic,and could segmented the feather quill successfully.