Image Segmentation Scheme Based on Graph-Cut for the Paint Bubbles
Shidu Dong, Jiang Qun, Cui Guan-xun · 2009
The interactive image segmentation scheme based on graph-cut, which is very popular, requires certain pixels as seeds for object and background respectively. It may be terrible for one to mark seeds of all paint bubbles when the amount is large. To overcome this deficit, this paper presents a two-phase scheme. First, with seeds of object and background marked by user, a bubble is segmented by the improved segmentation scheme based on graph-cut. Second, the color and texture feature extracted from the segmented bubble are utilized to detect other bubbles and determine their seeds, and then the bubbles are segmented. Repeating the two-phase, finally, all bubbles are segmented. Theoretical analysis and experimental results show that with the proposed scheme, the accuracy of segmentation is improved and workload of marking object seed is reduced.