关键词: 彩色图像, 分水岭, 最小生成树, 自适应局部阈值, 图论

Abstract: The using of minimum spanning tree based on graph theory can obtain global division effect and good operating efficiency. But the number of arcs is too large, and the method through fixed global threshold doesn’t adapt to all color images. This paper aims at these problems, combines the minimum spanning tree method with watershed algorithm, firstly, transform the initial segmentations into nodes and arcs of graph, then merge them through minimum spanning tree. By analyzing the local changes of the regions characteristics, adaptive local thresholds are derived. The experiment indicates that, the minimum spanning tree method based on adaptive local thresholds is better than the minimum spanning tree based on global threshold.

Key words: color images, watershed, minimum spanning tree, adaptive local thresholds, graph theory