一种改进的Split-Merge图像分割算法

Modified Split-Merge image segmentation algorithm

图像分割的经典算法Split-Merge算法思路简洁、运算效率高,但存在边界定位不准确和过分割等问题。针对过分割问题提出了一种改进的Split-Merge算法,在合并过程中引入全局目标函数,并在目标函数中定义了长度项,使用贪心法完成合并过程。长度项以及贪心法的使用促使相邻区域充分合并,有效地克服了过分割等问题,取得了较好的分割效果。

As a classical algorithm in image segmentation, Split-Merge algorithm is simple and effective. However, two problems arc often encountered, i.e. inaccuracy of edges and over-segmentation of the image. To eliminate oversegmentation, this paper proposed a modified Split-Merge algorithm. In the modified algorithm, a global objective function including a length term was introduced, and the greedy algorithm was used to merge the regions. The use of length term and greedy algorithm made it possible to merge the adjacent blocks effectively and eliminate over-segmentation. The experiments show that the new algorithm can yield good results.