基于过渡区提取的多阈值图像分割

Multilevel image segmentation based on transition region extraction

对基于过渡区提取的多阈值图像分割算法进行了研究 .首先使用局部熵方法提取过渡区 ,然后构造对称非最大抑制滤波器检测过渡区直方图的多个峰值 ,最后使用考虑差异度与位面数的代价函数决定最优类别个数及多分割阈值 .实验研究表明 ,算法得到的阈值准确、稳定 ,较好地解决了基于过渡区提取的图像分割方法中的多阈值分割问题 ,同时也为多阈值图像分割提供了一种新的思路 .

The transition region was extracted by local entropy and symmetric non maxima suppression filter was adopted to detect multiple peaks of transition region′s histogram. A criterion considering both discrepancy and bites was used for determining optimal classification numbers and multiple thresholds. Experimental results demonstrate that the thresholds obtained by proposed algorithm are accurate and robust. The proposed algorithm has offered effective solutions to multilevel segmentation problems in transition region extraction based on segmentation technique. It has provided a new idea for multilevel image segmentation.