图像分割中阈值选取方法比较研究

COMPARISON OF THRESHOLD SELECTION FOR IMAGE SEGMENTATION

以错分概率、相对均匀测度及相对形状测度为评价准则,利用合成图像测试分割算法的性能,客观、定量地分析比较了目前常用的几种整体阈值选取方法,实验结果表明:分割方法的通用性与目标大小、对比度、噪声等图像因素有关.最后,本文给出了各种方法的实际应用范围.

Based on synthetic images, a comparative performance study of several global thresholding algorithms for image segmentation is quantitatively investigated using the criterioa functions such as the probability of error, absolute uniformity measurer and absolute shape measure. Experiments show that the performance of global thresholding techniques is image-dependent. In other'words, it is limited by the object size, the contrast, noise, and so on. Finally, the scope of application of each method is proposed.