基于灰色聚类决策的图像分割性能评价

Performance evaluation based on grey clustering of segmentation

针对图像分割的性能分析问题,运用灰色聚类决策技术建立一种多层次的综合评估方法。该方法通过图像多个分割指标的特点,得到合适的白化函数,再通过白化函数和灰色聚类权值的分析计算,将收集到的分散信息生成灰色聚类矩阵,以此对图像分割性能进行分类;其评估结果的离散性高,便于区分不同分割算法的性能。应用实例表明,该方法很好地克服了以往评估方法的不灵敏性和奇异性,从而使评判更容易进行。

In view of the performance analyses of image segmentation, a hierarchical synthetic evaluation method is proposed based on the grey clustering. The method obtains the appropriate albinism function by target characteristic, then the scattered information produces cluster matrix. The new method can distinguish the performances from dif- ferent segmentation algorithms easily. The experiments show that the method has higher sensitivity and lower singu- larity than other methods, and can be more effectively used to analyze the performance of segmentation algorithms.