基于关联度分析的图像分割性能评估  

Synthetic Performance Evaluation of Image Segmentation Algorithms

作　　者：李保洲[1] 王建琦[1] 邓雁萍[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200030

出　　处：《上海交通大学学报》2000年第6期824-826,共3页Journal of Shanghai Jiaotong University

摘　　要：针对图像分割的性能分析问题 ,运用关联度分析技术建立了一种多层次的综合评估方法 ,运用此方法评估既可以在整体性能上进行 ,又可以从某个侧面进行 .该方法所采用的评估指标具有多层次的特点 ,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 ,该方法较好地克服了以往评估中存在的不灵敏性和奇异性 。To the performance analyses of image segmentation, a hierarchical synthetic evaluation method was proposed based on the relational grade. Using the method, the performance evaluation can be performed from both the global or local aspects. The metrics for evaluation is hierarchically determined. With the properties of high discreteness, the new method can distinguish the performances from different segmentation algorithms easily. The experiments show that the method has higher sensitivity and lower singularity than other methods, and it can be more effectively used to analyse the performance of segmentation algorithms.

关 键 词：图像分割 性能评估 关联度 综合评价 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]