【摘要】 为了克服传统分水岭算法固有的过度分割问题 ,采用了标记 分水岭变换 ,提出了一种新的标记提取方法 .通过计算图像中每个小区域 (5× 5 )的均匀性而得到一幅与原图同样大小的表征原图区域均匀性的均匀性图 ,取均匀性图中的低谷作为标记对均匀性图进行分水岭变换就得到最终的分割结果更多还原

【Abstract】 The marked watershed transform was adopted to overcome the inherent drawback of watershed algorithm over segmentation. By computing homogeneity of every 5×5 block in original image, a "homogeneity image" of same size with the original image was got. The low valleys taken as markers, the segmented result was obtained after watershed processing. The method is simple, effective and adaptable. Additionally, Various accuracy level results can be obtained by setting different thresholds.更多还原