摘要
针对分水岭算法对噪声敏感和易于产生过分割的问题,提出运用顶帽变换对图像进行Ostu局部阈值处理,改善光照不均和噪声对图像分割的影响;采用多尺度形态学梯度,解决结构元素的形状和尺寸对梯度图像产生的影响。实验结果表明,该算法既能有效地分割粘连颗粒,又能有效抑制过分割现象。
I-top-hat transtSrm is introduced to local threshold processing, so as to improve uneven light effect and noise on the image seginentation for solving the problems of noise-sensitive and over-segmentation watershed transformation. Multi-scale morphological gradient is used to solve the impact of the structural elements of the shape and size on gradient image. By compared with other watershed algorithms, the experimets show that this algorithm can segment the touched kernels more effectively and inhibit the over-segmentation phenomenon.
出处
《微型机与应用》
2012年第9期44-46,49,共4页
Microcomputer & Its Applications
关键词
图像分割
数学形态学
分水岭算法
顶帽变换
多尺度形态学梯度
image segmentation
mathematical morphology
watershed algorithm
hop-hat transform
multi-scale morphological gradient
