摘要
提出了一种基于分水岭变换和统计区域合并的图像分割方法.该方法综合利用高斯低通滤波、分水岭变换和统计区域合并,先对原始图像提取分割标记,然后利用Meyer分水岭变换对标记分水岭进行分割,最后利用概率统计的方法对过分割区域进行合并.该算法通过调节尺度参数可以实现由粗到细(coarse-to-fine)的分割.实验结果表明,这种简单可行的算法在分割噪声图像时依然有良好的效果,具有较强的鲁棒性.
A hybrid image segmentation using watershed transform and statistical region merging(ISRM) is proposed.This method takes a comprehensive utilization of Gaussian lowpass filters(GLPFs),watershed transform and statistical region merging(SRM).The segmentation markers will be extracted from the original image.Then Meyer watershed transform is applied on the original image using those labels.Finally,statistical region merging is used to merge the over-segmentation images.One input parameter is needed in this method to build a hierarchy of coarse-to-fine(multi-scale) segmentation of an image.The results show that the simplicity and robustness of the approach make it possible to cope with noise corruption.
出处
《中国计量学院学报》
2012年第4期373-378,共6页
Journal of China Jiliang University
