分水岭图像分割算法的研究

Research on watershed algorithm based image segmentation

针对分水岭算法在对复杂粘连物体图片的分割过程中产生局部极值过多和分割后区域边缘不能准确定位的弊端,提出对要分割的图像做基于形态学的开闭重建运算;然后对图像的形态学梯度图像进行滤波重建,通过对梯度幅值图像进行修改,标记出要分割区域的最小值,过滤掉无关的较小值;最后运用分水岭算法对标记的图像进行分割,取得了较好的分割效果。

In view of the disadvantages of watershed algorithm in the process of segmentation of complex conglutinated object image,such as too many local extremum and the edge of the segmented area cannot be located accurately,it is proposed that to do morphological open close reconstruction for the image to be segmented;then the morphological gradient image of the original image is filtered and reconstructed;by modifying the gradient amplitude image,the minimum value of the region to be segmented is marked and the irrelevant smaller value is filtered out;finally,watershed segmentation algorithm is used to segment the marked images,thus to obtain a good segmentation result.