基于形态学梯度的分水岭彩色图像分割

Color image segmentation based on morphology gradients and watershed algorithm

图像分割是从图像处理到图像分析的关键步骤之一。在改进了基于地形学距离的分水岭算法的基础上,提出了一种结合图像信息熵、形态学梯度与区域合并的图像分割方法。该算法首先利用信息熵在RGB颜色空间中对彩色图像求其形态学梯度,然后对彩色梯度图进行分水岭分割,最后对分水岭产生的过分割现象进行区域合并。通过Matlab对图像进行实验,结果证明该算法不仅能够减少分水岭算法的过分割现象,而且还提高了图像分割的精确性,同时在图像分割时具有很好的鲁棒性和适应性。

Image segmentation is one of the key steps from image processing to analyzing. Based on an improved algorithm of watershed by topographical distance, an image segmentation algorithm which combined with image information entropy, morphology gradients and region merging is proposed in this paper. The algorithm utilizes information entropy to calculate the morphology gradients in RGB color space firstly. Then, the color gradient image is done watershed segmentation. Finally, it merges the over-segmentation region which is generated by the watershed. Given by the images experiments which conducted on Matlab, the results show that the algorithm not only reduces the phenomenon of over-segmentation but also improves the accuracy of image segmentation, meanwhile has good robustness and adaptability in the image segmentation.