小波及分水岭变换在图像分割中的应用

Application of Wavelet and Watershed in Image Segmentation

在进行分水岭图像分割时,由于对噪声及灰度变化的敏感性,将得到过度分割的图像。为了克服这种缺点,利用小波多分辨率分析特性及图像增强对图像进行处理,去除大量的噪声信息,保留图像边缘的细节信息,并使图像边缘像素灰度值得到增强,减少由噪声带来的过度分割;对降噪增强后的图像使用形态学梯度算法及控制标记符的分水岭分割方法能够进一步减少由于灰度变化小带来的过度分割,从而较好地解决了过度分割问题。

Due to the effects of the noise and small retuse regions,the over-segmented image will be produced by the watershed algorithm.To overcome the drawback,the wavelet deniosing is used to reduce the over-segmentation by wiping off the noise,reserving and enhancing the edge information. A number of over-segmentations are reduced by using the marker-controlled watershed segmentation algorithm and morphological gradient.The experiment results show that this method is a solution for the over-segmentation problem.