基于改进分水岭的图像分割算法研究

针对传统的分水岭算法由于受到图像细节和噪声的干扰而存在着过分割现象，提出了一种改进的分水岭算法，并应用于图像分割。该算法首先采用双边滤波进行预处理，然后建立形态学梯度并对梯度图像进行开闭重建，在保留了轮廓极值信息的同时去除了噪声，最后对重建的梯度图像进行前景和背景的标记，依据标记对梯度图像进行修正并用分水岭算法对其进行分割。结果表明，该方法能有效抑制传统分水岭的过分割现象。

Image Segmentation Based on Improved Watershed Algorithm

Because of image details and noise interference, the traditional watershed transformation exist over-segmenta!ion. Therefore, an improved watershed image segmentation algorithm was put forward based on the traditional watershed algorithm. Firstly, it ＇ s preprocessed with bilateral filtering. Secondly, morphological double opening and closing reconstruction is performed after the morphological gradient＇ s calcula- tion, which keeps the outline of extreme value information and eliminates the noise. Then the reconstruction gradient image is marked with the prospects and backgrounds. According to the marks, the gradient image is amended. Finally, the watershed transformation of the modified gra- dient image is performed to achieve the segmentatlon. The results showed that the improved watershed algorithm can effectively restrain the o- ver-segmentation phenomenon.