一种改进的基于分水岭的图像分割算法

An Image Segmentation Based on Improved Watershed

摘要:

针对分水岭算法过分割现象,提出一种综合分水岭算法、中值过滤算法和归一化割算法的改进算法；该算法首先应用改进型的中值过滤算法对图像进行适当的除噪；然后通过分水岭变换对图像进行了初步分割,最后使用归一化割算法进行图像精度分割；算法集合了分水岭算法、中值过滤算法及归一化割算法的优点,既较好地解决了分水岭算法中过度分割的问题,又降低了归一化割算法的时间复杂度；实验结果表明该算法是一种切实可行的图像分割方法。

Abstract:

In order to overcome the shortcoming in over-segmentation of traditional watershed algorithm, this paper presented an improved image segmentation method, which is having performance advantage of watershed, median filter and normalized cuts. Firstly, an improved median filter is applied to reduce the image noise. Secondly, watershed transform is used to pre-segmentation. Finally, normalized cuts is employed to deal with detail segmentation. Segmentation experiments show that this method which contains good performance of the three steps, can not only effectively avoid the over-segmentation of watershed, but also reduce the time complexity of normalized cuts.