粘连颗粒图像分割方法综述

Summarization of Image Segmentation Methods for Adhesion Particles

摘要: 粘连颗粒图像分割是图像分割的一个重要应用领域，也是一个难题。本文首先总结了基于阈值、边缘、分水岭等大类的分割方法。基于阈值的分割方法算法简单，运算速度快，但分割结果依赖阈值选择。基于边缘的分割方法边缘定位准确，适用于表面光滑、形状规则的图像分割。粘连颗粒分割最广为应用的即为分水岭分割，相对其他分割方法能得到更为准确的结果。但若对粘连颗粒图像直接应用分水岭分割，则容易出现过分割现象。因此与其他方法相结合以实现分割则能够得到更好的分割结果。最后，本文认为由于粘连颗粒图像形状不规则、特征相同、表面粗糙、堆叠严重等因素的影响，未来算法还应在准确性、精确度、客观评价等方面继续改进。

Abstract: Image segmentation of adhesive particles is an important application in the field of image seg-mentation, and it is also a difficult problem for researchers. In this paper, we first summarize the segmentation methods based on threshold, edge, watershed and so on. The threshold-based method is simple and fast, but the result depends on the threshold. The edge-based method can detect the edge accurately, and is suitable for the image with smooth surfaces and regular shapes. Watershed is the most widely used segmentation, which is more accurate than other methods. However, if the watershed segmentation is directly applied to the image of the adherent particles, over-segmentation is likely to occur. Therefore, combining with other methods to achieve seg-mentation can result in better segmentation results. Finally, the future algorithm should be im-proved in accuracy, precision and objective evaluation, due to the irregular shape, the same char-acteristics, the rough surface and the serious stacking.