基于熵率超像素和区域合并的岩屑颗粒图像分割

Image segmentation of cutting grains based on entropy rate superpixel and region merging

针对岩屑颗粒密集和颗粒表面纹理复杂的特点,提出一种基于熵率超像素分割和区域合并的分割方法。熵率超像素分割将图像分为一系列紧凑的、具有区域一致性的区域,不仅边缘定位准确且降低图像计算的复杂度;针对存在的过分割情况,提出一种结合颜色直方图和形状信息的合并准则,进行基于RAG结构的快速区域合并,得到最后分割结果。实验结果表明,将该方法用于岩屑颗粒图像分割,能够取得较好的实验效果。

As the distribution of cuttings grains is dense and its surface texture is complex,a segmentation algorithm was proposed based on the entropy rate superpixel segmentation and the region merging.Firstly,the image was divided into a series of compacts with regional consistency using the entropy rate superpixel segmentation algorithm.This algorithm not only made edges accurately,but also greatly reduced the complexity of image calculation,and it is suitable for the segmentation of cutting grains.In view of the unavoidable over-segmentation,a merging rule was proposed which combined the color histogram and the shape information together for fast region merging based on the structure of RAG.The experiment results show that applying this method to cutting image produces a better result.