基于SLIC和动态区域合并的岩屑图像分割

岩屑图像分割要求精度高、速度快和鲁棒性强。针对这些要求,提出了基于SLIC(simple linear iterative clustering)和动态区域合并的分割算法。SLIC算法能产生形状规则、大小均匀、排列紧凑的超像素区域。但是SLIC分割后的图像过分割问题严重,为了降低过分割率,提出了基于NNR的动态区域合并算法,将超像素区域进行相似性合并。实验结果表明,将该算法用于岩屑颗粒图像分割,能够取得较好的效果。

Rock Image Segmentation Based on SLIC and Dynamic Region Merging

Rock image segmentation algorithms should meet the requirements of high precision, high efficiency and high robustness. Aiming at these requirements, a segmentation algorithm based on SLIC (simple linear iterative clustering) and dynamic region merging is proposed. Using SLIC algorithm, the image can be divided into different regions, and each super-pixel region is a regular shape, uniform size, compact arrangement. However, SLIC algorithm exists a serious over-segmentation problem. In order to reduce the rate of over-segmentation, we propose a merging algorithm based on dynamic region of NNR, which employs similarity merging algorithm to merge these super-pixel area. The experiment results show that this algorithm performs very well for rock images.