基于改进熵率超像素和区域合并的岩屑图像分割精选.pdf

Journal of Image and Signal Processing 图像与信号处理, 2016, 5(1), 16-24 Published Online January 2016 in Hans. /journal/jisp /10.12677/jisp.2016.51003 Rock Image Segmentation Based on Improved Entropyrate Superpixel and Region Merging * Jie Li, Xiaoqiang Wu , Shuhua Xiong, Yajing Wang, Xiaohai He School of Electronics and Information, Sichuan University, Chengdu Sichuan th th th Received: Dec. 26 , 2015; accepted: Jan. 10 , 2016; published: Jan. 13 , 2016 Copyright © 2016 by authors and Hans Publishers Inc. This work is licensed under the Creative Commons Attribution International License (CC BY). /licenses/by/4.0/ Abstract Detection and segmentation of rocks plays an important role in many applications such as geolog- ical analysis and mining processes, and it is the foundation of the analysis and recognition. Rocks are usually segmented using a variety of features such as texture, shading, shape and edges. In this paper, an improved algorithm based on entropy rate superpixel and region merging is provided. The entropy rate superpixel segmentation algorithm proposes an objective function, which favors formation of compact and homogeneous clusters. Considering the shape of Rocks, the mirror sym- metry coefficient is integrated into the objective function. If rock images were then segmented by superpixel algorithms with this objective function, the results would adhere more accurately to rock boundaries. Furthermore, we propose an improved maximal similar region merging algo- rithm based on the sum and difference histograms for function reducing the over-segmentation. The experiment results show that applying this algorithm to rock images performs very w