基于改进SLIC方法的彩色图像分割  

Color Image Segmentation Based on Improved SLIC Method

作　　者：张亚亚[1] 刘小伟[1] 刘福太[1] 张建廷[1] 

机构地区：[1]海军航空工程学院电子信息工程系,山东烟台264001

出　　处：《计算机工程》2015年第4期205-209,共5页Computer Engineering

基　　金：国家自然科学基金资助项目(61102167)

摘　　要：针对简单线性迭代聚类(SLIC)方法边界分割不精确的问题,提出一种改进的SLIC方法。为降低纹理、噪声等信息对图像边界定位的影响,利用非线性扩散滤波器对图像进行滤波处理,在保持边缘信息的同时滤除噪声、平滑图像,并在第一次迭代后运用Sigma滤波器的特性增加聚类的限定条件,从而改进SLIC超像素聚类效果,提高迭代过程中聚类中心的准确度,减少误分割的现象,并使用像素的亮度距离合并孤立点。仿真结果表明,该方法在分割图像细节方面表现较好,与传统SLIC算法相比,在未增加时间复杂度的同时,能提高边界命中率。In view of Simple Linear Iterative Clustering（ SLIC） method sometimes does not well adhere to the edge,this paper proposes an improved SLIC method to solve this problem. In order to reduce the influence of texture and noise information to images boundary location,the nonlinear diffusion filtering algorithm is introduced to smooth the images w hile preserving the boundaries and decreasing the noises. After the first iteration by using the principles of the Sigma filter,it increases clustering condition,improves the result of SLIC superpixels clustering,and enhances the accuracy of clustering center w hile reducing the inaccurate segmentation in the process of iteration. Simulation results show that the superpixels segmentation results of the proposed method have better performance in image details,and this method also increases the boundary hit rate w ithout significantly increasing the processing time compared w ith the conventional SLIC method.

关 键 词：图像分割 超像素 非线性扩散滤波 Sigma滤波器 聚类 边界命中率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]