CAJ | 학술논문

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。
침대간단선성질대취류( SLIC)방법변계분할불정학적문제，제출일충개진적SLIC방법。위강저문리、조성등신식대도상변계정위적영향，이용비선성확산려파기대도상진행려파처리，재보지변연신식적동시려제조성、평활도상，병재제일차질대후운용Sigma려파기적특성증가취류적한정조건，종이개진SLIC초상소취류효과，제고질대과정중취류중심적준학도，감소오분할적현상，병사용상소적량도거리합병고립점。방진결과표명，해방법재분할도상세절방면표현교호，여전통SLIC산법상비，재미증가시간복잡도적동시，능제고변계명중솔。
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 while 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 while 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 without significantly increasing the processing time compared with the conventional SLIC method.