基于稀疏矩阵的谱聚类图像分割算法

发布时间：2018-11-11 07:37

【摘要】：在图像分割中谱聚类算法需要计算像素之间的相似度矩阵,构造数据量大,并且要对拉普拉斯矩阵进行特征分解,计算比较耗时。针对这一问题,提出了一种基于稀疏矩阵的谱聚类图像分割算法。算法结合图像特征信息在不同尺度上对谱聚类进行误差分析,设计了一种新的样本信息选取方案,并利用选取的图像信息直接创建稀疏相似度矩阵。理论分析以及图像分割实验结果表明,该算法能够有效降低谱聚类的计算复杂度,同时,提高了分割的准确性和鲁棒性。
[Abstract]:In image segmentation, spectral clustering algorithm needs to calculate the similarity matrix between pixels, construct a large amount of data, and decompose the Laplace matrix, which is time-consuming. To solve this problem, a spectral clustering image segmentation algorithm based on sparse matrix is proposed. Based on the error analysis of spectral clustering based on image feature information, a new sample information selection scheme is designed, and the sparse similarity matrix is created directly by using the selected image information. Theoretical analysis and image segmentation experiments show that the proposed algorithm can effectively reduce the computational complexity of spectral clustering and improve the accuracy and robustness of the segmentation.
【作者单位】：兰州理工大学电气工程与信息工程学院;西北民族大学数学与计算机学院;
【基金】：国家自然科学基金项目(61561042) 西北民族大学引进人才基金项目(1310RJYA013)
【分类号】：TP391.41

【相似文献】