基于低秩优化的CT图像目标检测方法  

Low-rank optimization method for object recognition in CT images

作　　者：张敏辉 杨剑[2] ZHANG Min-hui 1,YANG Jian 2(1.School of Computer Science,Chengdu Normal University,Chengdu 611130,China;2.School of Computer,Chengdu College of University of Electronic Science and Technology of China,Chengdu 611130,Chin)

机构地区：[1]成都师范学院计算机科学学院,四川成都611130 [2]电子科技大学成都学院计算机学院,四川成都611130

出　　处：《计算机工程与设计》2018年第7期2080-2083,共4页Computer Engineering and Design

基　　金：四川省教育厅自然科学重点基金项目(17ZA0053)

摘　　要：针对传统的模式识别方法难以识别CT图像中不规则目标的问题,提出一种基于低秩优化的目标识别方法。利用病灶部位在影像中的稀疏性与多样性,将众多CT图像配准到标准图像中并连接为一个矩阵;利用矩阵中正常组织部分的低秩性和病灶组织部分的稀疏性,将矩阵分解为低秩成分和稀疏成分;通过低秩优化寻找矩阵中的低秩成分和稀疏成分,直接分离出病灶组织。实验结果表明,该方法相对于传统的分类和聚类算法可以极大减少误诊率,具有更快的运行效率,可更有效地运用于辅助诊断。Automatic diagnosis with CT images has been widely investigated.However,traditional pattern recognition based approaches view objects as individual regions and it is difficult to detect irregular objects in CT images.An object recognition algorithm based on low rank optimization was proposed.CT images were mapped to normal images by image diffeomorphism.Images were transformed into a low-rank and sparse matrix.By optimizing the low-rank and sparse components respectively,the objects were decomposed from CT images.Experimental results show that the proposed approach achieves higher precision and speed than the state-of-the-art methods,indicating great prospects on CT image based diagnosis.

关 键 词：低秩优化 CT图像诊断 目标检测 图像处理 模式识别 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]