基于奇异值分解和小波变换的医学图像识别  

Medical Image Retrieval Based on Singular Value Decomposition and Wavelet Transform

作　　者：常兆凯[1] 齐敏[1] 李大健[2] 严欣[1] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072 [2]西北工业大学365研究所,陕西西安710072

出　　处：《计算机仿真》2011年第4期254-256,383,共4页Computer Simulation

基　　金：西北工业大学研究生创新种子基金(Z2010044)

摘　　要：在提高诊断与治疗准确性的研究中,医学影像识别是重要手段。为了克服由于光照条件不同以及成像设备的限制,使得不同条件下获取的图像之间不仅存在较大的灰度差异,而且还存在旋转、偏移等几何形变而导致无法识别匹配的问题,提出了一种利用奇异值分解得到图像的空间信息,然后用小波变换提取纹理信息,再利用奇异值分解得到图像的能量信息来进行识别匹配的方法。仿真结果表明,可有效地解决诊断的准确性,而且对图像的旋转和偏移也具有很好的鲁棒性,能够准确和高效地查找所需的图像,与其它算法相比具有较高的识别率。In order to overcome the homogeneous or a different device to obtain medical images,the different lighting conditions may result in potential retrieve large errors of the issue,a new kind of medical image retrieval algorithm was presented by using singular value decomposition to obtain the space information.Then by using wavelet transform,texture features of image were extracted based on singular value decomposition of image matching methods of energy retrieval.The image retrieval experiments indicate not only that the brightness does not have the sensitivity,but also that the image rotation and offset have a good robustness.It can find out the needed images accurately and efficiently,reduce the requirements of the treatment of retrieving images,and has a higher retrieval-rate than other methods.

关 键 词：医学图像识别 奇异值分解 小波变换 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]