基于内容的医学图像检索中对Tamura纹理特征的算法改进  

Algorithm Improvement of Tamura Texture Features in Content-based Medical Image Retrieval

作　　者：辛民宣[1] 史正星[2] 崔光彬[3] 卢虹冰[2] 

机构地区：[1]东南大学网络中心,南京210018 [2]第四军医大学生物医学工程系 [3]第四军医大学唐都医院放射科

出　　处：《医疗卫生装备》2010年第2期32-35,共4页Chinese Medical Equipment Journal

摘　　要：目的:对Tamura纹理特征的算法进行改进,使之在医学图像检索中适用于不规则图形的计算。方法:对乳腺图像进行分割,提取出感兴趣区域(ROI)。在样本(包含正常乳腺图片和乳腺肿瘤图片共281幅)中使用传统算法Tamura纹理特征和改进算法的Tamura纹理特征进行图像内容检索,比较2种算法的检索结果。结果:初次检索时,传统Tamura算法在样本中检索出93幅相关图像,175幅不相关图像;改进后的算法检索出91幅相关图像,131幅不相关图像。结论:使用改进算法的Tamura纹理特征进行查询,在保证查全率的基础上查准率有较大提高。提示改进算法后的Tamura纹理特征更适用于医学图像中不规则区域的特征计算。Objective To improve the algorithm of Tamura texture features, which makes it more suitable for the content- based medical image retrieval of Irregular graph . Mathods First, the breast tissue area（ROI region of interest）was segmented from the breast images. Second, such items were calculated as the original Tamura texture features（OTF）and improved Tamura texture features（ITF）of breast tissue area of breast images from the set which contains 281 normal or breast cancer images. Third, the OTF values were compared with the ITF values, Results When the primary retrieval was performed with the OTF, 93 related images and 175 unrelated images were found. In addition, when ITF was used, the number was 91 and 131. Conclusion The fact that when OTF is compared with ITF, the latter shows a higher precision than the former and they have similar recall, indicates that ITF is more suitable for the retrieval of Irregular graph. [Chinese Medical Equipment Journal,2010,31 （2）：32-35] Key words content-based image retrieval; Tamura texture feature; algorithm improvement

关 键 词：基于内容的图像检索(CBIR) Tamura纹理特征 算法改进 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]