针对胸部CT扫描图像库,提出了一种基于感兴趣区域提取的图像检索方法.首先为了提取感兴趣区域,提出一种基于灰度层共现矩阵的区域增长算法,分割出病灶区域,再通过闽值算法进行边界的磨合,然后针对感兴趣区域提取形状和分布特性作为图像匹配准则的客观依据.最后,将该算法与其他底层特征算法进行比较和分析.实验结果表明,该算法能够比较有效地应用于基于内容的医学图像检索系统中.
This paper presents a method of medical image retrieval technique based on extraction of region of interest. First of all, in order to extract the focus region, we propose a texture-based region growing algorithm and the threshold segmentation is applied to improve the edge of segmented region. Then we extract the shape and distribution features of the region of interest for image retrieval. At last, we compare this approach with the method of retrieval based on other features and make a detailed analysis in which the result indicates that the method discussed in this paper is much more effective.