摘要： 针对胸部CT扫描图像库，提出了一种融合区域和全局特征提取的医学图像检索方法。为了提取局部感兴趣区域，给出了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，再结合迭代阈值算法进行病灶边界的磨合。为了避免身体姿势问题造成的图像角度差异，利用具有旋转不变性的Zernike矩提取图像的全局特征，融合感兴趣区域的形状和分布特性以及整幅图像Zernike矩全局特征作为图像匹配准则的客观依据。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。

关键词: 医学图像检索, 感兴趣区域, 灰度层共现矩阵, 区域增长法, 阈值法

Abstract: This paper presents a method of medical image retrieval based on regions of interest and global features. It presents a new region growing algorithm with gray level co-occurrence matrix to extract the focus region, the threshold segmentation is applied to improve the edge of segmented region. In order to avoid the image’s angle difference for the bodily position, the paper uses Zernike moment to extract the global features of the image. It combines the shape and distribution features of the regions of interest with Zernike feature for image retrieval. Experiment results show that the method is much more effective.

Key words: Medical image retrieval, Regions of interest, Gray level co-occurrence matrix, Region growing algorithm, Threshold segmentation