基于多通道Gabor纹理特征的遥感图像检索

Remote sensing image retrieval based on multi-channel Gabor texture features.

遥感图像数据量的急剧增加,给研究者们提出了有效检索和管理图像的难题.为了快速准确地检索海量遥感图像,提出一种基于五叉树分解的多通道Gabor纹理特征图像检索新方法.该方法用五叉树分解法层层分解图像,在利用多通道Gabor滤波器与图像做卷积得到滤波能量值基础上,提取各子图像24个滤波能量均值和均方差作为纹理特征,并计算查询图像和与其大小相当的数据库子图像的纹理特征相似性距离,从而获得与查询图像相似的数据库子图像.将该方法用于高分辨率卫星和航空图像数据库的检索,实验结果证明了方法的有效性.

The rapid increase in remote sensing image data has posed a tough task of retrieving and managing image efficiently for researchers. A novel remote sensing image retrieval approach based on multi-channel Gabor texture features and quin-tree decomposition is proposed. A given image is decomposed by using quin-tree, and texture features of each subimage are extracted via multi-channel Gabor filter. The similarity between a given query image and the database subimages, whose sizes are approximately equal to the former, is measured by using Euclid texture distance. Then the most similar subimages are returned as query results. This approach is applied to retrieve high resolution remote sensing images, and its efficiency is confirmed by the experiments.