Feature selection and re-weighting in content-based SAR image retrieval

With the development of synthetic aperture radar (SAR) in recent years, the explosion of SAR images has urged people to find efficient means for searching and organizing mass amounts of images. In this paper, we propose an approach to content-based retrieval of SAR images, which contains feature selection and relevance feedback. In the process of retrieval, a low-dimensional feature subset is selected from original feature set by feature selection technique based on linear support vector machines (SVM). And the relevance feedback technique employs feature re-weighting method to set appropriate weights for each component of the selected feature subset. Lastly, this feature subset with different weights is used to retrieve relevant images in database which are similar to sample images submitted by users. The experiment results prove that the proposed method is efficient for querying pure terrain in SAR image.