Research on content-based remote sensing image retrieval: the strategy for visual feature selection, extraction, description and similarity measurement

There is rich information contained in image data, while rarely text information can be used to retrieve images in a database, therefore, content-based image retrieval (CBIR) is widely studied in the image database (IDB) area. However, research about content-based remote sensing image retrieval is seldom reported in the literature. This paper puts emphasis on this subject, and the strategies for visual feature selection, detection, similarity model, implementation procedure of CBIR on a remote sensing imagery database are thoroughly investigated. Based on principal component analysis and a clustering technique, the cluster of color (spectrum) and texture feature is proposed to index the RS image and a similarity model for similarity measurement between images with varying dimension is proposed. Finally, the validity and usefulness of the proposed theory and technique are proved by the experiment.