Content based Image Retrieval for Satellite Images

Abstract

The amount of visual information available in digital format has grown exponen- tially in recent years. Content based Image Retrieval systems (CBIR) have become a reliable tool for many image database applications. Today, the need for reliable, automated satellite image classi cation and browsing systems is more than ever be- fore. Everyday there is a massive amount of remotely sensed data being collected and sent by terrestrial satellites for analysis. The use of automated tools for this analysis has become imperative due to the large volumes required to be processed on a daily basis. Our Goal is to develop a prototype of Content based Image Retrieval System for Satellite Images, wherein the user can enter the query in form of text or as a sample image and the relevant similar images based on image content are displayed as an output. The Work presents the image retrieval based on colour and texture. Texture in- formation is obtained using co-occurrence matrix of the gray scaled images. The properties like correlation and homogeneity are good parameters to measure the tex- tural property; these are evaluated at proper o set and angle. Colour is inevitable feature that dominates the human perception most. To a human observer, images with similar colour scheme and reasonably similar in other features is voted higher in similarity list, than images measuring better in other feature descriptors. Making this as the foundation, the work presented here uses the two features independently and also groups colour with texture. Results have shown that the retrieval based on combination of texture and colour based features match better with the human perception. The work further gives an idea about the consequences of Clustering and search post cluster formation on the e ective Content based Image Retrieval. It facilitates reduction in candidate search space and time required to retrieve similar images based on their image content and in turn improves the e ciency of the Re- trieval procedure. The Work nally illustrates few remote sensing applications which can be integrated to enhance the search procedure. MATLAB 7 is the tool used to implement image processing algorithms and Postgresql 9.0.2-1 for windows is the open source database management system used for managing the database of image features.