Research Of Relevance Feedback Of Image Retrieval Based On SVM And Rough Set

With the application of Internet and the continually development of multimedia and internet technique, digital image should be regarded with its abroad, The traditional text-based image retrieval is no more appropriated for large image databases. When people trying to find one of the most effective search methods, the Content-Based Image Retrieval emerged and became and become a more active research areas.In this dissertation, lots of exploratory research work for relevance feedback has been done around some key techniques of CBIR, The main contributions of this dissertation are summarized as follows:1. Has studied relevance feedback algorithm based on the SVM. For the selected issues of SVM kernel function and parameters, proposed a feedback method combining with rough set, This method carries on the feedback using SVM to the primary search image first, then uses the up and down approximation of rough set to classify and sort once more, then return results.2. This paper uses the histogram-based color features and co-occurrence matrix-based texture features to describe the whole feature.3. A content-based image retrieval system is designed. This system supports image retrieval based on color and texture features. The system also supports relevance feedback. Two relevance feedback methods are: SVM-based and SVM combined with rough set-based.