Study On Content-Based Image Retrieval

We are living in a digitalized world. More and more audio and video information are digitalized and stored. This is especially so after the emergence of World Wide Web. The amount of information grows in an exponential way which makes information retrieval a crucial problem. Since most of the information one receives is from vision, image retrieval has become an important part of information retrieval technology. Therefore, Content-based Image Retrieval (CBIR) system has gained more and more attention in the scientific research world.This paper studies several kernel techniques of Content-based Image Retrieval and proposed a relevance feedback method based the RBF neural network. It is shown in the experiment that this method can enhance the retrieval accuracy in an efficient way. And finally a content-based retrieval system from huge trademark database is designed.