Image Sparse Representation Method And Its Application To Image Retrieval Research

With the rapid development of computer technology, multimedia technology and network technology, image retrieval is becoming one of the most active focus in the field of multimedia. Form the image retrieval based on text, to image retrieval based on the image-content(CBIR),to image retrieval based on sparse representation, every development of image retrieval techniques have attracted many scholars’attention.sparse decomposition based on over-complete dictionary was first proposed in1993, by S. Mallat and Z. Zhang, providing a new direction for the study of signal and image processing. In2004, Candes, Donoho and Tao et al proposed a compressed sensing theory, laid a solid theoretical foundation for its Widely use in the field of image processing. In this paper, after introducing the development of image retrieval and analyzing the principle of image sparse decomposition, we mainly research the image retrieval algorithms based on Sparse Representation. The main content of this paper are summarized as follows:1. we comprehensively analysis the principle and method of signal and image sparse decomposition based over-complete dictionary. To matching pursuit as an example, We analyze the complexity of the algorithm, then propose a image sparse decomposition method based on intelligent algorithm and low-dimensional space.2. proposing a image classification based sparse representation(SRC). We Analysis two important aspects of the retrieval system:feature extraction and robustness, then propose the improvement algorithm3. proposing a image retrieval algorithm based sparse representation. As sparse representation-based classification for core, we propose a new method of image retrieval. combined with the traditional image pretreatment, the performance of the system will be improving4. Using matlab to design and build a image retrieval system based on sparse decomposition.