The Research And Application Of Content-based Image Retrieval Technology

The basic thoughts of Content based image retrieval is according to a certain algorithm to extract has given the image feature, then relies on similarity measure function, the image database image features and examples of image feature matching, then according to the similarity level sort, sorting according to display the search results. In the retrieval process, feature extraction and similarity measurement is the key link, the former will affect the image content description, and the latter to calculation of real-time put forward higher requirements.This article in view of the above questions, key research based on color and shape features of image retrieval.Main contents:1. Gives a brief overview of content based image retrieval technology development and the research status at home and abroad, analyses the current image retrieval research focus and difficulty.2. Introduces the content based image retrieval, commonly used image feature description, image similarity measure, the retrieval feedback mechanism, retrieval performance evaluation index and so on.3. Research on the image retrieval based on color features, according to the color feature extraction, using HS V color space and color quantization of images, according to people of different degree of concern, the use of global and local union of the main color of image block, and to all the pieces given to different weight, thus overcoming the histogram method to ignore spatial information deficiencies, so as to improve the retrieval efficiency. The second part is the research on the shape feature based image retrieval technology, on the image shape feature extraction algorithm, this paper puts forward a comprehensive outline of the image features and Hu moment invariants method, can be both a performance of the image edge features and does not lose the entire area of an image retrieval algorithm based on characteristics of. The algorithm first through the improved Canny operator to get the ideal image edge; by eight neighborhood tracking criterion of image outline, the final statistics seven moment invariants to describe the image shape feature. The data show that the algorithm can effectively extract the target shape feature, obtain ideal retrieval results.4. In the study of the classic CBIR. system basis, put forward an image retrieval framework, including feature extraction, feature preservation, image search, the results showed that, the frame has module replacement, flexible interface, easy maintenance and expansion characteristics, and to other multimedia mining algorithm provides a platform.