The Research On Indexing And Optimization Technology In Content-based Image Retrieval

Content-Based Image Retrieval is a hot subject of information retrieval, belonging to multi-disciplinary subject, involved in computer graphics, image processing, data mining and pattern recognizing. It is mostly used to solve the problem of "semantic gap" in Text- Based Image Retrieval.In the aspect of cluster indexing, we research and improve the K-means clustering algorithm, reduce the effect of the dependence of the cluster centers in traditional K-means clustering algorithm for clustering through improvements in the initial cluster algorithm. On the other side, aiming at EMD image similarity calculation algorithm, a new adaptive clustering algorithm is dug out, so that we will no longer rely on the cluster centers and the number of clusters, reduce the average algorithm matching number from n to k+n/k with high precision and recall rates (k>>n).In the search engine optimization, a user-based search engine optimization technology is brought forward on the base of traditional search engine optimization technology. The user can gain the hot image retrieval result from the server end in the shorter response time with the memory cache; Relying on the user log learning, the user can judge and improve the result.Based on the performance tests, clustering results will no longer be restricted by the number of clusters and initial center. The way, which is to determine the number of clustering by the distance between the image feature vectors, could provide more accurate retrieval result in a shorter time. If cache hit, the response speed will be faster. Meanwhile, the precision can be raised to 75% by user log learning.