IMAGE RETRIEVAL BASED ON IMPROVED HIERARCHICAL CLUSTERING ALGORITHM

The traditional CBIR is sequential retrieval. However, for large and high-dimension image databases, it is obvious that this retrieval method has been unable to meet efficiency. It is more important that the image database should be preprocessed and establish indexing to improve retrieval efficiency. Focus on the hierarchical clustering algorithm's high computational complexity, this paper introduces ART2 clustering algorithm for image database preprocessing, which reduces the computational complexity, and makes the Algorithm more efficient. In order to avoid the clustering center offset of ART2, K-means algorithm is used to calculate the pattern center, improving the accuracy of clustering. Compared by retrieval efficiency and retrieval result, it is convincingly proved that hierarchical index structure based on clustering is efficient and applicable in CBIR.