Abstract: A method of image retrieval based on the feature fusion of region of interest was proposed to realize the semantic correlation of images content. First, the regions of interest were divided and the integrated underlying characteristics of image were extracted. Second, the characteristics were used as training data to classify the images by semisupervised learning, then the mapping between images and categories of semantic was established. Finally, the quadratic distance and the improved Canberra distance were respectively used for measuring lowlevel features, and the cluster centers of images in the feature space were updated iteratively through positive feedback. The experiments compared with other algorithms showed that the proposed image retrieval algorithm had higher accuracy and performed more effectively than traditional algorithms.

Key words: image retrieval, feature fusion, semantic correlation, positive feedback,  region of interest, semi-supervised learning