Image Retrieval Based On Regional Semantics

At present, the image retrieval based on image semantic hierarchy has become to be an very important research field.In this paper,the main work is the image retrieval based on region semanticannotation. The main contribution of this paper are the following.Firstly, we have achieved the JSEG algorithm for image segmentation.Using this algorithm for image segmentation,the segmentation region will represent the true object region well.Secondly, fusion the color histogram feature and the texture feature with a weight to represent the image feature.Thirdly,we propose a method which fusion the K-means clustering algorithm and K-nearest neighbor algorithm in this paper to achieve the automatic semantic annotation.Experimental results demonstrate that proposed method is efficient.