Image Retrieval Based On Salient Region And SVM Relevance Feedback

With the rapid development of computer and Internet techniques, digital imagelibraries are growing larger and larger, so that traditional text-based image retrievalmethods have been unable to meet the needs of people. Content-based imageretrieval come into people’s attention and quickly become a research hotspot.In the past, people mainly describe images from a global perspective in CBIR. Infact, the images often have a lot of redundant information, but only the local imageobject users care. Based on this consideration, this paper researched image retrievalbased on salient region, the main work is summarized as follows:(1) A novel salient region based method is presented. It searches for the localmaximum density region of interest points which is called salient region after thepoints were extracted. And within the salient region interest points named as salientpoints. Then, it integrates the color and shape features of the salient region forretrieval. Experimental results have shown that our method does better in both colorretrieval and color and shape retrieval.(2) To further explore the user’s search intention, SVM relevance feedback willbe introduced into the image retrieval. The method that base on salient region weproposed is used as the initial result, and then SVM relevance feedback is added for asecond search, experimental results show that the retrieval efficiency has beenimproved after the introduction of SVM.(3) In order to verify the validity of the algorithm, this paper designed an imageretrieval system towards the local database.