Study On The Image Retrieval Technology Based On Shape Features And Relevance Feedback Mechanism

Along with the continuous development of all kinds of digital devices, huge amountsof images go into people’s life. How to effectively retrieve and identify the informationneeded become a problem which people more and more focus on. Text retrieval alreadycannot satisfy people’s needs, so more and more scholars focus on the research ofcontent-based image retrieval. Representation of image content features include color,texture, shape, and so on. The users prefer to identify objects according to shape features,therefore the shape feature is got more attention. Representation and description of theshape is the core event of the shape-based image retrieval. The complexity of the shapefeatures and the existence of the semantic gap make the realization of shape-basedretrieval have certain difficulty. Therefore, the research of shape-based image retrievalhas great significance.First of all, the development and the key technology of content-based image retrievalare summarized and analyzed. Especially about shape-based retrieval, more deeptheoretical study is made. Based on the existing research achievement, a kind of imageretrieval method is put forward which combine the region feature with the contour feature.Zernike moment is chosen as the region feature which describes global information of animage. PGH and centroid distance histogram is chosen as contour feature whichdescribes detail information of an image.Then, relevance feedback is introduced into image retrieval. The user evaluates thefirst retrieval result. According to the result of evaluation, the feature that users pay moreattention to is recognized. After the weight of prior feature is improved, a new retrieval iscarried out.Finally, an image retrieval prototype system is implemented based on the abovealgorithm. The performance of the proposed method is evaluated. The experimentalresults show that the proposed method is more outstanding than single feature, andrelevance feedback mechanism is eliminated semantic gap to a certain extent.