Font Size: a A A

Image Retrieval Techniques By Combing Multiple Features And Relevance Feedback Based On ROI

Posted on:2008-01-17Degree:MasterType:Thesis
Country:ChinaCandidate:H TianFull Text:PDF
GTID:2178360215970894Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the rapid development of computer, multimedia, and Internettechniques, Content-Based Image Retrieval (CBIR) becomes one of the mostactive research focuses of implementation of multimedia.In this paper, firstly the background, meanings of CBIR, the current statesand active research areas are introduced, and the classical Content-Based ImageRetrieval System (CBIRS) are presented, after which, a brief review of thesystem frame and some important techniques in CBIRS are introduced.By using the technologies of image processing, pattern recognition,computer vision and database, this dissertation studies some key problems in thefield of image retrieval. The major contents include:The low-level feature descriptions including color, texture, shape and thesimilarity matching measures are deeply analyzed, and the feature extractionsand image retrievals based on the single low-level feature are realized byexperimentation.Region Of Interest (ROI) is a hotspot of CBIR at present. An algorithm byusing combined features based on Region Of Interest is proposed in this paper.This algorithm firstly decomposes the sample image into sub-images byquad-tree, and then the feature of the sub-images of interest selected by the useris extracted for similarity measure. Meanwhile, a method of feature extractionby Color Moment, Co-occurrence Matrix, Invariant Moment based on Region Of Interest is proposed.In order to capture the user's intention and achieve the better performance,the techniques of relevance feedback are analyzed in detail. A region weightadjusting algorithm mingling ROI is proposed.An image retrieval system model of combing multiple features andrelevance feedback based on ROI is proposed. And the multilevel CBIR modelbased on ROI and the Region weight model are proposed. Furthermore, theprocedures of the multilevel retrieval and region weight adjusting mingling ROIare described in detail.A image retrieval system prototype of combing multiple features andrelevance feedback based on ROI is designed and realized, it can reduce the gapbetween low-level features and high-level semantic concepts and be closer to thehuman perception by using ROI and relevance feedback. This system providesan experimental frame platform.
Keywords/Search Tags:Content-Based Image Retrieval (CBIR), Region Of Interest (ROI), low-level feature, feature extraction, similarity matching, relevance feedback, multiple features, region weight
PDF Full Text Request
Related items