Font Size: a A A

Research On Multiple Feature Based Image Retrieval And Relevance Feedback Methods

Posted on:2007-11-14Degree:MasterType:Thesis
Country:ChinaCandidate:S F ZhangFull Text:PDF
GTID:2178360212957570Subject:Industrial Engineering
Abstract/Summary:PDF Full Text Request
With the development of multimedia and internet technology, the application of the image is more and more extensive. How to retrieve necessary information from large amount image information efficiently and quickly needs to be solved urgently. Traditional image retrieval methods are no more appropriated for large image databases. In order to overcome the limitation of the traditional searching method, the CBIR (Content-Based Image Retrieval) has become one of the hot research areas in image domain.CBIR analyses the image on the basis of the color, texture and shape embedded in the image. Index of the image is built on the feature vector. The system retrieves the above information from the image database to satisfy the users' requirement.On the basis of widely referring to the material about CBIR, this dissertation summarizes CBIR system architecture and some key techniques of the image retrieval. The key research issues of this dissertation are feature extraction and relevance feedback.The key technique of CBIR is extracting feature. Integrated two methods for image retrieval based on single color or shape feature, this dissertation presents a new method for image retrieval based on multiple features of salient points, which utilizes the location information of the salient points, extracts the local color histogram and Hu invariant moment, matches the feature vector to retrieve the image. The results of experiments show that the method utilized multiple features overcomes the limitation of the method utilized single feature and improves the efficiency and universality of the system.It can improve retrieval performance to introduce relevance feedback to CBIR. This dissertation introduces the theory of rough set to relevance feedback and presents a new feedback algorithm based on district feature re-weighting. The decision-making table is constructed according to the feedback images. The weights of the district features are adjusted according to the precision of decision-making rule. The results of the experiments show the effectiveness of the algorithm.A content-based image retrieval system which implements the technique of relevance feedback is constructed. The results of experiments prove the validity of the proposed methods.
Keywords/Search Tags:CBIR, Multiple Feature, Relevance Feedback, Rough Set
PDF Full Text Request
Related items