Font Size: a A A

Design And Implementation Of Image Retrieval System Based On Multi-feature Fusion

Posted on:2013-07-04Degree:MasterType:Thesis
Country:ChinaCandidate:C XingFull Text:PDF
GTID:2248330395986925Subject:Detection Technology and Automation
Abstract/Summary:PDF Full Text Request
Content-based image retrieval is a process of finding a similar image to thetarget images, which mainly reflects the characteristics of image information throughsimilarity criterion. The process of feature extraction and similarity measurementcriterion are the most important for the result of image retrieve. The former reflectedthe representation of image content, the latter has large impact on real time of thealgorithm, these two points are the key research content.This paper narrates status at home and abroad and the new research in the fieldof image retrieval research. And introduce the key technique of the image retrieval.The content of this article is divided into several parts:Firstly, this paper constructs image retrieval system and introduce imagepreprocessing algorithm and the classical algorithm for image feature extraction. Ituses median filtering as preprocessing algorithm and common feature extractionalgorithm mainly contains color feature, texture feature, shape feature.Secondly, it improves the classic algorithm of color feature extraction andtexture feature extraction. For color feature extraction algorithm, this paper integratesedge information into histogram. The experimental results show that the improvedretrieval algorithm has higher accuracy than the original retrieval algorithm. Fortexture feature extraction algorithm, this paper proposed a method based on graylevel co-occurrence matrix as texture feature representation. Firstly, calculate graylevel co-occurrence matrix, then calculate the image of energy, entropy, moment ofinertia, and the local balance, correlation and texture feature respectively accordingto the gray level co-occurrence matrix. Compared with the original algorithms, thisalgorithm significantly improves retrieval effectiveness.Finally, a new image retrieval algorithm based on multiple features fusion isproposed, which integrates two improved algorithms mentioned above. Meanwhile,an image retrieval system based on the algorithm presented in this paper is constructed. The experiment results on the Corel database show that the image recalland precision are87.6%and88.4%respectively, which improves the accuracy ofimage retrieval effectively, and satisfies the actual requires perfectly.
Keywords/Search Tags:image retrieval, fusion of multiple features, histogram, gray levelco-occurrence matrix
PDF Full Text Request
Related items