Font Size: a A A

Design And Implementation On Region-based Nature Image Retrieval System

Posted on:2012-06-06Degree:MasterType:Thesis
Country:ChinaCandidate:Y D ZhangFull Text:PDF
GTID:2178330335450938Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the development of multimedia technology and the improvement of image acquisition devices, image data grows at an unprecedented rate. Now, how to effectively analyze, organize, and manage these vast amounts of image resources makes more and more scientific research agencies focus on content-based image retrieval systems. Image retrieval research involves many areas, how to make computer understand and match the image information according to the visual perspective from the people is one of the difficulties of current research. Region-based image retrieval (RBIR) is a new stage of Content Based Image Retrieval (CBIR).Describing and extracting in the regional characteristics of the image information as well as regional characteristic matching are the two core issues of image retrieval. This work is to focus on these two issues and realize the algorithm.Image segmentation is a reasonable and effective image feature description and extraction region a prerequisite. In this paper, based on the traditional segmentation method, we use JSEG segmentation algorithm for segmentation image. In extraction salient region, we use JSEG algorithm combined with significant attention mechanism map extraction algorithm to extract the salient areas of the image. This system can use RGB, YUV, LAB and HSV color model to describe the regional characteristics of the image information. In this way, we improve the retrieval accuracy and express the intention of the user.In the period of region feature matching, in begin with we analysis comprehensive regional of the original Integrate Region Matching (IRM) algorithm. In the system we realized a mutil-characteristic of the IRM retrieval method and took experiments on image database and analysis the result of IRM algorithm. After this, we improved the original algorithm using salient region. In feedback in image retrieval process, we firstly introduced a Multiple-instance learning method, and realized Diversity-Density algorithm, and then we used this function in our retrieval system.Based on the researching algorithm, We built up a content-based image retrieval experiment platform, and specified the system composition and function. At last, we used statistical method to analysis the experiment results.
Keywords/Search Tags:Image segmentation, Salient region, IRM, Multiple-instance learning
PDF Full Text Request
Related items