Font Size: a A A

Research And Implementation Of Image Retrieval System Based On Harris-Affine

Posted on:2009-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:J YangFull Text:PDF
GTID:2178360272470383Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
With the popularization of digital equipment and the development of Internet, the amount of image data grows dramatically fast. Therefore, the content-based image retrieval (CBIR) has emerged to be one of the hot research areas in image domain. At first, most people make use of global information to match and retrieval, such as gray, color, texture and so on. In practice, there are many kinds of noises between the image pairs that need to be matched, such as viewpoint changing, scaling, shading etc. As a result, the retrieval results are far from satisfactory.This paper deeply researches the Harris-Affine algorithm and makes detailed description of its principle and characteristics. It says that the Harris-Affine local characteristic regions are very invariant in visual. The regions are not only robust enough to deal with blur, brightness, noise, but also stable enough with the presence of perspective and scale changing. Compared to the traditional methods, its precision and recall rate have been improved to a higher level.Because of the enormous amount of data and the complex matching process, the conventional content-based image retrieval system are very time consuming. In order to resolve that problem, this paper makes a novel local region based image retrieval framework, combining with the idea in text retrieval. The framework puts forward concepts such as the visual word and the word frequency vector, combining with knowledge in pattern recognition and text retrieval field. The main purpose of the framework is accelerating match and retrieval through data compression.In order to verify the validity of the above-mentioned methods and framework, this paper carries out an experiment in an image database which contains 1000 images. The image retrieval result shows that the method and the framework in this paper are both utility. Compared to the traditional methods, its precision and recall rate have been improved to a higher level. In addition, the retrieval speed can satisfy the needs of the users. The experiment illustrates that the Harris-Affine algorithm and the image retrieval framework in this paper are very useful especially in many kinds of noise including strong perspective transform.
Keywords/Search Tags:Harris-Affine, Visual Invariant, Visual Keyword, Content-Based Image Retrieval
PDF Full Text Request
Related items