Research Of Necktie Pattern Retrieval Based On Image Retrieval

With the rapid advancement of the multimedia and Internet technology, more and more digital images are used in many fields, so the amount of various digital images increases at a surprising speed. People are confronted with the problems about how to analyze, extract, store and retrieve the useful and relevant information from the large amount of digital images efficiently and effectively while they benefit from the adoption of digital images. Over the past few years, many researches are performed on the analysis and retrieval of digital images and image retrieval system, and it's is progressing rapidly. In this paper, we focus on key techniques of image retrieval and image retrieval system and employ these techniques on the retrieval of necktie pattern.Firstly, this article reviews the techniques on image retrieval. This article discusses the image data model, image features and the extraction and indexing of image features. This article also refers to the relevance feedback and similarity measure. Then several key issues on the image retrieval system are introduced, such as system architecture, image andfeature database, query interface and appraisal of system performance etc.According to analysis of necktie features, this article selects the color histogram and texture as the main features for the retrieval of necktie pattern. This article computes the color histogram in HSV color space and extracts texture feature by Gabor wavelet algorithm. In this article, the similar feature matching algorithm for image retrieval is established by reducing the scope of images and different weights for features in feature matching computing and multiple retrievals. To improve the system performance, the user relevance feedback information from user is adopted. At last a necktie pattern retrieval system was designed according to above research and software engineering techniques.