The Technological Research Of Content Based Trademark Pictures Retrieval

Trademark is an important symbol of goods, and plays an important role in market economy. With the sharply rising of trademark registration number and widely using of image database technology in many fields, it requests higher for the trademark management, and how to retrieval similar trademarks from mass data quickly and accurately is a hot spot in current research. Content Based Image Retrieval method provides an effective way for solving this problem. It means retrieving images based on the image content description features directly. The present research on CBIR focuses on the similarity matching of low-level features. This paper summarized the key techniques of CBIR, combining with the characteristics of trademark image, proposes a new brand retrieval method. The main work is as follows:We propose an improved image retrieval technology based on shape context, which collect a group of limited discrete points to represent the shape of image, and obtain the space information using the vector set getting from one point to the rest of other points. Improving the boundary extraction algorithm and the matching point computational method, after normalized processing, then match two feather points based on shape histograms. Applied to the trademark retrieval, it has good two-dimensional invariance and shape visual consistency.Color information in trademark images is a very important message in the judgment of similarity. We extract color histogram as color features, which is convenient and easy to calculate, and embodies the mass-tone attune of the trademark image.In order to overcome the deficiencies of retrieval on individual features, on the analysis of the shape and color feature extraction algorithm, this paper introduces a method of combining shape and color characteristics to retrieval trademark image. We introduce the normalization method of different characteristics and the relevance feedback technology, and design a retrieval system as a test platform with explanations of the structure, modules function and the users interface. A small trademark image database is established for system performance test, The results of the experiments make senses that composes of multi-features get good result than single one, and with good satisfaction.