Design And Implementation Of Registered Trademarks Retrieval System Based On Image Retrieval

A trademark is a notable indicator, which commodity producers or operators use on their products for consumers to identify the source of goods. A registered trademark is private, and protected by the law. In the process of trademark classification and registration, each trademark should be not only unique, but also easily distinguished. Therefore, before registering, it is necessary to query and compare trademarks in order to be sure that the trademark to be registered has not been registered. Along with the surge in the number of trademarks, it is more difficult to query the trademarks. Traditional Text-Based Trademark Retrieval can not meet the requirement of the application in this area, so people begin to apply the content-based image retrieval technology to retrieval trademark.In the content-based image retrieval, it is particularly critical for feature selection and extraction. Taking into account the characteristics of trademark images: monotonous color, simple shape and regular size, in the paper, we utilize comprehensively the most notable features of trademark images: color and shape, and the texture analysis for image segmentation. Besides, we add the spatial information into texture feature. On the basis of improving the retrieval accuracy and efficiency, we make full use of the integrated features of trademark images.This paper presents a new method of trademark image retrieval based on the shape feature. Firstly, texture image will be divided into the same size and non-overlapping blocks. Secondly, we use such a gray co-occurrence matrix that has been proven to be more effective texture features, and also add the spatial information into each texture block. In the end, we use Fuzzy Clustering Method (FCM) for texture segmentation, by the number of classification as a parameter, and then, we extract the features of segmented regions which including center, quality, dispersion, mean and variance, and produce a 30-dimensional feature vector used in the image retrieval.In order to achieve complementary results of different features, we have adopted the concept of color coherence vector as the features of trademark image. The image content is divided into two types of aggregation and non-aggregation, which considered the spatial information of image.Based on the image retrieval methods discussed and combining the existing research results, we have designed a experimental content-based trademark image retrieval system. Experiments indicate the effectiveness and the preferable performance of our trademark image retrieval system.