Trademark Retrieval Based On V-system

Trademark must be officially registered in national trademark administrations in order to get the protection of the law. The new trademark must have enough unique and cannot be confused with the registered trademark before it is registered. The core issue of trademark retrieval is to solve how to detect a similar trademark images. With the rapid development of the economy, the number of trademarks is growing surge. The traditional handicraft trademark retrieval methods have been incapable of a huge amount of trademark search. Along with the advancement of technology, content-based image retrieval applied to trademark search field, reflecting the huge advantage and overcoming the drawbacks of the manual retrieval methods.Trademark image retrieval method based on the content of the image extract the feature of the image with characteristic attributes such as color, shape, texture, and spatial relationship, and then use these feature to retrieval. Without complex texture and background, the shape characteristics of trademark retrieval are more significant than other characteristics, people tend to use more shape characteristics to retrieval trademark.In this paper, starting from a class of complete orthogonal functions-V system, we mainly research the application of V-system in trademark retrieval. First, a new feature expression based on the characteristics of the boundary, named as V descriptor, is introduced. The advantage of V descriptor in the boundary characteristic expression is depicted,and then,we compare V descriptor with Fourier descriptor in image retrieval and the result prove the advantages of the V descriptor in the group expression.Then, we introduce a new expression based on the characteristics of the region-V moment, which is also based on V-system theory. In this article, we describe the basic forms of expression of the V moments and perform a series of experiments. The experimental results show that the advantages of the V moments at the regional expression. Finally, because features cannot be completely characterized only by shape characteristics, so we explore a new algorithm based on V descriptors, named as "B&R", which is the fusion of boundary characteristics and regional characteristics. The "B&R" algorithm obtain features of trademark through the integration of boundary characteristics based on V descriptor and regional characteristics based on blocked image according to the weight. We designed three types of experiments, which are visual consistency experimental, distortion and deformation experiments as well as the classification experiments. We compared our algorithm with Fourier descriptors, Zernike moment, invariant moments and "F&Z" method respectively, which combine Fourier descriptors and Zernike moments, and the experimental results show that our algorithm has obvious advantages.