Research On Content Based Trademark Image Retrieval

Trademark Office retrievals the trademark to be registered by the traditional text-based and category-based trademark image retrieval method.This manual classificational and keyword-based method is of subjectivity, and time consuming.For the staffing constraints of the Trademark Office, using the traditional method to retrieval the trademark has caused a large backlog of applications for trademark registration.The trademark registration period is extended,which is not conducive to rapid development of market economy.In recent years, the development of content-based image retrieval (CBIR) method breaks through the drawbacks of this traditional method, which uses the characteristics of image itself (color, texture, shape, spatial structure information, semantic etc.). Many researchers at home and abroad have researched trademark images retrieval using CBIR in order to achieve efficient trademark images retrieval. Therefore, this paper design and implement a trademark image retrieval prototype system.First,this paper reviews the content-based trademark image retrieval and analyzed the related technologies.The paper focuses on the commonly used shape feature extraction methods, such as shape invariant moment, edge direction histogram, eccentricity and so on.Second,this paper presents an trademarkimage retrieval method based on integrated multi-features.First eccentricity,7 HU invariant moments,and edge direction histogram are extracted, and use the weighted combination of these three features as the initial search. Second use the blocked features as the second search method.Finally,this paper design a trademark image retrieval prototype system. The system consists of human-computer interaction, image preprocessing, feature extraction, image retrieval module.