Trademark Image Retrieval System using Neural Networks

The Image retrieval plays an important role in several applications such as trademark registration, fingerprint classification and face recognition. Trademarks are considered as valuable intellectual properties. They play very important roles for successful business or companies. A huge amount of trademarks have been registered, and they are protected from imitation through legal proceedings. Therefore, registering a newly designed trademark without conflicting with the previously registered trademarks has become more and more difficult. To avoid inadvertent infringement of copyright, one must have an automatic trademark retrieval system for examining the content of the newly designed trademark with the ones that have been registered. In Chamber of Commerce in Saudi Arabia, the trademarks are classified into groups and the closest trademarks are identified manually to the one we want to register it as a new trademark. This process takes very long time. In this research project, a new approach for retrieving Trademark images is proposed based on integrating multiple classifiers. This approach is suggested to speed up the retrieving process and to improve retrieving accuracy. The proposed system is divided into feature extraction and classification or retrieving. Feature extraction is the most important step in this system for obtaining good selected features to utilize them in the classification stage. The classification process is based on multiple classifiers which get different features from the previous stage. The main goal of this research is to develop a complete automatic trademark retrieval system with intelligent issues surpass the systems introduced in literature and to reduce many of the restrictions in the working environment.