Pulse Coupled Neural Network Application In The Trademark Image Retrieval Research

As we all know, the cerebral cortex is organisms of the highest level of feeling central, is the key parts which the visual system processes the images information. Simulation and imitation visual system of human being or mammalian has been the hot topic of the research areas of computer vision and pattern recognition at home and abroad. From this research starting to design a more in line with the human visual system structure of the computer model is built, a kind of more close to the human brain cortex of processing system to solve practical engineering and scientific research in the field of Von Nouma computer difficult problem to solve, Pulse coupled neural network (PCNN) in the1990s was born.PCNN (Pulse Coupled Neural Network) has two properties, namely, the special neurons linear additivity and biological neurons unique nonlinear multiplication modulation, is a different from the traditional Neural Network of a new generation of artificial Neural Network. Sum up with synchronous pulse PCNN issuing phenomenon, dynamic pulse issuing phenomenon, capture characteristics, threshold dynamic characteristics, nonlinear modulation characteristics and comprehensive space-time characteristics, can very well simulate the biological neurons motivation, fatigue and refractory period, and so on. Therefore, enables it to good to simulate the real biological neurons, the signal processing capabilities become stronger. At present, PCNN is widely used in image processing fields, shows unique superiority, and on the further research has important theoretical and practical significance.This master’s thesis discusses some PCNN models and the basic principle of PCNN. Using advantages of PCNN in image segmentation and binary image and combining the histogram equalization and density distribution function, this master’s thesis puts forward two retrieval methods of trademark images, and as a shape similarity criterion, and application to the trademark image searching. The research work and research results of this master’s thesis are as follows:First, review the classical model and the principle of pulse coupled neural networks; the improved PCNN model.Second, analyzes application and principle of PCNN in image retrieval and image segmentation. Third, combining with PCNN image segmentation technology and the histogram equalization technique, put forward a kind of effective image preprocessing algorithm to reduce influence of color for trademark gray distribution difference. And then put forward an edge time sequence concept based on PCNN, and used to extract the form features of the trademark image, and finally realize quickly effective retrieval of the trademark images.Fourth, analyze the binary image algorithms of PCNN and its application study.Fifth, propose a new trademark retrieval method combining with pulse coupled neural network model (PCNN) and regional shape characteristics (density distribution characteristics). In order to ignore the influence of color for trademark image retrieval, segmentation the trademark image into black and white binary image using PCNN, extract the target shape area of the image. In the obtained binary image by extracting the density distribution characteristics of the target area, which is used to describe the distribution characteristics of the target area for the trademark image in space, and finally achieve quickly effective retrieval of the trademark images. This method can effectively retrieve out the trademark image corresponding to be retrieved trademark image, can be to adapt to the changes of the color of trademark image, translation changes, rotation angle and local shape change, reflect the good retrieval performance.