Based On The Gabor Wavelet Transform And Moments Transform Neural Network Image Recognition Technology

With the development of computer technologies and human sensation researches, the technology of image recognition became one significant field of image processing. It can be applied in almost all human lives. But, it is limited in its shortcomings of lacking adaptation and not being used in different field. Thus, how enhance its adaptation, robustness, even universality is a problem.On the basis of enhancing adaptation, robustness, even universality of image recognition technology, this dissertation researched the image recognition technology. According to the phases of image recognition, it researched and discussed in three sections such as, image preprocessing, feature extraction and pattern recognition.In the phase of image processing, this dissertation introduced, discussed and simulated the traditional methods of image restoration, image enhancement and image segmentation. Furthermore, it proposed a new adaptive threshold median filter to remove multiple-impulse noise. Moreover, it processes better ability of attenuating multi-layered impulse noise than other methods.In the phase of feature extraction, this dissertation proposed to use Gabor wavelet transform to get global feature of image and use moment transform to get local feature. Then, combining the global feature and the local feature will fuse into new feature to be processed in the pattern recognition phase.In the phase of pattern recognition, this dissertation adopted artificial neural network simulating the structure and function of human neural network to judge and assort. It took the fusing feature to be the input of neural network to train and recognize. Last, the new technology is completed. The experiment results show that it owns better adaptation, robustness, even universality than neural network image recognition based on the Gabor wavelet transform or the moment transform.