Application Of Image Reverse Data Process Based On Nueral Networks

Recent year with neural network have many characteristics, the applicationfields of it also have extended widely. In this paper neural networks arithmeticand digital image process technique are addressed and applied in the imagereverse processing to solve some key questions of engineering image, such asmass data, feature exaction, and defect image. The purpose of it is to satisfy theimage reverse request.Although arithmetic of neural network is filtered through to some aspects ofthe image processing, but some of the effect is not as good as user expected, suchas distortion in the processing that makes engineering disastrous and so on. Sinceno best way exits, there is a lot of work to do to improve the image processing.The main job of this paper has three aspects. First of all image digitalizing andsmoothing processing, use one technology of the noise Median filtering in thesmoothing processing;Secondly image feature extraction by the SOFM neuralnetwork and emulating image proceeding. Because of the numerous image dates,image compression is the key technology which ensures image processing workeffectively. A new type two-layer BP neural network (momentum type two-layerBP network) is proposed to compress image which its compress effect is comparewith other BP neural net;Thirdly according to both RBF neural network and BPneural network have some disadvantages in the image inpainting, a BP-RBFcombined neural network is proposed. Compare with the each of the formerneural network. Matlab is used in the Windows 2000 for application programdevelopment and emulating image proceeding.Image reverse proceeding technology based on neural network integratedwith the neural network technology and the Image proceeding technology. Inthis paper, it is innovative to propose the RBF-BP combined neural networkwhich is significant in theory and practice in the field of image inpainting.