Research On Image Fusion Technology Based On PCNN

Pulse Coupled Neural Network(PCNN) is a kind of Neural Network based on biological background. As the third generation of artificial neural network, PCNN has very extensive application in the field of image processing, mainly used in image denoising, image segmentation, image enhancement, it mainly used in image denoising, image segmentation, image enhancement. Because of PCNN is closer to biological neural network, make it has the characteristics that traditional neural networks without them, such as pulse distribution of synchronization, variable threshold, wave formation and propagation of features, these features can be very good application in image processing, which is based on the image fusion technology is used in pulse coupled neural network provides the theoretical significance and support.This paper first introduce the basic principle of PCNN model, elaborate its operation mode; analyze the characteristics of PCNN, discusses the application method of the model in the field of image processing and its advantage; through discussing the characteristics of PCNN, we can get that combined multi-scale Nonsubsampled Contourlet Transform(NSCT) with PCNN fusion algorithm and its application value will greater.Compared with traditional wavelet transform, discrete wavelet transform, the ridgelet transform, NSCT have translation invariance can get a better result in image fusion processing. This paper introduce the theoretical knowledge and its components in detail of NSCT transform, summed many excellent features of NSCT, show that NSCT transform have obvious advantages in image processing.Image fusion of this paper is improved base on traditional combination of PCNN and NSCT transform. NSCT transformed the source image(the two images to be fused), judgement about their focus situation:(1) the focus both are clear / fuzzy area are weighted according to the energy coefficient;(2) the focus clear / fuzzy situations are on the contrary, use regional variance as the low frequency sub-band fusion coefficients selection rule. The selection rule of bandpass directional subband fusion coefficients is use regional variance as the low frequency sub-band fusion coefficients selection rule, Respectively, using the sharpness and regional energy as the PCNN external input and links intensity, then, after NSCT inverse transformation to get the final fusion result.According to combine the translation invariance of NSCT and global coupling pulse synchronization features of PCNN, this paper proposed the based on dual-channel PCNN and NSCT combination image fusion method. NSML representative at the edge feature of the low-frequency sub-graph and used to motivate the adaptive PCNN neurons. For high frequency sub, Spatial frequency that after NSCT decomposited as PCNN external input. Through a large number of experiments, to evaluate the performance of fusion, the result shows that both from the visual quality and objective evaluation, the proposed scheme is superior to wavelet and PCNN, NSCT and PCNN fusion method, has a realistic significance.