Research Of Non-subsampled Contourlet Transform And Adaptive PCNN In Image Fusion

As a technique for refining multi-channel information,image fusion can enhance image information in the same scene.For a limited scene or object,multi-source images obtained with several identical collectors or several collectors contain rich redundant information and complementary information.Fusion of redundant information and complementary information is beneficial to improve the signal-to-noise ratio of the image,that can also save more texture and detail information in the fused result.It's precisely because of the extraction of information from multi-source images that the fused image has a more accurate and comprehensive description of the scene than any single image with a fixed angel,and the result can better meet the human eye and application.In this paper,Contourlet transform and Non-Subsampled Contourlet transform(NSCT)are applied in image's multiscale analysis,and we strive to study flexible and stable image fusion algorithms.Here are several innovations:In order to solve the problem that the traditional Pulse Coupled Neural Network(PCNN)algorithm cannot adapt to the characteristics of the image well,an image fusion framework based on Non-Subsampled Contourlet transform and adaptive Pulse Coupled Neural Network is proposed in this thesis,the role of adaptive pulse coupled neural network in multi-focus image fusion is analyzed from subjective and objective evaluation.The comparison of the experimental results proves that the image generated by the algorithm fusion is superior in both intuitive observation and objective performance.In order to solve the problem that the clarity index is not sensitive to the correlation between image pixels,this thesis proposes a multi-focus image fusion algorithm based on Non-Subsampled Contourlet transform and Kernel Density Estimation(KDE).The kernel density estimate value of pixels in the subband image of the Non-Subsampled Contourlet transform,is used as the basis for the calculation of the sharpness index,and fuse the subband image according to the principle of large corresponding index.In the experimental results of image fusion,the algorithm is superior in image visual effect and objective criteria.Finally,for the sparseness of the subbands obtained after Non-Subsampled Contourlet transform,sparse representation is applied to guide the fusion of subband images in this thesis,.The K-means singular value decomposition(KSVD)joint training is performed on all subband coefficients by region to obtain a local joint dictionary.According to this dictionary,the orthogonal matching tracking algorithm(OMP)is used to solve the sparse coefficients of each subband coefficient,and the norm of the sparse coefficients is calculated to guide the fusion subband image with the principle of large norm.Through the analysis of the objective criteria of the algorithm before and after the guidance,the effectiveness of the sparse representation in retaining the background information of the image is demonstrated.