Research On Image Fusion Method Based On NSCT And PCNN

An image from a single sensor only can show one-side feature of scene or object. But for the same scene or object, a plurality of images obtained by different sensors, which contain redundant information and complementary information. Therefore the process of image fusion is effectively integrated redundant information and complementary information of images, the final fused image with high quality which has more substantial and accurate information. With the help of image fusion technology, the viewers have a comprehensive understanding and analysis on the measured object or scene and then make accurate identification, location, diagnosis, etc. Image fusion has been widely used in various fields of military, medical and so on.Multi-scale geometric analysis is an effective tool of image fusion, non-subsampled contourlet transform (NSCT) of multi-scale geometric analysis methods has multi-scale, multi direction, translation invariance and other advantages compared with the wavelet, Contourlet Transform (CT), and thus NSCT is more suitable for image fusion. Pulse Coupled Neural Network (PCNN) with bionic mechanism analog optic nerve work of cat, which has changed threshold characteristics, capture feature, pulse synchronization issuance characteristics and so on, which conducive to achieving image fusion. In a word, the method of image fusion which based on NSCT of multi-scale and PCNN has a larger application value.The research contents of this thesis is the fusion algorithm of combination of NSCT and PCNN, rational and effective fusion rules are proposed for low-frequency and high-frequency sub-band coefficients respectively, which aims to improve the sharpness of the fused image and fusion effect. Firstly, the medical images which have been registration strictly are decomposed into low-frequency and high-frequency sub-band images using NSCT. Taking into account the impact of low frequency sub-band fusion strategy on the edge sharpness, information content of the final fused image, fusion algorithm based on the active measurement of information entropy is used to fuse coefficient of low-frequency sub-band. Adaptive PCNN is used to fuse coefficient of high-frequency sub-band, which use modified spatial frequency as external input and weighted sum-modified-laplacian as linking strength, removing block effect due to the discontinuity of blur and clear area of image caused. Fusion coefficient of high frequency sub-band is determined by the number of ignition of PCNN. Finally, the fusion image achieved by inverse transformation of NSCT. In this thesis, to verify the applicability and effectiveness of the proposed algorithm, two sets multi-source image contained brain CT, MRI image and infrared light, visible light images through simulation experiment respectively, fusion results are analyzed and compared with other fusion methods. The experimental results show that the proposed fusion algorithm is more in line with the human visual system, and performs better in subjective and objective assessments.