Research On Image Fusion Based On NSCT And PCNN

Image fusion technology, proposed in the late1970s as well as grew as an emergingtechnology, has been widely applied in the fields of military and civilian. Fusion algorithmbased on image transform domain is a hot research topic at present. With features ofmulti-resolution, multi-direction, anisotropy and translation invariance, the nonsubsampledcontourlet transform (NSCT) is more suitable for image fusion than other multi-scale analysistools. In recent years, image fusion methods based on multi-scale decomposition and pulsecoupled neural network (PCNN) have received better fusion effects.This paper studied the basic principles of NSCT and PCNN, and proposed an imagefusion algorithm based on adaptive PCNN in image NSCT domain. The main contents are asfollows:Firstly, we studied the principles of Contourlet Transform and NSCT, analyzed thefactors in NSCT which affect image fusion quality, and selected the filter combination inNSCT which has the best fusion performance by multiple experiments.Secondly, we proposed an adaptive PCNN model whose neurons’ linking coefficients arespatial frequency (SF) in image NSCT domain and external inputs areSum-modified-Laplacian (SML) in image NSCT domain to overcome the shortcomings oftypical PCNN for image fusion whose neurons’ linking coefficients are same.Thirdly, we proposed a new image fusion algorithm using adaptive PCNN in imageNSCT domain to select the Sub-band coefficients of the fused image, and emphasizedconsistency check for the decision map in the algorithm.Finally, the new image fusion algorithm was used to fuse multi-focus images, infraredand visible images. The results of simulation experiments show that the proposed algorithm hasbetter performance than discrete-wavelet-based, stationary-wavelet-based, NSCT-based, andtypical NSCT-PCNN-based fusion algorithms.