Multi-sensor image fusion by NSCT-PCNN transform

With the character of homologous and heterologous multi-sensor images, a novel image fusion algorithm by NSCT-PCNN transform was proposed. Above all, the registered input images are decomposed by nonsubsampled Contourlet transform (NSCT) and the edge textures of two-dimension or high dimension are accurately extracted. Then, the improved pulse coupled neural network (PCNN) is applied to high frequent subband coefficients integration, while the regional variance integration rules are for the low-pass subband part. Finally, the fusion image is achieved by inverse NSCT on the above-mentioned subband coefficients. The simulation experiments show that compared with the result of Laplacian pyramid transform, Mallat wavelet transform and Contourlet transform algorithm, that of the proposed method have the better visual effect and objective quantitative indicators, meanwhile solve the problem of information loss in subsampled process.