Novel multi-focus image fusion method in nonsubsampled Contourlet transform domain

Abstract

A fusion method using pulse-coupled neural network(PCNN) in non-subsampled Contourlet transform(NSCT) domain was proposed in order to solve the problem of multi-focus image fusion in the same scene.Both the low-pass sub-band coefficient and the band-pass directional sub-band coefficient of source image by NSCT were inputted into PCNN.The ignition mapping image was obtained via the ignition frequency generated by the neuron iteration.Then the approach degree function was adopted to describe the association degree of the neighborhood characteristic in ignition mapping image,and the appropriate sub-band coefficient was selected according to the neighbor approach degree.The fused results were obtained through the inverse NSCT.Experimental results demonstrate that the method greatly retains the clear region and the feature information of multi-focus image.The method has better fusion performance than the classical wavelet transform,the Contourlet transform and the conventional NSCT.