Fusion technique for multi-focus images based on NSCT–ISCM

Abstract

The issue of fusing multi-focus images is meaningful and undoubtedly suitable for visual effects and further image processing tasks. In this study, a novel multi-focus image fusion technique based on non-subsampled contourlet transform (NSCT) and improved spiking cortical model (ISCM) is presented. Compared with the current multi-resolution geometric analysis (MRGA) tools, NSCT not only has much better competences of information capturing and feature extracting, but also overcomes the drawback of shift-invariance lacking from which the contourlet transform suffers. As a recently developed biological model, SCM combines the advantages of both pulse coupled neural network (PCNN) and intersecting cortical model (ICM), and has been considered to be an optimal neuron network model recently. The proposed technique is composed of three main phases. Firstly, by using NSCT, each source image is decomposed into a low-frequency sub-image and a series of high-frequency sub-images in different directions. Then, the classic SCM is improved to be ISCM with a less complex structure and much more effective function mechanism, which is responsible for obtaining the fused sub-images. Finally, inverse NSCT is utilized to reconstruct the final fused image. The NSCT–ISCM based fusion algorithm is devised. Experimental results indicate that the proposed technique is superior to other current popular ones in both aspects of subjective visual and objective performance.