A Gradient-based Multi-focus Image Fusion Method Using Multiwavelets Transform

This paper presents an approach of multi-focus pixel level image fusion, which is based on multi-wavelets transform and a gradient image representation. By comparing and analyzing all sorts of fusion schemes, two different fusion rules for different sub-band region are chosen. One is a region-based root mean square error weighted average scheme in low frequency region. The other one is an energy-based region weighted method in high frequency region. In order to evaluate fusion image quality objectively, the spatial frequency and definition is adopted as the criterion. Experimental results show that the gradient-based fusion method using multi-wavelets is superior to the scalar wavelet method and the multi-wavelets method.