Image fusion based on nonlinear structure tensor

A novel image fusion method based on nonlinear structure tensor is proposed within the wavelet framework. Local geometric information in detail wavelet sub-band is extracted and recognized by structure tensor. A nonlinear TV flow is applied to structure tensor element before calculating the eigenvalue and eigenvector. The source data with larger eigenvalue indicates the direction with more features. An adaptive weight function is constructed to obtain new detail coefficients for the reconstruction of the fused image. Experimental results are given to demonstrate the performance of the new method.