Fusion performance measures and a lifting wavelet transform based algorithm for image fusion

This paper introduces the concept of fusion symmetry (FS) as a measure of evaluating performance of fusion algorithms. The fusion symmetry measure quantifies the relative distance (in terms of mutual information) of the fused image with respect to input images. The smaller the FS the more symmetric is the fused image, i.e., it captures information from both the input images. The traditional criterion of maximizing the joint mutual information is also quantified and a definition called fusion factor is evolved. An algorithm for image fusion using a lifting wavelet filter is proposed. In this algorithm fusion is performed in the transformed domain. The performance of this algorithm is compared with that obtained using average, Laplacian pyramid based approaches and guidelines for selecting an appropriate image fusion algorithm for different sensor conditions are evolved.