A Projection-Based Metric for the Quantitative Evaluation of Pixel-Level Image Fusion

Pixel-level image fusion has been investigated in various applications and many algorithms have been proposed. However, few authors have addressed the problem of how to evaluate the performance of those algorithms. In this paper, we propose a novel metric for blind quantitative evaluation of pixel level image fusion, i.e., no reference image is needed. Firstly the signal characterization of input images are projected onto their counterparts of the fused image to obtain projection-based values (PVs). Then we use structure distortion (SD) intensity to measure how many information have been transferred from each source into fused result by the difference of PVs between them. Finally all the SD pairs are incorporated into one expression according to Weber-Fechner law. Experimental results demonstrate that our proposed metric is compliant with subjective evaluations and out performs other recently developed blind evaluation metrics of pixel-level image fusion.