Objective content-dependent quality measures for image fusion of optical data

A new approach to obtain objective measurements for evaluating the performance of the actual fusion is proposed. The measures are based on clear physical meanings, straightforward to implement and do not require any ground-truth. The proposed algorithm utilises a content-based approach, splitting the original image into different regions with various spatial characteristics. The spectral and spatial changes for each region between the original multispectral image and the fused product are used to assess the performance of the fusion process. A few commonly used fusion techniques were evaluated using the proposed framework. The results show that the suggested measures are objective, in agreement with previous empirical studies and can easily be used to evaluate the ratio of the spectral/spatial information coming from the multispectral and panchromatic images into the fused product.