A comparison of image fusion quality metrics

Abstract

Measuring the fitness of the fused image plays a key role in image fusion applications. For a learning process performed in a machine learning algorithm, the result of the fusion should be evaluated numerically. In the literature, there are well-known quality metrics developed for this purpose. Each metric evaluates the quality of the image using a different method. However, to be used in the learning process, the quality metrics must be able to provide results compatible with the change in the image's visual quality. In this study, synthetic images with known quality levels were created for this purpose. The scoring accuracy of six quality metrics commonly used in the literature was compared with these test images and the results were evaluated.