Objective evaluation of color fusion of visible and IR imagery by measuring image contrast

Color night vision through fusion of images derived from visible and thermal infrared sensors has come into application. A lot of research of perceptual or subjective evaluation of color fusion algorithms is conducted by the TNO, MIT and NRL^[1-5]. Objective evaluation is becoming an increasing issue. A good fusion algorithm should preserve or enhance all of the useful features, i.e., targets pop out and content or details from the source images. Objective: The goal of our research is to predict details and target detectability of color fusion images by measuring image contrast. Methods: A color image contrast metric^[11], which was designed to compute how much details for natural images, is employed to measure details and target detectability of color fusion images in this paper. A sub-band contrast index is defined, the first band contrast index indicates the level of details, and higher band contrast indexes indicate the level of target detectability. Results: The perceptual details have high correlation with the first sub-band contrast indexes, and the perceptual target detectability has correlation with higher sub-band contrast indexes. This means the sub-band contrast indexes may predict details and target detectability, however, noise effect on details need to be considered and prediction of target detectability need to be improved.