Two comprehensive evaluation metrics, image perceptual quality based on target detectability (PQTD) and perceptual quality based on scene understanding (PQSU), are proposed to measure image quality for visible and infrared color fusion images of typical scenes. A psychophysical experiment is performed to explore the relationship between conventional quality attributes and the proposed evaluation metrics. The prediction models for PQTD and PQSU are derived by multiple linear regression statistical analyses. Results show that the variation of PQTD can be predicted by sharpness and perceptual contrast between the target and background, and that color harmony and sharpness can predict PQSU. The proposed evaluation metrics and their prediction models provide a foundation for further developing objective quality evaluation of color fusion images.OCIS Visible and infrared color image fusion combines two source images into a single composite false-color image that is suitable for special visual tasks. Many fusion algorithms and systems have been successfully applied. At present, however, no generally accepted methods of color fusion image quality evaluation exist. Image quality assessment has been investigated using subjective and objective approaches. Subjective approaches evaluate the image quality based on subjective perception by observers, whereas the goal of objective image quality assessment research is to design computational models that can predict the perceived image quality accurately and automatically. The numerical measure of quality an algorithm provides should correlate well with human subjectivity [1] . Many factors are able to influence perceptual image quality. For true-color images, Choi et al. performed a psychophysical experiment in which colorfulness, contrast, and naturalness were the key attributes controlling image quality [2] . Pedersen et al. found that color, lightness, sharpness, contrast, physical attributes, and artifacts were the most meaningful attributes for the print image quality [3] . However, unlike true-color images, color-fused images contain dual-band information. Thus, the purpose of image fusion is not to obtain the color image that completely corresponds with the truecolor image of the same scene, but to improve the suitability for special vision tasks (i.e., target detection and scene understanding) [4] . Until recently, most subjective evaluations have been investigated based on different visual tasks, such as target detection and recognition, scene recognition, and situational awareness [5−7] . Shi et al. presented three influence factors (i.e., target detection, detail, and colorfulness) to evaluate the color fusion image quality [8] . Thus far, however, no agreement on which quality metric should be used to evaluate the quality of visible and infrared color fusion images has been reached. For the same image, perceptual quality assessment results differ according to different visual tasks [9] . Therefore, to evaluate the color fusion image quality comprehensively, we prop...