Both infrared and visible cameras are deployed in a modern night-vision surveillance system. Fusion methods incorporated with the color transfer technology have been widely used in such a system, aiming to give the multiband fused image a natural day-time color appearance. However, there always exist the following problems: blurry visual effects and low luminance contrast between targets and background, when conventional methods are applied. To alleviate the aforementioned problems, we propose an adaptive color fusion method for night vision in this paper. An intermediate fused grayscale image is produced from the source images via the S-shaped function and the local energy metric in the NSCT (the nonsubsampled contourlet transform) domain. Subsequently, the fused grayscale image is assigned to the luminance channel of YUV color space during the color fusion procedure, so as to improve the target detectability, as well as keep the abundant details shown in each band. Experimental results indeed accord well with our expectation.