Natural color fusion of long-wave infrared and visible images is helpful to observers for quick perception of the environment and discovery of targets. The coaxial optical structure enjoys the unique advantages in terms of improving the quality of the fused image and simplifying the registration algorithm. According to the requirements in battlefield detection, surveillance, and so on, a real-time coaxial color fusion system based on TMS320DM642 DSP and ATmega32 MCU was designed in the present work. In the fusion system, the long-wave infrared imaging system is combined in parallel with the visible imaging system. The optical path was separated by a beam splitter and a mirror. The dual-band images were absolutely registered using field-of-view matching and optical-axis calibration. The image processing program was developed based on a DSP/BIOS embedded operating system. A natural color fusion algorithm based on a CbCr look-up table was adopted according to the features of the video image. The fusion algorithm was optimized according to the hardware features. The fusion time of a single-frame image with resolution of 720 pixel×576 pixel is 9.18 ms, which satisfies the real-time fusion requirements of 25 frames per second. The fusion image exhibits a natural color, and the hot targets in the scene can be readily identified. The research on related problems presented in this paper can provide references to the design and implementation of a coaxial image fusion system.