Night vision technology is becoming ever-more widely used in military and civil fields, and it will be more accurate for target detection and recognition through color fusion of infrared and low light level images. Since the classic Waxman fusion model-only simulates the rattlesnake's IR-depressed Visual Cell and the target in fusion image is not obvious, a novel fusion model is proposed in this paper. We enhance the edge information through the ON neural network for the infrared and low light-level images and then establish a mathematical model to process the rattlesnake's “enhanced cells”and “depressed cells”. Next, we input the ON-central receptive field for fusion and RGB spatial mapping, which can fully realize the union function of the “enhanced cells”and “depressed cells”. Finally, we conduct comparative experiments and image quality evaluation with the classical Waxman fusion model. The results show that image targets are more obvious obtained by our algorithm and increased by an average of 51.97%, 4.07%, and 7.62% than Waxman algorithm in terms of color, mutual information, and structural similarity, respectively. It turned out that our fusion images are richer in color than the Waxman fusion images, which contain more source image information, and more similar to the source image structure.