Adding a narrowband filter in a night vision system can considerably reduce the signal-to-noise ratio(SNR) of output images. To solve this problem, a true-color night vision system based on minus filters is designed. Two comparative experiments involving the detection of a standard colorimeter and natural scenery using the true-color night vision system based on narrowband filters under different ambient illumination are performed. The SNR is indirectly evaluated using the correlation coefficients between the edge maps of the reference and output true-color night vision images. In experiment 1, the results show similar correlation coefficients for the two systems when the ambient illumination is 0.51 lx. For illumination levels of 0.12, 0.04, and 0.01 lx, the correlation coefficients of the proposed system are 14.9%, 26.8%, and 128.0% higher, respectively, than those of the system with the narrowband filters. In addition, the correlation coefficient of the proposed system is also higher than that of the traditional system in experiment 2. Therefore, compared with the traditional true-color night vision system, the true-color night vision system based on minus filters can work under conditions of lower illumination and produce true-color night vision images with higher SNR.