The invention discloses a noise space characteristic based nonlinear filter denoising method, which comprises the following processes of: (1) inputting an image f detected by an infrared imaging detector; (2) establishing an image g which has the same size as that of the image f and has gray value of 0; (3) performing single-point denoising on the image f with noise; (4) performing two-point denoising on the image f with noise; (5) performing three-point denoising on the image f with noise; and (6) taking the gray value of a pixel, of which the gray value is zero and which is amended through the single-point denoising, the two-point denoising and the three-point denoising, in the image g as the gray value of the point in the image f. By adopting the method, blindness during denoising of the conventional nonlinear filter denoising method is made up, and difference of the noise space characteristic and the target space characteristic is fully utilized, so that detail and edge of the image can be kept well during denoising, and subsequent works such as image correction, target identification and the like are facilitated.