In order to improve the overall effect of low illumination image denoising, a low illumination image smoothing denoising method based on fuzzy mean difference is proposed in this study. Sobel gradient is used to detect the image edge information, and low illumination images were segmented into flat and detail regions by OTSU (Nobuyuki otsu) threshold segmentation; The noise standard deviation of two regional images is estimated by fuzzy mean difference method; The wavelet domain method is used to smooth the soft threshold denoising of the image to realize the smooth denoising of the low illumination image. The experimental results show that this method can obtain the noise standard deviation close to the real value under different noise levels. When the noise level is the largest, the signal-to-noise ratio reaches 27.97db, the denoising effect reaches 92.1%, the proportion of the number of images with good quality reaches 80.58%, the image information loss is small, the denoising effect is good, and it has great application value.