Abstract: Aiming at the problem that the results of optical remote sensing imaging are degraded by noise and haze, a fast noise reduction and haze reduction algorithm combining multi-frame fusion noise reduction and dark channel prior method for haze reduction is proposed. In view of the low main frequency and limited computing power of space-borne computers, this paper changes the pixel-by-pixel registration of the traditional multi-frame fusion algorithm to two-level registration in the noise reduction stage, and the absolute error and phase-related registration are respectively used globally and locally. A new registration quality evaluation method based on the registration result is proposed, which reduces the evaluation time; in the haze removal stage, a Gaussian kernel are used to convolve with the transmittance template, instead of the traditional refinement method. The experimental results show that the algorithm proposed in this paper can increase the average peak signal-to-noise ratio (PSNR) of the image by 8.97 dB. Through the optimization of the amount of calculation, the time consumption of haze removal is reduced by 73.41%.