Abstract: In this paper, we propose an innovation denoising method named fractional integral denoising algorithm (FIDA) in order to remove noise as largely as possible. Our approach is based on the Riemann-Liouville definition of fractional calculus. The structures of FIDA on eight directions are discussed first. In the first aspect,the structures of fractional integral masks for FIDA on eight directions are constructed respectively. The eight directions used in our algorithm are 135 degrees, 90 degress, 45 degrees, 0 degrees, 180 degrees, 315 degrees, 270 degrees and 225 degrees. In addition, we also present the numerical implementation rules of FIDA for digital image. The experimental results show the effectiveness of our method according to the visual perception and peak signal noise ratio (PSNR) subjectively and objectively. Those results also demonstrate that FIDA can effectively remove noise while preserving the image's significant information simultaneously, especially for the edges and texture information with weak variation on gray intensity.