Abstract: The noise is eliminated from the images by carrying out a first Fourier transform of the images of an original batch of images. Each of these transformed images is then broken down into Fourier series to reveal the main harmonics of a periodic evolution of the original images and of the transformed images. This harmonic breakdown is used to recompose transformed images that are thus filtered by the abandoning of the non-useful harmonics. The filtered transformed images are used to carry out a reverse Fourier transform to obtain, ultimately, filtered original images. It is shown that the filtering in this case removes the noise without lowering the quality of the structures of the image, without distributing the noise among all these structures.