Simulating Film Grain using the Noise-Power Spectrum

Abstract

Film grain is an essential part of real images. The artifacts it introduces add character to images, which can otherwise appear too perfect. This paper considers the synthesis of film grain based upon its noise-power spectrum, producing grain fields which approximate real film over a range of enlargements and densities. We also identify limitations of the noise-power spectrum in that it fails to fully characterize grain. Simulated grain fields have spectra which match real grain, but lack the phase and density correlations required to create the appearance of individual grains at high magnifications.