Signal-dependent film grain noise generation using homomorphic adaptive filtering

Abstract

A technique for the generation of signal-dependent film grain noise is presented. It finds applications in television and motion picture productions where artificial film grain noise is added to computer-generated synthetic images to give them a realistic appearance. After decoupling the signal from the noise, using the generalised homomorphic transformation, the noise is removed from the noisy image using the proposed adaptive filtering scheme. The noise parameter is then estimated using the higher-order statistics (skewness or kurtosis) of the observed image and the filtered image statistics, and it is used to generate film grain noise. Computer simulation results demonstrate the effectiveness of the proposed method.