Removal of grid-pattern noise introduced by M-channel subband image coding

Abstract

A method is proposed for reducing grid-pattern noise introduced by low bit-rate coding of images using integer-modulated filter banks. The noise level can be reduced significantly by applying a predictive subband coefficient compensation algorithm in the iterative synthesis filtering stage. Simulation results show the improvement in visual quality and a decrease in mean square error values of about 2 to 5. The coefficient compensation algorithm is suitable for real-time implementation since it is computationally simple and uses only integer operations.