Modified Kalman filtering filtering for image super-resolution: experimental convergence results

Abstract

Video super-resolution is the process of estimating a highresolution image from a motion sequence of low-resolution image frames. This paper examines the convergence properties of the modified Kalman filter super-resolution algorithm introduced in [1]. Most significantly, the ratio of the system and measurement noise variances is shown to be a highly useful control parameter of the convergence rate, super-resolution image sharpness, and the algorithm's behaviour in the presence of noise.