Primal-dual algorithm for solving a convex image dejittering model with hybrid finite differences

Abstract

Jittering is a common phenomenon arising from the area of multimedia data compression and wireless video transmission. The visual abnormality of a jittered image is the jag in edge and loss of synchronization in latitudinal direction. Typically, the problem of intrinsic image dejittering is challenging to be tackled because of the ubiquitous noise in jittered data. In this paper, we develop a convex variational model for solving image dejittering problem by exerting high-order finite differences regularizer in objective function and exploiting linearization to constraints. Upon the recent progress in convex optimization community, the proposed model can be efficiently solved by the first-order primal-dual algorithm. Numerical simulations on recovering both noiseless and noisy jittered data demonstrate the compelling performance of the proposed model.