Total variation regularization filtering for video signal processing

Modern consumer television sets need a high quality format conversion, image enhancement, and artifact reduction. Especially artifacts due to block based coding schemes may degrade image quality and make artifact reduction mandatory. A promising approach to high quality filter design is given by formulating the filter task via a variational problem and use partial differential equations to solve it. In this paper the design of a generalized total variation regularization filter is described. This filter can be applied in many video signal processing steps in consumer television as pre-or post-processing filter, for example as pre-processing filter prior to a motion estimation to improve image quality or analysis results. The usage in coding artifact reduction without knowledge of decoding parameters is presented in greater detail and some results are presented.