Spatio-temporal post-filtering of 3-D coded video sequences

Abstract

Lossy video coding using wavelet transforms at low bit-rates results in coding artifacts mainly observed as ringing effects. The performance of a coding system can be additionally enhanced if a spatio-temporal post-processing algorithm is applied at the end of the coding chain. The description of an adaptive spatio-temporal algorithm is the objective of this paper. The algorithm takes into account the motion information to avoid temporal blurring of moving edges and spatial edge information to avoid spatial blurring while implementing temporal and spatial filtering in order to alleviate the presence of the artifacts. Motion information is used in order to tune the algorithm adaptively so that motion compensated temporal filtering is applied only in non-motion areas and motion areas with low prediction error. The proposed algorithm uses adaptive spatial filtering that is based on edge detection and ringing area estimation for pixel classification in edge and non-edge corrupted pixels. Spatial filtering leaves edges unaffected to preserve image details. Experimented results show significant increase in perceptual quality of wide sequences.