Systems Seminar

Recent Progress in Decompressed Image Artifact Reduction

Prof. Yu Hen Hu
UW ECE Department

Abstract

Image and video coding standards such as JPEG and MPEG are the pillar technologies facilitating digital library deployment. An emerging trend is to develop very low bit rate image and video coding methods to transmit visual information through wireless channels. With stringent bit-rate constraint, the image and video coding algorithm can transmit mainly the low frequency components of the image or video frame where most energy concentrate. As a result, visually annoying coding artifacts such as blocking, ringing, and smearing will appear in the reconstructed image or video, degrading the quality of the transmitted visual information.

In this presentation, some recent results developed here at Madison will be briefly reviewed. They include (a) Dithering Based Blocking Effect Reduction; and (b) Ringing effect reduction using Mathematical Morphology.

Dithering is a process of injecting high frequency noise into the image. While applying a regularization approach to eliminate unwanted blocking boundaries, the resulting image looks "artificial" as a result of missing high frequency information. We incorporate dithering as part of a constrained minimization framework and show that the optimal solution is visually pleasing.

Ringing is an annoying artifact arise in low bit-rate wavelet compressed image. Ringing appears as false parallel edges to a major edge in the image. Our approach is to use mathematical morphological operators to develop a mask of regions in an image which are susceptible to ringing artifact. Then a nonlinear morphology filter is applied to reduce the ringing effect.

Time and Place: Wed., Nov. 11, 3:45-4:45 pm in 4610 Engr. Hall.

SYSTEMS SEMINAR WEB PAGE: http://www.cae.wisc.edu/~gubner/seminar/