Share Email Print

Proceedings Paper

Image restoration using vector classified adaptive filtering
Author(s): Paul Richardson
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

This paper describes a novel adaptive filtering technique for image reconstruction `Vector Classified Adaptive Filtering' (VCAF) and provides comparative results when reconstructing images corrupted by additive Gaussian noise. Each sample in the image being reconstructed is first classified by mapping a classification vector onto a codebook as is done in Vector Quantization (VQ) image coding. The classification vector is a set of samples from the neighborhood of the one being reconstructed, and the codebook is a set of vectors designed off line using representative images, a model of the distortion function, and VQ codebook design techniques. Once the classification of the local region has been determined an optimal reconstruction filter is used to estimate the correct sample value. In this paper Wiener filtering techniques and a priori information from representative images are used to design least square optimal reconstruction filters for each class in the codebook. Thus VCAF is a very highly adaptive filter that incorporates a priori knowledge of typical image statistics in the form of a classification codebook and optimal reconstruction filter for each class. Experimental results indicate that VCAF performs very well with respect to the techniques with which it was compared; specifically non-adaptive Wiener filtering and edge orientation adaptive Wiener filtering.

Paper Details

Date Published: 22 October 1993
PDF: 11 pages
Proc. SPIE 2094, Visual Communications and Image Processing '93, (22 October 1993); doi: 10.1117/12.157918
Show Author Affiliations
Paul Richardson, Monash Univ. (Australia)

Published in SPIE Proceedings Vol. 2094:
Visual Communications and Image Processing '93
Barry G. Haskell; Hsueh-Ming Hang, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?