Share Email Print
cover

Proceedings Paper

Vector quantization with distance constraints for enhanced post-processing
Author(s): Thomas P. O'Rourke; Robert L. Stevenson
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

We have previously used a MAP estimation technique to remove quantization noise from decompressed images. This constructed optimization type of post-processor has a tendency to over-smooth the image removing detail information. The proposed quantizer provides additional information about the original image vectors to reduce the size of the constraint space. For each image vector, in addition to the index of the codebook vector, the quantizer also transmits the distance from the original image vector to the codebook vector. This distance information reduces the constraint space to a solid hyper-sphere which will generally be smaller than the polyhedron defined by the codebook vectors alone. The projection onto the solid hyper-sphere also requires much less computational effort than the projection onto the polyhedron. Preliminary experimental results have shown that the VQ with additional distance constraints allows the post-processor to reduce the visibility of quantization errors while retaining detail information that otherwise would have been smoothed out of the post- processed image. The size of the constraint regions could also have been reduced by increasing the codebook size. VQ with distance constraints will be compared with VQ with a larger codebook size at the same data rate. Experimental results will be shown.

Paper Details

Date Published: 13 March 1996
PDF: 12 pages
Proc. SPIE 2669, Still-Image Compression II, (13 March 1996); doi: 10.1117/12.234766
Show Author Affiliations
Thomas P. O'Rourke, Univ. of Notre Dame (United States)
Robert L. Stevenson, Univ. of Notre Dame (United States)


Published in SPIE Proceedings Vol. 2669:
Still-Image Compression II
Robert L. Stevenson; Alexander I. Drukarev; Thomas R. Gardos, Editor(s)

© SPIE. Terms of Use
Back to Top