Image Compression
Abstract
During May and June we investigated correlation methods for determining affine iterated function sets. Our basic method is to compare portions of an image with other portions of the same image. The affine transformations which we choose by this process must therefore specify range and domain as well as gray level. Although the compressions we attain by such codes run from 10:1, to 20:1, this method has the advantage of being completely automatic and requiring no pre-processing. Furthermore our image encoding is relatively rapid, requiring approximately 30 minutes for a 256 by 256 image, when the algorithm runs on a Sun 4 computer. We have run the algorithm on 10 separate images. During the next two months we will investigate several potential improvements in the code. These include modifications of the tiling, relaxing conditions on the multiplicity of the mappings (the trade is size of the tiles vs the number of segments mapping into a given tile) and the use of Fourier transform method to improve the choice of affine maps.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1990
- Accession Number
- ADA224242