Evolved Multiresolution Transforms for Optimized Image Compression and Reconstruction Under Quantization
Abstract
State-of-the-art image compression and reconstruction techniques utilize wavelets. Beginning in 2004, however, ongoing research at Wright-Patterson Air Force Base (WPAFB), the University of Alaska Anchorage (UAA), and the Air Force Institute of Technology (AFIT) has demonstrated that a genetic algorithm (GA) is capable of evolving nonwavelet transforms that consistently outperform wavelets when applied to a broad class of images under conditions subject to quantization error. This report describes recent research that builds upon those previous results in each of the following ways: 1) First, this research demonstrates that a GA can evolve a single set of coefficients describing a single matched forward and inverse transform pair that can be used at each level of a multiresolution transform to simultaneously minimize the size of the compressed file and the squared error (SE) in the reconstructed file. 2) Second, this research examines the relationship between the specified quantization level and the performance of the evolved transform relative to the wavelet. 3) Third, this research extends the GA to simultaneously evolve k sets of coefficients - one set for each level of a k-level multiresolution transform-that further reduce error in reconstructed images. 4) Fourth, this research attempts to evolve k-level multiresolution transforms using highly specific images as the training population, in hope that the resulting transform would exhibit better performance (in terms of reduced error) on similar images, but poorer performance against dissimilar images.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2005
- Accession Number
- ADA451265
Entities
People
- Frank Moore
Organizations
- University of Alaska Anchorage