Image Compression Research.
Abstract
The vast quantities and high generation rates of tactical imagery require very efficient data compression in order to conserve precious bandwidth for transmission and to limit the required storage volume for archiving. This report describes the results of efficiency and image quality comparisons for several transform image coding techniques. Research effort focused on developing the Singular Valve Decomposition (SVD) as an approach to image compression. Detailed comparisons were to the two-dimensional cosine transform, Hadamard transform and Karhunen-Loeve techniques. High and low altitude aerial imagery, and IR and SAR imagery were included. Images were coded with rates in the range .25 to 1.5 bits per pixel (bpp). (Original images were digitized at 8 bpp.) The order of performance, as measured by rms error (reconstructed image versus original) versus bpp, as well as by visual image quality judgements, was first, cosine and Karahunen-Loeve (nearly the same), second, SVD, and third, Hadamard. Computational burdens, however, are least for the Hadamard, intermediate for cosine and Karhunen-Loeve and most for the SVD technique.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1981
- Accession Number
- ADA110811
Entities
People
- Michael S. Murphy