Knowledge-Based Oceanographic Image Compression
Abstract
The Navy has a requirement for image compression on the order of 30:1 for transmission of satellite imagery to the Fleet. Many common image compression algorithms will not retain the desired image fidelity at this compression ratio. This study investigates the use of domain specific oceanographic knowledge in the compression process as one means of achieving the required performance. A transform-based compression algorithm, which draws upon a large, representative image archive for a statistical description of oceanographic knowledge, is proposed and demonstrated. Very high compression ratios seem to be achievable by this method (much greater than 30:1), but the exact evaluation of performance potential requires further study. Among the unique characteristics of this approach is the ability to fill in cloudy areas with estimated image features. While knowledge-based image compression shows much promise for very high compression ratios, it does have significant drawbacks. The most significant of these are the requirement for a large image archive and much computation in order to build the region specific, knowledge- based algorithm. For these reasons, the compression approach demonstrated here is not seen as the single solution to all of the Navy's satellite image compression problems. However, in many applications the proposed algorithm could be very beneficial. In concert with other more generic image compression algorithms, knowledge-based image compression should play an important role.... Remote sensing, Image compression.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1993
- Accession Number
- ADA265752
Entities
People
- Juanita R. Chase
- Ronald J. Holyer
Organizations
- United States Naval Research Laboratory