Statistics – Computation
Scientific paper
Sep 1999
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=1999aas...19410806f&link_type=abstract
American Astronomical Society Meeting 194, #108.06
Statistics
Computation
Scientific paper
We review the requirements for a compression algorithm compatible with the FITS Standard. We propose an object-oriented generic mechanism for the encoding of data structures with multiple fields and stochastic properties. Using syntax reminiscent of the World Coordinate System FITS extension, we decompose the data into an approximate component and residuals. We propose an adaptive wavelet image coder based on Universal Trellis-Coded Quantization (UTCQ). This algorithm outperforms other vector and lattice quantization techniques of much higher complexity. Coupled with Diagonal Coding of residuals, it permits high-performance data compression with random access to subsets of exact reconstructed data. UTCQ requires no training algorithm, a simple modeling procedure is required for a given source statistical distribution. The model allows UTCQ access to continuous rates without code book storage. The coder is quite robust and even at 64:1 compression performs within 0.3 dB of the rate-distortion bound for a memoryless gaussian source. The compressed imagery is robust to post-processing including histogram equalization, sharpening and edge enhancement without blocking artifact. An embedded zero-tree algorithm permits control of the compression to any desired bit-rate or quantitative quality criterion. Such control ensures the ability to meet any target data quality requirement to preserve the overall error budget in scientific data processing. We thus reduce expenditure on storage devices and computer networks and their manageability is improved. When combined with hierarchical region-based encoding such as the k-d tree considered by AXAF, the computational resource requirements improve by a factor of ln(N)/N where N is the dimensionality of the data. We accelerate calculation of wavelet coefficients for scientific analysis via ultra-low cost commercially-available wavelet compression hardware (e.g. Analog Devices' ADV601) may. We demonstrate some results based on astronomical archive imagery and ancillary data considering future development projects.
No associations
LandOfFree
Proposal for Preliminary Evaluation of an Approximate Data Compression Algorithm for the FITS Standard. does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Proposal for Preliminary Evaluation of an Approximate Data Compression Algorithm for the FITS Standard., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Proposal for Preliminary Evaluation of an Approximate Data Compression Algorithm for the FITS Standard. will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1767205