|
 |
What is USES?
Universal Source Encoding for Science Data (USES), originally developed, as
Universal Source Encoding for Space (USES), is a lossless data compression
technique best suited for science data of various instruments. It compresses
science data into a smaller volume and after decompression (decoding) is applied,
the reconstructed data is EXACTLY the original data; no distortion or loss of
even one bit is incurred in the whole process.
This technique is based on an
algorithmic architecture originally devised by Robert Rice at Jet Propulsion
Laboratories in the early 70's. It was later fully analyzed and enhanced by a
research group led by Dr. Pen-Shu Yeh at Goddard Space Flight Center in the
late 80's and early 90's.
The final version won the NASA Government Invention of the Year.
Features:
- No code tables needed, no update on statistics
- Adapts to local statistics automatically, optimal over whole data set
- Works on data of any quantization levels, be it 8-bit, 10-bit, 16-bit or more
- Speed relatively independent of quantization levels
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
How does it work?
As Figure 1.1 shows, the architecture generally consists of two parts: a
preprocessor and an entropy coder
- Preprocessor: used to decorrelate data, can be user-defined
- Entropy coder: has a set of options, each is optimal in an entropy increment range of 1 bit
- Processes a block of data, say 8, 10 or 16 samples at a time, selects the entropy coder option
with the best performance, signals the option with ID bits, then codes data with this option
- Each entropy option is a unique Huffman coder WITHOUT a code table
What types of data is this technique best for?
This technique is best suited for raw science data in integer format. Simulations have also
shown that it can work on real type science data as well. Some examples are:
- Remote sensing data including various kinds of instrument onboard space craft, such as
Landsat Thematic Mapper, EOS, MODIS, Solar Spectrometer, Hubble Space Telescope instruments, ...
- Medical imaging data: MRI, CT, digital angiograph, microscopic video, digital mammogram, ultrasound, ...
- Seismic data, seismic traces
Note that the entropy coder options in Figure 1.1 work best theoretically when the decorrelated data
samples have statistics that are unimodal and centered at integer value 0. Therefore, the function of
the preprocessor in Figure 1.1 is to decorrelate input data into statistics as close as possible to the
unimodal distribution.
How much compression can be achieved?
Definition of Compression Ratio (CR): Data Quantization Level in Bits / Average Coded Bits per Data.
- If original data is 8-bit, average coded bits per data is 4.2, then CR = 8/4.2 = 1.9.
- If original data is 12-bit, average coded bits per data is 4.0, then CR = 12/4.0 = 3.0
- If the original data is 12-bit, but is stored as 16-bit data words, and the average coded bits per data is 4.0 bits, then
- CR (actual) = 12.0/4.0 = 3.0
- CR (for file) = 16.0/4/0 = 4.0.
Examples:
• Landsat Thematic Mapper Data: 8-bit data, CR ~ 1.8
 Thematic Mapper Data
• GOES imager data: 10-bit data, CR = 10/5.19=1.93
 GOES Image Data
• Seismic data: 16-bit data, CR = 16.0/8.12 = 1.97
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CCSDS LOSSLESS DATA COMPRESSION TECHNOLOGY USERS
| |
Projects |
Launch Date |
Lead Agency |
HW/SW |
|
| |
Mars Observer |
09/92 |
NASA/JPL |
SW |
|
| * |
SERT-96 (Sounding Rocket) |
11/96 |
NASA/GSFC |
HW |
+ |
| * |
Mars-96 |
11/96 |
RSA |
SW |
|
| |
COBRA |
/97 |
DOE |
HW |
+ |
| * |
LEWIS/SSTI |
08/97 |
NASA/HQ |
HW |
+ |
| * |
CASSINI/CDA Cosmic Dust Analyzer |
10/97 |
NASA/JPL |
SW |
|
| * |
SERTS-97 |
11/97 |
NASA/GSFC |
HW |
+ |
| * |
SWAS/SMEX-3 |
01/99 |
NASA/GSFC |
SW |
+ |
| |
KOMPSAT-1 |
/99 |
KARI |
HW |
+ |
| * |
IMAGE/MIDEX |
02/00 |
NASA/JPL |
SW |
|
| * |
Mars Odyssey/THEMIS |
04/01 |
NASA/JPL |
HW |
+ |
| * |
MAP/MIDEX |
06/01 |
NASA/GSFC |
SW |
+ |
| |
INTEGRAL |
10/02 |
ESA |
SW in ADA |
|
| * |
ROSETTA |
01/03 |
ESA |
HW |
+ |
| * |
SBIRS |
Multiple |
DOD |
HW |
+ |
| * |
ESDIS-HDF5 |
07/03 |
NASA/NCSA |
SW |
+ |
| |
SIRTF |
08/03 |
NASA/JPL |
SW |
|
| * |
AURA/EOS |
07/04 |
NASA/GSFC |
HW |
+ |
| * |
MESSENGER |
08/04 |
NASA/APL |
SW |
|
| |
PICARD |
2005 |
CNES |
SW on DSP |
|
| * |
NPP |
? |
NOAA/NASA |
HW |
+ |
| |
GOES/ABI |
? |
NOAA/NASA |
HW |
|
| * |
JWST |
? |
NASA |
HW |
|
| * |
GLORY |
? |
NASA/GSFC |
HW |
|
* Project supported by GSFC
+ H/W, S/W by CAMBR, U. Idaho
| |
