GENERIC ARCHITECTURE FOR SPATIAL, SPECTRAL, AND TEMPORAL COMPRESSION

THERE IS A NEED FOR HIGH-SPEED, EFFICIENT, FLEXIBLE IMAGE COMPRESSION ALGORITHMS FOR A RANGE OF APPLICATIONS INCLUDING SATELLITE IMAGING SYSTEMS AND MULTIMEDIA TELECOMMUNICATIONS. EOSDIS, THE DATA TRANSFER AND ARCHIVING SYSTEM FOR FUTURE NASA EARTH OBSERVING SATELLITES, WILL REQUIRE PROCESSING AND/OR ARCHIVING OF 260-1200 GIGABYTES/DAY FOR PERIODS OF DECADES. FUTURE LANDSAT MULTISPECTRAL IMAGING SATELLITES WILL PRODUCE RAW DATA RATES OF 5 MEGASAMPLES/SEC, THEREBY REQUIRING ON-BOARD COMPRESSION. REAL-TIME VIDEO IMAGE TRANSMISSION INVOLVES DATA RATES OF 30-250 MEGABITS/SEC. IMAGE COMPRESSION INVOLVES REMOVAL OF SPATIAL, SPECTRAL, AND TEMPORAL REDUNDANCY. THIS PROJECT ADDRESSES A GENERIC ARCHITECTURE CONCEPT UTILIZING A LIBRARY OF FUNDAMENTAL BUILDING BLOCKS WHICH CAN BE COMBINED TO COMPRESS MULTIPLE IMAGE SOURCES. POTENTIAL BUILDING BLOCKS ARE PREDICTORS, DCTS, DSTS, QUANTIZERS, ENTROPY CODERS, BUFFERS, ETC. THE FEASIBILITY OF PROGRAMMING A COMMERCIALLY-AVAILABLE DIGITAL VIDEO PROCESSOR CHIP (E.G., C-CUBE CL4000) DEVELOPED FOR REAL-TIME VIDEO IMAGERY FOR COMPRESSION OF OTHER TYPES OF IMAGERY, SUCH AS LANDSAT MULTISPECTRAL DATA, WILL BE DETERMINED. USE OF A COMMON NONLINEAR PREDICTION ALGORITHM TO REMOVE THE SPECTRAL REDUNDANCY IN MULTIPLE BANDS OF MULTISPECTRAL IMAGERY AND THE TEMPORAL REDUNDANCY IN MULTIPLE FRAMES OF AN MPEG VIDEO IMAGE SEQUENCE WILL BE INVESTIGATED. THE VALUE OF USING ONE POWERFUL CHIP FOR HIGH BANDWIDTH COMPRESSION APPLICATIONS SUCH AS REAL-TIME VIDEO AND MULTISPECTRAL DATA IS THAT IMAGE DISTRIBUTION SYSTEMS COULD BE IMPLEMENTED FOR EDUCATORS AND SCIENTISTS AS WELL AS COMMERCIAL USERS FOR ON-DEMAND IMAGE LIBRARY SERVICES.