IMPROVED MPEG-BASED ALGORITHM FOR VIDEO DATA COMPRESSION

EFFICIENT COMPRESSION OF IMAGE SEQUENCE DATA IS A LIMITING TECHNOLOGY FOR MANY DIGITAL VIDEO TRANSMISSION AND STORAGE SYSTEMS. VIDEO IMAGES TYPICALLY CONTAIN A LARGE AMOUNT OF SPATIAL AND TEMPORAL REDUNDANCY; THUS VERY HIGH COMPRESSION RATIOS ARE FEASIBLE IN PRINCIPLE. UNFORTUNATELY, EMERGING VIDEO COMPRESSION TECHNIQUES (SUCH AS THOSE BASED ON THE PRELIMINARY MPEG STANDARD) CANNOT CONSISTENTLY ACHIEVE MAXIMUM PERFORMANCE BECAUSE THEY UTILIZE FIXED IMAGE PARTITIONS AND SIMPLE MOTION COMPENSATION ALGORITHMS. WE HAVE SHOWN THAT SIGNIFICANTLY HIGHER COMPRESSION RATIOS WILLBE OBTAINED BY USING OUR VIDEO COMPRESSION APPROACH BASED ONADAPTIVE SUB-PIXEL MOTION PREDICTION AND MULTI-RESOLUTION IMAGE REPRESENTATIONS. THESE ENHANCEMENTS RESULT IN TWO-FOLD OR GREATER REDUCTIONS IN THE NUMBER OF BITS REQUIRED TO ENCODE THE PREDICTION ERROR AND THE MOTION DESCRIPTION - THE TWO MAJOR COMPONENTS OF THE MPEG INTER-FRAME TRANSMISSION. THE PROPOSED PHASE I STUDY WILL INTEGRATE SUB-PIXEL MOTION PREDICTION AND MULTI-RESOLUTION DESCRIPTIONS WITHIN THE FRAMEWORK OF THE EMERGING MPEG VIDEOCOMPRESSION PROTOCOL, AND WILL DEVELOP A SOFTWARE TOOL FOR OFF-LINE TESTING ON ACTUAL VIDEO DATA. PHASE II WILL REFINEAND SELECT A FINAL SET OF VIDEO COMPRESSION ALGORITHMS, AND INVESTIGATE THEIR IMPLEMENTATION IN SPECIAL-PURPOSE VLSI HARDWARE.