OPTIMAL PERFECT RECONSTRUCTION FILTER BANKS FOR MULTIRESOLUTION CODING

MULTIRESOLUTION COMPRESSION ALGORITHMS OFFER POTENTIAL ADVANTAGES OVER MORE TRADITIONAL VECTOR BASED ALGORITHMS (TRANSFORM CODING, VECTOR QUANTIZATION) SINCE THE BLOCKING ARTIFACT IS NOT PRESENT AND EDGE FIDELITY CAN BE IMPROVED. THE MULTIRESOLUTION REPRESENTATION ITSELF PROVIDES A SOLUTION TO MANY IMAGING APPLICATIONS WHICH NEED TO HANDLE THE SAME IMAGE AS DIFFERENT RESOLUTIONS FOR DISPLAY, PRINTING, BROWSING, ETC. A MULTIRESOLUTION COMPRESSION ALGORITHM, BASED ON PERFECT RECONSTRUCTION QUADRATURE MIRROR FILTERS (QMFS) WHICH ARE USED IN A TREE BASED STRUCTURE TO SPLIT RECURSIVELY THE LOW FREQUENCY BAND, IS BEING CONSIDERED. WAVELETS ARE A SPECIAL CASE OF THESE FILTERS. THE RESEARCH PROVIDES A TECHNIQUE WHERE OPTIMAL FILTERS ARE DESIGNED AND APPLIED TO TYPICAL IMAGERY. THESE FILTERS ARE COMPARED WITH DAUBECHIES' AND MALLAT'S WAVELETS WHICH ARE COMMONLY USED. BOTH OBJECTIVE AND SUBJECTIVE MEASURES ARE MADE AND AN OPTIMAL QMF BANK SELECTED.