Wavelet packet decomposition for the SPIHT

Abstract

<P>THIS THESIS CONTRIBUTES AN INNOVATIVE COMBINATION OF THE STATE-OF-THE-ART IMAGE CODING ALGORITHM SET PARTITIONING IN HIERARCHICAL TREES (SPIHT) AND THE WAVELET PACKET TRANSFORM (WPT), WHICH SIGNIFICANTLY IMPROVES THE CODING EFFICIENCY OF THE SPIHT ALGORITHM. </P><P>AIMING AT IMPROVING SPIHT BY MERELY A SLIGHT MODIFICATION, THE WPT AND THE SPIHT ARE MATCHED IN A NOVEL WAY. THE NEW METHOD PUTS EMPHASIS ON AVOIDING THE PARENTAL-CONFLICTS PROBLEM. FIRSTLY, THE SUBBAND STRUCTURE OF THE WPT IS THOROUGHLY STUDIED, AND ONLY THE TRANSFORMS WITH THE CROSS SUBBAND SIMILARITY ARE CONSIDERED AS THE CANDIDATES. SECONDLY, AMONG THESE CANDIDATES, THE BEST TRANSFORM IS SELECTED, IN THE SENSE THAT THE ENERGY IS CONCENTRATED TO THE FEWEST TREES (THE BASIC DATA STRUCTURE OF SPIHT); TO ACCOMPLISH THE SELECTION, A PROPER COST FUNCTION IS PROVIDED TO MEASURE THE ENERGY CONCENTRATION ABILITY, AND A FAST SEARCHING ALGORITHM IS PROPOSED TO GOVERN THE SELECTION. IN THIS WAY, THE SELECTED WPT GUARANTEES HIGH