DESIGN OF SHORT TAP PERFECT RECONSTRUCTION FILTERS FOR SUBBAND CODINGOF IMAGES

Short tap filters reduce the computational complexity of filtering pro cesses and also reduce image degradations associated with ringing effe cts near edges in image subband coding. When the number of taps is red uced, however, coding performance generally deteriorates. In this pape r, short tap filters that satisfy isolated perfect reconstruction cons traints are designed so as to improve coding performance. For the sake of comparisons, linear phase- and orthogonal filters are also designe d. Filters with coefficients determined by the values 2(n) are designe d to reduce further the computational burden of the filter process. As suming that the input can be modeled by an AR(I) process, the set of p arameters which determine the tap coefficients are selected to maximiz e the coding gain. The relationship between signal-to-noise ratio and entropy is determined to evaluate coding performance. The superior cod ing performance of the perfect reconstruction filters shows no degrada tion when subjected to an integer coefficient constraint. Also, filter s with coefficients determined by the values 2(n) and excellent coding performance are obtained.