Quantization Effect of Symmetric Wavelet Filter Banks for Image Compression

S.-w. Park (USA)

Keywords

image compression, subband coding, wavelet transform, QMF

Abstract

In this paper, one orthogonal and three symmetric wavelet filters have been applied to images and resulting wavelet coefficients were quantized for testing the quantization effect of the different filter banks. Symmetric wavelet filters have advantage over orthogonal wavelet filters: it does not result in coefficient expansion when symmetric extension is used. Symmetric extension is preferred to periodic extension, as the former does not introduce discontinuity that makes highpass filter output coefficients unnecessarily large. In the JPEG2000 image compression standard, symmetric biorthogonal wavelet filters are widely used. It is empirically shown in this paper that a near-orthogonal symmetric QMF filter performed as good as or little better than the popular biorthogonal wavelet filter in terms of SNR after quantization. One notable advantage is that the QMF filter shares the same coefficients for both the decomposition and reconstruction filters and for both the lowpass and the highpass filters except for signs.

Important Links:



Go Back