9/7 discrete wavelet transform with binary coefficients in JPEG2000

Based on the binary representation for real number, the lifting coefficients of the 9/7 biorthogonal wavelet filter pair of CDF (Cohen, Daubechies and Feauveau) are expressed in binary representation. A simple shift-add operation is used to substitute for the multiplier with complicated architecture operating on floating-point, so that the 9/7 discrete wavelet transform in JPEG2000 is implemented through fixed-point computing. Compared with the floating-point computing method, the shift-add operation is simpler, and can be implemented easily by VLSI, which enables the real-time image processing to be realized by hardware. The simulation results show that, in the case of low compression ratio, the PSNR (peak signal-to-noise ratio) of the reconstructed image by the shift-add operation is only 0.1 dB lower than the floating-point computing method. When the compression ratio is high, our method is a little superior to the latter.