Optimization of orthogonal wavelets for image compression

A technique is developed where a wavelet is matched to a 1D or 2D signal, with the goal of increasing coding gain. This is done by maximizing the projection of the signal onto the scaling subspace. For 1D signals, the results show that the scaling subspace (low pass) signal closely matches (subjectively and objectively) the original signal. For 2D signals, the optimized wavelets give superior coding results when compared with the Daubechies wavelets. For low compression ratios the PSNR gain is over 1.5 dB. This gain drops off at higher compression ratios, possibly due to the lower number of vanishing moments in the optimized wavelets.