The Genetic Algorithm Use in the Searching of the Best Fitting Wavelet Visual Weighting Function

The discrete wavelet transform (DWT) is suitable for still image compression, and by using visually tuned quantization the quality of such images can be improved. We use measured results of the human vision system (HVS) sensitivity to the wavelet quantization noise to establish wavelet visual weighting function W_v(u,v). By using a genetic algorithm we estimate parameters of the function Wv(u,v) to give the best approximation in the L_1 norm of the measured values.