VISIBILITY OF WAVELET QUANTIZATION NOISE

The discrete wavelet transform (DWT) decomposes an image into bands th at vary in spatial frequency and orientation., it is widely used for i mage compression, Measures of the visibility of DWT quantization error s are required to achieve optimal compression, Uniform quantization of a single band of coefficients results in an artifact that we call DWT uniform quantization noise; it is the sum of a lattice of random ampl itude basis functions of the corresponding DWT synthesis filter. We me asured visual detection thresholds for samples of DWT uniform quantiza tion noise in Y, Cb, and Cr color channels, The spatial frequency of a wavelet is r2(-lambda), where r is display visual resolution in pixel s/degree, and lambda is the wavelet level, Thresholds increase rapidly with wavelet spatial frequency. Thresholds also increase from Y to Cr to Ch, and with orientation from lowpass to horizontal/vertical to di agonal, We construct a mathematical model for DWT noise detection thre sholds that is a function of level, orientation, and display visual re solution, This allows calculation of a ''perceptually lossless'' quant ization matrix for which ail errors are in theory below the visual thr eshold, The model may also be used as the basis for adaptive quantizat ion schemes.