Design of Quantizers for Real-Time Hadamard-Transform Coding of Pictures

A methodology is developed in this paper to establish fidelity criteria that characterize human observers' perception of noisy transform-coded pictures and to obtain optimum quantizers for the transform coefficients based on these fidelity criteria. The perceptual effects of impairments introduced in the transform domain are, in general, quite different from the impairments introduced in the picture domain. Our experiments, which are performed with the Hadamard transform of a stationary picture, determine the visibility of impairments in the reconstructed picture when noise (//-noise), which simulates the quantization noise, is added to a Hadamard coefficient. Functions that give the appropriate 21 subjective weighting of the quantization noise as a function of the quantity to be quantized are derived from these experiments. These functions, called visibility functions, are used in a systematic way to design quantizers for PCM and DPCM coding of the transform coefficients. These quantizers are compared with minimum mean-square error (MMSE) quantizers, both in terms of picture quality and bit rates. 1.1 Relationship to previous work