Zerotree wavelet image coding based on the human visual system model

In this paper, we consider an application of perceptual optimization to the zerotree wavelet image coding. Three main properties of the Human Visual System (HVS) are incorporated into the coder: frequency sensitivity, luminance sensitivity and texture masking. Optimization is implemented in two phases. In the first phase, all wavelet coefficients are weighted by the base frequency sensitivities of the corresponding subbands. In the second stage, local adjustment of the quantizer is achieved by the comparison of a just noticeable distortion (JND) threshold with the current quantizer step size. Values of the JND threshold are calculated at every pixel location using luminance sensitivity and texture masking factors. Only quantized pixel values are used to avoid overhead information and preserve the iterative nature of the algorithm. Our proposed coder has been tested on a set of images. Simulation results show that the superior subjective image quality can be achieved at the same bitrate, compared to the PSNR-optimized encoder.