Optimal previsualized image vector quantization
Thomas G. Stockham, Zhong Xie · International Conference on Acoustics, Speech, and Signal Processing · 2003
The authors present a novel technique for image vector quantization (VQ) at a bit rate of 0.75 bits/pixel or below with moderately low to very low subjective distortion. The encoding scheme incorporates the multiplicative visual model as part of the distortion measure. By modeling the quantization noise as an additive signal-dependent noise process, an optimal pre- and postprocessing system, which minimizes the mean-squared error measured inside the visual model, is derived. The design algorithm for this optimal system is also developed. Using this optimal system, a set of experiments was conducted. Images of surprising quality have been produced at the bit rate of 0.1 bits/pixel with a compression rate of 80:1 relative to a normal 8 bits/pixel original.>