LATTICE VECTOR QUANTIZATION OF GENERALIZED GAUSSIAN SOURCES

Many important sources of data including subband image and speech coef ficients are highly peaked and can be well-modeled by the family of ge neralized Gaussian (GG) PDF's parameterized by a shape parameter nu. W e present here algorithms for defining and quantizing to a Z lattice i n which the boundary is optimized to the characteristics of GG sources , and show that these techniques lead to high performance and low comp lexity for bit rates and dimensions that are of interest in a number o f practical coding applications. We derive an analytical description o f the granular and overload distortion valid for low and moderate bit rates, and also provide a description of quantizer performance in the limit of high rate and dimension.