Multidimensional rotations for quantization

Laplacian and generalized Gaussian data arise in image and speech coding. Simple rotations of independent, identically distributed Laplacian and generalized Gaussian data in multiple dimensions can improve the granular and overload characteristics for quantization. In this paper, we describe the practical properties of multidimensional rotations for both scalar and lattice quantization and then apply them to image compression over noisy channels.<
>