Lattice vector quantization of image wavelet coefficient vectors using a simplified form of entropy coding

Lattice vector quantization has recently attracted some interest as an alternative to full-search VQ for signal and image coding problems. It is considerably more computationally efficient, and it avoids the difficult codebook design problem. Furthermore, it has been noted that the optimal high bit rate entropy constrained vector quantizer will approximate a lattice. Indeed for lattice VQ to be competitive, the quantized vectors should be entropy coded. This has traditionally been performed on a per image basis, a complex and inefficient process. We propose a general image-independent coding scheme which we apply to the coding of lattice quantized wavelet coefficient vectors. We also demonstrate that the quantizing and coding complexity can be reduced through reducing the dimension of the wavelet vectors by means of principal components analysis and a perceptually based nonlinearity