2-dimensional Scalar Quantization with Performance Close to Vector Quantization

The principle of subband coding of images has been studied intensively for years. In our work on subband coding of images we have focused on a particular IIR lter bank structure that produces complex valued subband signals. In previously reported coders, data compression is achieved as a result of scalar quantization or vector quantization (VQ) of the encoder sub-band signals 1. In this paper we present new results pertaining to the quantization of the complex valued subband signals employing a new two dimensional (2-D) scalar quantizer. In the new quan-tizer we assume the signal to be 2-D vectors represented on polar form. The new quantization scheme can be applied to any signal provided the signal samples are statistically independent and grouped in 2-D vectors. The quantizer performs close to that of VQ in terms of SNR, where as it is characterized by the low computional complexity of scalar quantizers.