Advantages of uniform scalar dead-zone quantization in image coding system

Quantization is an element of lossy image compression systems. We introduce three basic quantization methods commonly used in the JPEG or JPEG-2000 standards: USQ (uniform scalar quantization); USDZQ (uniform scalar dead-zone quantization); UTCQ (universal trellis coded quantization) which has recently become regarded as a competitive quantization strategy. We apply these three types of quantizers to the DWT transformed coefficients of images. The performance comparisons show that UTCQ can quantize data more precisely than the two other quantizers with the same step size. However, when they are combined with zero or high order entropy coders, USDZQ is the best. We show that if the dead-zone is designed carefully, USDZQ can effectively reduce the output bits of the entropy coder, although it reduces quantization precision by discarding some data around zero. These results show that the threshold we choose for USDZQ is reasonable, and it is suitable for a transform based image compression system.