Quantization Error Minimization by Reducing Median Difference at Quantization Interval Class

Abstract

In this paper, a new technique to define the size of quantization interval is defined. In general, high quantization error will occur if large interval is used at a large difference value class whereas low quantization error will occur if a small interval is used at a large difference value class. However, the existence of too many class intervals will lead to a higher system complexity. Thus, this research is mainly about designing a quantization algorithm that can provide an efficient interval as possible to reduce the quantization error. The novelty of the proposed algorithm is to utilize the high occurrence of zero coefficient by re-allocating the non-zero coefficient in a group for quantization. From the experimental results provided, this new algorithm is able to produce a high compressed image without compromising with the image quality.