Modified Golomb coding algorithm for asymmetric two-sided geometric distribution data

Golomb coding is useful for encoding geometrically distributed data and plays an important role in advanced compression techniques, such as JPEG-LS and H.264. In this paper, a new Golomb coding method for asymmetric two-sided geometrically distributed data is proposed. An asymmetrical model means that the value of x can be positive or negative, but the ratio P(x = -n)/P(x = -n-1) is unequal to P(x = n)/P(x = n +1). The asymmetric model is more suitable for modeling the practical case because many data have higher probability to be positive (or negative) than to be negative (or positive) in nature. Two simulation examples are given: One is to encode the object boundaries for binary image compression and the other one is to encode the DC differences in JPEG. Both simulation results show that the proposed asymmetric two-sided Golomb coding algorithm outperforms other methods and has higher ability for data compression.