Proposal of a fast multialphabet dynamic arithmetic coding

Abstract

Image coding is discussed in two phases; one is transformation of image signals using a coding model, and the other is entropy coding which assigns a code to a transformed sequence. A well-known example of the latter is arithmetic coding. Because arithmetic coding makes dynamic adaptation rather easy, it is used in efficient entropy coding of signals, such as image signals, whose statistical characteristics vary locally. Though adaptation is an effective measure to achieve efficient coding, it requires longer processing time and larger hardware size. The authors have proposed multialphabet arithmetic coding with a 1/2n rate approximation, whose algorithm is simple. This arithmetic coding can be performed with an extremely small amount of multiplication, and it can be extended to adaptation simply by recalculation of cumulative approximate probability. In this article, we will propose a method that optimizes update timing in adaptation, and show that we obtained satisfactory results in efficiency and coding speed. © 2000 Scripta Technica, Electron Comm Jpn Pt 1, 83(11): 66–75, 2000