Improved adaptive arithmetic coding based on optimal segmentation of code symbols for lossless motion vector coding

Adaptive arithmetic coding is a general technique for coding the symbols of a stochastic process based on an adaptive model. The adaptive model provides the code symbol statistics and is updated along with encoding/decoding processes when more encoded/decoded symbols are fed as samples to the adaptive model. The coding performance depends on how well the adaptive model fits the symbol statistics. If the number of code symbols is large and the samples of code symbols are limited, the adaptive model may not be able to provide an accurate symbol statistics, which leads to the inefficient coding performance of the adaptive arithmetic coder. An example is lossless motion vector coding for video transmission when the motion searching range is very large. This paper presents an improved adaptive arithmetic coder used for lossless motion vector coding. A novel representation of motion vector differences (MVDs) is proposed, in which a MVD is divided into two segments, namely significant segment and non-significant segment. Each segment is separately coded with an adaptive arithmetic coder. With this division, the possible values of each segment are concentrated within a small range. This concentration leads to a good fit of the adaptive model to the symbol statistics and therefore to an improvement of the adaptive arithmetic coding efficiency.