Advanced Context-Modeling and Frequency Table Adjusting Methods for Motion Vector Encoding
Chun-Hung Lin, Yi-Shang Lu, Jian–Jiun Ding · 2021 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) · 2021
In this paper, an advanced entropy coding method for motion vector (MV) encoding in video compression is proposed. By observing the features of motion vectors and utilizing the correlation among frames, we proposed five techniques to improve the efficiency of MV encoding, including alternative raster scanning, context modeling according to the probability of zeros of adjacent pixels and frames, varied reference distances for different frames, the initial frequency table, and symmetric frequency table updating. Experiments show that, compared to the frequently used exp-Golomb coding, the proposed method can achieve 25% of bitrate reduction and is very effective for MV encoding.