A Fast Transform Algorithm based on VVC
Keke Ding, Dong Liang Jiang, Feiyang Zeng, Jucai Lin, Jun Yin · 2020
Versatile Video Coding and Test Model (VTM) reference codecs, which is developed by Joint Video Experts Team of ISO/IEC and ITU-T, utilize multiple types of integer transforms based on DCT and DST of various transform sizes for intra and inter-predictive coding. It has brought a considerable gain in BDBR performance. VVC adopts three types of integer transform kernels (DCT-II, DCT-VIII and DST-VII). In addition to the fast algorithm of DCT-II, there is little research on fast algorithms of other transform kernels such as DCT-VIII, DST-VII, DCT-IV, DST-IV. In this paper, we present a fast transform algorithm based on VVC. First of all, we extract cyclic convolution matrices from the transform matrix. Then, according to the properties of cyclic convolution, we reduce the required number of multiplications and additions for transform. In various experiments, the proposed fast calculation method effectively reduces the total matrix complexity without any loss in BDBR performance. In particular, the proposed method achieves 30% average time saving in transform part compared to the original VTM 10.0.