Combined Prediction Mode for AVS

Zhiyi Zhu,Xin Jin
DOI: https://doi.org/10.1007/978-3-540-89796-5_3
2008-01-01
Abstract:Audio Video Coding Standard (AVS) is Chinese new audio and videocoding standard, which provides close performance to H.264/AVC mainprofile with lower complexity. Like H.264/AVC, prediction plays animportant role in AVS. A macroblock (MB) uses either Intra or Interprediction to exploit spatial and temporal correlation to achievegood compression efficiency. In this paper, a new coding techniquebased on combined prediction is proposed for AVS. Based on thedistribution analysis of residuals generated by Intra predictionand Inter prediction, the characteristics in both spatial andtemporal domain are utilized to combine two selected modesIntra_8×8_DC and Inter_P_16×16 together. The limitedcomplexity increase and the standard compatibility are discussedfor the proposed combined prediction mode (CPM). Experimentalresults show that by applying the proposed CPM to the AVSrate-distortion optimized encoding flow, up to 2.70% bit ratereduction for CIF sequences and 1.57% for 720p sequences areachieved, because it generates the more precise prediction and hasminimal complexity increase both in the encoder and decoder.
What problem does this paper attempt to address?