Name
Abstract | ||
|---|---|---|
In industrial and practical application, the robust rate control method is highly required in field of video coding. In this paper, a Long-Short Term Memory (LSTM)-based method is proposed to optimize x264 rate control. The bitrate and distortion information are used to construct the LSTM recurrent neural network to predict a quantization parameter (QP) at inter frame level. Then a QP refinement strategy is utilized to further improve the relationship between the QP and bitrate. Finally, the average bitrate control (ABR) method is optimized to provide more accurate bitrate while encoding. Experimental results show that the proposed method achieves 1.2% BD-rate reduction in HM test sequence, and 1.0% in surveillance video. Meanwhile, the target bit matching rate is up to 98.9% and 99.7%, respectively.
|
Year | DOI | Venue |
|---|---|---|
2018 | 10.1145/3301506.3301524 | Proceedings of the 2018 the 2nd International Conference on Video and Image Processing |
Keywords | Field | DocType |
Long-Short Term Memory Neural Network (LSTM), rate control, video coding, x264 | Average bitrate,Computer science,Recurrent neural network,Algorithm,Coding (social sciences),Inter frame,Quantization (signal processing),Artificial neural network,Distortion,Encoding (memory) | Conference |
ISBN | Citations | PageRank |
978-1-4503-6613-7 | 0 | 0.34 |
References | Authors | |
0 | 5 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Zheng-Teng Zhang | 1 | 0 | 0.34 |
| Jucai Lin | 2 | 0 | 0.34 |
| Ruidong Fang | 3 | 0 | 0.34 |
| Juan Lu | 4 | 1 | 2.04 |
| yao chen | 5 | 24 | 9.82 |