Quality-Constant Per-Shot Encoding by Two-Pass Learning-based Rate Factor Prediction
Chunlei Cai, Yi Wang, Xiaobo Li, Tianxiao Ye · 2022 IEEE International Conference on Visual Communications and Image Processing (VCIP) · 2022
Providing quality-constant streams can simultaneously guarantee user experience and prevent wasting bit-rate. In this paper, we propose a novel deep learning based two-pass encoder parameter prediction framework to decide rate factor (RF), with which encoder can output streams with constant quality. In first-pass, an RF is predicted based on spatial-temporal and pre-coding features of video segment. Then video segment is encoded using the predicted RF and then its VMAF is measured. If first pass VMAF doesn't meet target quality, a second pass prediction is performed using another model, in where results of first pass is added to features. Experiments show the proposed method requires only 1.55 times encoding complexity on average, meanwhile the accuracy, that the compressed video's actual VMAF is within ±1 around the target VMAF, reaches 98.88%. Compared with average rate mode, this method can both improve visual Quality and save ~10% bit-rate, as shown in demos11Subjective comparison videos can be downloaded at https://drive.google.com/drive/folders/1BJNZ5HssxFaKBcXMenEdioLJ77hnZ-vx?usp=sharing..