Quality-Constant Per-Shot Encoding by Two-Pass Learning-based Rate Factor Prediction

TL;DR

This paper introduces a two-pass deep learning framework for per-shot video encoding that predicts rate factors to maintain constant quality, significantly improving accuracy with minimal additional complexity.

Contribution

The proposed method is the first to use a two-pass neural network approach for per-shot rate factor prediction, achieving high accuracy and efficiency in quality-constant video streaming.

Findings

Achieves 98.88% accuracy in maintaining target VMAF within ±1.

Requires only 1.55 times the encoding complexity of single-pass methods.

Effective in real-time scenarios with minimal additional computational cost.

Abstract

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. For each one-shot segment in a video, the proposed method firstly extracts spatial, temporal and pre-coding features by an ultra fast pre-process. Based on these features, a RF parameter is predicted by a deep neural network. Video encoder uses the RF to compress segment as the first encoding pass. Then VMAF quality of the first pass encoding is measured. If the quality doesn't meet target, a second pass RF prediction and encoding will be performed. With the help of first pass predicted RF and corresponding actual quality as feedback, the second pass prediction will be highly accurate. 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 $\pm1$ around the target VMAF, reaches 98.88%.