Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Subjective Evaluation of Frame Rate in Bitrate-Constrained Live Streaming

Jiaqi He, Zhengfang Duanmu, Kede Ma

TL;DR

The paper tackles how frame rate and bitrate trade off perceptual quality in bitrate-constrained live streaming. It introduces the HFR-LS dataset, comprising 384 1080p clips derived from 32 120fps sources encoded at four bitrates and three frame rates, and it employs a single-stimulus hidden-reference subjective study to obtain DMOS ratings. Key findings reveal that frame rate significantly affects quality and interacts with bitrate and content, with high frame rates potentially producing more artifacts when bitrate is insufficient; VQA models struggle to predict quality across varying frame rates, though MinimalisticVQA shows the best performance among those tested. The work highlights the need for bitrate-aware encoding strategies and more nuanced VQA predictors that capture frame-rate, content, and bitrate interactions to improve live streaming quality assessment and optimization.

Abstract

Bandwidth constraints in live streaming require video codecs to balance compression strength and frame rate, yet the perceptual consequences of this trade-off remain underexplored. We present the high frame rate live streaming (HFR-LS) dataset, comprising 384 subject-rated 1080p videos encoded at multiple target bitrates by systematically varying compression strength and frame rate. A single-stimulus, hidden-reference subjective study shows that frame rate has a noticeable effect on perceived quality, and interacts with both bitrate and source content. The HFR-LS dataset is available at https://github.com/real-hjq/HFR-LS to facilitate research on bitrate-constrained live streaming.

Subjective Evaluation of Frame Rate in Bitrate-Constrained Live Streaming

TL;DR

The paper tackles how frame rate and bitrate trade off perceptual quality in bitrate-constrained live streaming. It introduces the HFR-LS dataset, comprising 384 1080p clips derived from 32 120fps sources encoded at four bitrates and three frame rates, and it employs a single-stimulus hidden-reference subjective study to obtain DMOS ratings. Key findings reveal that frame rate significantly affects quality and interacts with bitrate and content, with high frame rates potentially producing more artifacts when bitrate is insufficient; VQA models struggle to predict quality across varying frame rates, though MinimalisticVQA shows the best performance among those tested. The work highlights the need for bitrate-aware encoding strategies and more nuanced VQA predictors that capture frame-rate, content, and bitrate interactions to improve live streaming quality assessment and optimization.

Abstract

Bandwidth constraints in live streaming require video codecs to balance compression strength and frame rate, yet the perceptual consequences of this trade-off remain underexplored. We present the high frame rate live streaming (HFR-LS) dataset, comprising 384 subject-rated 1080p videos encoded at multiple target bitrates by systematically varying compression strength and frame rate. A single-stimulus, hidden-reference subjective study shows that frame rate has a noticeable effect on perceived quality, and interacts with both bitrate and source content. The HFR-LS dataset is available at https://github.com/real-hjq/HFR-LS to facilitate research on bitrate-constrained live streaming.
Paper Structure (10 sections, 1 equation, 4 figures, 2 tables)

This paper contains 10 sections, 1 equation, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Subjective User Study
  3. Dataset Construction
  4. Subjective Study
  5. Subjective Data Analysis
  6. Qualitative Analysis
  7. Statistical Significant Tests
  8. VQA Model Evaluation
  9. Conclusion
  10. Acknowledgments

Figures (4)

  • Figure 1: Sample frames from the proposed HFR-LS dataset, among which (a)-(l) were captured with camera motion.
  • Figure 2: Scatter plot of SI vs. TI for HFR-LS.
  • Figure 3: (a) Histogram of DMOSs in $20$ equally spaced bins. (b) Scatter plot of DMOSs between two equal, disjoint groups of subjects.
  • Figure 4: Average DMOS as a function of bitrate for different frame rates and source content: (a) $12$ source videos with camera motion ($\text{TI} > 6$), (b) $9$ source videos with significant temporal variations ($\text{TI} > 6$) and no camera motion, and (c) $11$ source videos with slight temporal variations ($\text{TI} < 6$) and no camera motion. Error bars indicate the standard errors.