Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Convex-hull Estimation using XPSNR for Versatile Video Coding

Vignesh V Menon, Christian R. Helmrich, Adam Wieckowski, Benjamin Bross, Detlev Marpe

TL;DR

This paper explores using the eXtended Peak Signal-to-Noise Ratio (XPSNR) metric as an alternative to the popular Video Multimethod Assessment Fusion (VMAF) metric for determining optimized bitrate-resolution pairs in the context of Versatile Video Coding (VVC).

Abstract

As adaptive streaming becomes crucial for delivering high-quality video content across diverse network conditions, accurate metrics to assess perceptual quality are essential. This paper explores using the eXtended Peak Signal-to-Noise Ratio (XPSNR) metric as an alternative to the popular Video Multimethod Assessment Fusion (VMAF) metric for determining optimized bitrate-resolution pairs in the context of Versatile Video Coding (VVC). Our study is rooted in the observation that XPSNR shows a superior correlation with subjective quality scores for VVC-coded Ultra-High Definition (UHD) content compared to VMAF. We predict the average XPSNR of VVC-coded bitstreams using spatiotemporal complexity features of the video and the target encoding configuration and then determine the convex-hull online. On average, the proposed convex-hull using XPSNR (VEXUS) achieves an overall quality improvement of 5.84 dB PSNR and 0.62 dB XPSNR while maintaining the same bitrate, compared to the default UHD encoding using the VVenC encoder, accompanied by an encoding time reduction of 44.43% and a decoding time reduction of 65.46%. This shift towards XPSNR as a guiding metric shall enhance the effectiveness of adaptive streaming algorithms, ensuring an optimal balance between bitrate efficiency and perceptual fidelity with advanced video coding standards.

Convex-hull Estimation using XPSNR for Versatile Video Coding

TL;DR

This paper explores using the eXtended Peak Signal-to-Noise Ratio (XPSNR) metric as an alternative to the popular Video Multimethod Assessment Fusion (VMAF) metric for determining optimized bitrate-resolution pairs in the context of Versatile Video Coding (VVC).

Abstract

As adaptive streaming becomes crucial for delivering high-quality video content across diverse network conditions, accurate metrics to assess perceptual quality are essential. This paper explores using the eXtended Peak Signal-to-Noise Ratio (XPSNR) metric as an alternative to the popular Video Multimethod Assessment Fusion (VMAF) metric for determining optimized bitrate-resolution pairs in the context of Versatile Video Coding (VVC). Our study is rooted in the observation that XPSNR shows a superior correlation with subjective quality scores for VVC-coded Ultra-High Definition (UHD) content compared to VMAF. We predict the average XPSNR of VVC-coded bitstreams using spatiotemporal complexity features of the video and the target encoding configuration and then determine the convex-hull online. On average, the proposed convex-hull using XPSNR (VEXUS) achieves an overall quality improvement of 5.84 dB PSNR and 0.62 dB XPSNR while maintaining the same bitrate, compared to the default UHD encoding using the VVenC encoder, accompanied by an encoding time reduction of 44.43% and a decoding time reduction of 65.46%. This shift towards XPSNR as a guiding metric shall enhance the effectiveness of adaptive streaming algorithms, ensuring an optimal balance between bitrate efficiency and perceptual fidelity with advanced video coding standards.
Paper Structure (12 sections, 7 equations, 4 figures, 2 tables)

This paper contains 12 sections, 7 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. VEXUS architecture
  3. Spatiotemporal complexity feature extraction
  4. XPSNR-optimized resolution prediction
  5. Optimized QP prediction
  6. Encoding
  7. Experimental design
  8. Dataset preparation
  9. Model training
  10. Evaluation setup
  11. Evaluation results
  12. Conclusions

Figures (4)

  • Figure 1: Rate-distortion (RD) and rate-decoding time curves of representative sequences of Inter-4K dataset inter4k_ref encoded at 540p, 1080p, and 2160p resolutions using VVenC at faster preset, and decoded using VTM decoder.
  • Figure 2: Convex-hull estimation and encoding architecture using VEXUS.
  • Figure 3: Calculation of the groundtruth PSNR, XPSNR, and bitrate to train the prediction models. This example shows the ground truth calculation of a video encoded at 1080p with qp 30.
  • Figure 4: RD curves, encoding and decoding times of representative video sequences using Default (green line), FixedLadder (blue line), Bruteforce (black line), and VEXUS (red line) encodings.