Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Benchmarking Conventional and Learned Video Codecs with a Low-Delay Configuration

Siyue Teng, Yuxuan Jiang, Ge Gao, Fan Zhang, Thomas Davis, Zoe Liu, David Bull

TL;DR

This study benchmarks six codecs—VVC VTM, JVET ECM, AV1 libaom, AVM, and neural DCVC-DC/DCVC-FM—under a strictly low-delay configuration using MPEG JVET/AOM test conditions in YCbCr 4:2:0. It leverages BD-rate based on PSNR and VMAF to compare performance while also quantifying encoding/decoding complexity on a standard hardware platform. Results indicate that ECM-LDB provides the strongest PSNR BD-rate gains (around 18% over VTM-LDB, 18.8% over DCVC-DC, and 16% over AVM), with neural codecs approaching ECM in VMAF but requiring higher computational resources. Across resolutions, higher-quality gains emerge for some codecs, yet learned methods exhibit content-dependent variability, particularly on sequences with large background motion, underscoring the need for low-latency, low-complexity improvements for real-time streaming scenarios.

Abstract

Recent advances in video compression have seen significant coding performance improvements with the development of new standards and learning-based video codecs. However, most of these works focus on application scenarios that allow a certain amount of system delay (e.g., Random Access mode in MPEG codecs), which is not always acceptable for live delivery. This paper conducts a comparative study of state-of-the-art conventional and learned video coding methods based on a low delay configuration. Specifically, this study includes two MPEG standard codecs (H.266/VVC VTM and JVET ECM), two AOM codecs (AV1 libaom and AVM), and two recent neural video coding models (DCVC-DC and DCVC-FM). To allow a fair and meaningful comparison, the evaluation was performed on test sequences defined in the AOM and MPEG common test conditions in the YCbCr 4:2:0 color space. The evaluation results show that the JVET ECM codecs offer the best overall coding performance among all codecs tested, with a 16.1% (based on PSNR) average BD-rate saving over AOM AVM, and 11.0% over DCVC-FM. We also observed inconsistent performance with the learned video codecs, DCVC-DC and DCVC-FM, for test content with large background motions.

Benchmarking Conventional and Learned Video Codecs with a Low-Delay Configuration

TL;DR

This study benchmarks six codecs—VVC VTM, JVET ECM, AV1 libaom, AVM, and neural DCVC-DC/DCVC-FM—under a strictly low-delay configuration using MPEG JVET/AOM test conditions in YCbCr 4:2:0. It leverages BD-rate based on PSNR and VMAF to compare performance while also quantifying encoding/decoding complexity on a standard hardware platform. Results indicate that ECM-LDB provides the strongest PSNR BD-rate gains (around 18% over VTM-LDB, 18.8% over DCVC-DC, and 16% over AVM), with neural codecs approaching ECM in VMAF but requiring higher computational resources. Across resolutions, higher-quality gains emerge for some codecs, yet learned methods exhibit content-dependent variability, particularly on sequences with large background motion, underscoring the need for low-latency, low-complexity improvements for real-time streaming scenarios.

Abstract

Recent advances in video compression have seen significant coding performance improvements with the development of new standards and learning-based video codecs. However, most of these works focus on application scenarios that allow a certain amount of system delay (e.g., Random Access mode in MPEG codecs), which is not always acceptable for live delivery. This paper conducts a comparative study of state-of-the-art conventional and learned video coding methods based on a low delay configuration. Specifically, this study includes two MPEG standard codecs (H.266/VVC VTM and JVET ECM), two AOM codecs (AV1 libaom and AVM), and two recent neural video coding models (DCVC-DC and DCVC-FM). To allow a fair and meaningful comparison, the evaluation was performed on test sequences defined in the AOM and MPEG common test conditions in the YCbCr 4:2:0 color space. The evaluation results show that the JVET ECM codecs offer the best overall coding performance among all codecs tested, with a 16.1% (based on PSNR) average BD-rate saving over AOM AVM, and 11.0% over DCVC-FM. We also observed inconsistent performance with the learned video codecs, DCVC-DC and DCVC-FM, for test content with large background motions.
Paper Structure (10 sections, 4 figures, 1 table)

This paper contains 10 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Experimental setup
  3. Codec Selection and Configuration
  4. Test Sequences
  5. Evaluation Metrics
  6. Results and Discussion
  7. Overall Coding Performance
  8. Per-resolution and Per-sequence Performance
  9. Coding Complexity
  10. Conclusion

Figures (4)

  • Figure 1: Radar chart plot for six tested codecs in terms of BD-rate (PSNR), BD-rate (VMAF), relative encoding complexity and decoding complexity (all against libaom). Larger shaded area indicates better performance-complexity trade off for the corresponding codec.
  • Figure 2: Rate-ristortion curve of HD sequences.
  • Figure 3: Per sequence BD-rate results. Here libaom is used as the anchor.
  • Figure 4: Example frames of the sequences on which two neural video codecs did not perform well.