MSU Video Codecs Comparisons 2021:
Call for Codecs

You can apply for participation in comparison of 2023-2024 ->>

#1 codecs comparisons in the world

Video group head:	Dr. Dmitriy Vatolin
Project head:	Dr. Dmitriy Kulikov
Measurements, analysis:	Dr. Mikhail Erofeev, Anastasia Antsiferova, Egor Sklyarov, Alexander Yakovenko, Nickolay Safonov, Alexander Gushin Nikita Alutis

compression.ru

Lomonosov
Moscow State University (MSU)
Graphics and Media Lab

Dubna International
State University

Institute for Information
Transmission Problems,
Russian Academy of Science

Important Dates

You can apply for participation using our form:

Or if you want you can contact us at

Navigation

About

2021 Comparison parts

New features
Comparison Rules and Methodology
Participate
Feedback
Thanks
Contact

About Annual MSU Video Codecs Comparisons

We perform comparative unbiased analysis for

software implementations
hardware (GPU-based) implementations
cloud-based implementations

of different video coding standards (H.265/HEVC, H.264/MPEG-4 AVC, AV1, VP9, VP8, MPEG-4 and other) using objective metrics (SSIM, VMAF, PSNR and other).

With MSU Codecs Comparison developers can verify the performance of their codec. We share test sequences, encoding parameters and codecs versions so all developers can reproduce the results of the comparison. Participation with publishing of all results is for FREE. Private participation is for fee (contact us for the details).

MSU codecs analysis team has been conducting video codec analysis, testing and optimization since 2004. Some facts about previous MSU Video Codecs Comparisons:

There were more than 400.000 downloads of previous H.264/MPEG-4 AVC and H.265/HEVC video codec comparisons reports
Many codecs' bugs were found and reported to developers
210+ encoders were tested
More than 35 private reports for codec developers (description of codec's weak and strong points) after public report versions

Summary report topics:

Objective measurements
Subjective analysis
Encoding time
Bitrate keeping for evaluating rate-control mechanism
Speed/Quality trade-off analysis
Averaged objective results analysis
Leaders in different use cases

2021 Comparison parts and main points

FullHD Comparison

4K Comparison

Subjective Comparison

Ultra-fast (FPGA) Comparison

Cloud Comparison

Use Cases

Fast - 1080p@30fps
Slow - 1080p@1fps

Fast - UHD@30fps
Slow - UHD@1fps

Fast - 1080p@30fps
Slow - 1080p@1fps

Ultra-fast - 1080p@60fps

No speed limits, offline encoding scenario, several resolutions

Videos

60 FullHD videos

10 4K videos

15 out of 60 FullHD videos

60 FullHD videos

15 out of 60 FullHD videos

Objective metrics

Objective metrics: SSIM, PSNR, VMAF
3 color-planes - Y,U,V with different proportions: 4:11, 6:1:1, 8:11, 10:11
Powered by VQMT

Subjective metrics

Yes, powered by: Subjectify.us

Target Bitrate

1-12 Mbps

2-20 Mbps

1-12 Mbps

Hardware

CPU: Intel Core i7 8700K (Coffee Lake)
Mainboard: ASRock Z370M Pro4
RAM: 2x16GB DDR4 2400MHz
OS: Windows 10 x64 and Linux (Ubuntu)

FPGA Xilinx Alveo U250
GPU: NVIDIA Titan RTX

No hardware required

Download requirements

Will be released soon

New features in MSU Video Codecs Comparison

Since 2020	Automatic codecs submission system Open to compare Linux-based solutions Open to compare cloud-based solutions Enlarging of video collection
Since 2019	Increasing the number of test videos up to 100+, adding UGC (user-generated) videos (vlogs, noise videos, etc.) and HDR videos Open to compare cloud-based solutions Enlarging of video collection Going to use new no-reference metrics
Since 2018	Increased target bitrates for comparison on 4K videos Started to use VMAF objective metric Upgraded testing hardware to Coffee Lake Enlarged video collection (15833 videos)

Comparison Rules and Methodology

Please pay attention that we will use multi-core CPU for encoding, so you can use multi-threading

Decoding is performed with reference decoder (H.264/AVC, H.265/HEVC, VP9, AV1). For other standards participants should provide decoder
We don't limit GOP size and intra-period
We don't limit number of passes in the encoding. Total encoding time should fulfill use case speed requirements

Full Rules
Applying for participating, you agree with comparison rules.

Codec Requirements

Presets for different speed requirements should be provided by the developers
Codec should allow to set arbitrary bitrate of resulting stream in constant quality mode
Preferable codec interface - console codec version (with batch processing support вЂ” bitrate and file names must be possible to assign from the command line)
Encoder should be compatible with reference decoder

Developers Deliverables The following deliverables should be provided by each developer:

Codec files (CLI executable file is preferable)
Codec's presets

Full comparison methodology

FAQ
If you have any more questions about rules or methodology please refer to FAQ or contact us at

Take part in 2021 Video Codecs Comparison!

Deepest codecs review: 5 reports, including subjective with 1000+ viewers and 7000+ charts

General process of your codec testing:

We send you instructions for choosing presets for this year use cases (there are limitations for speed) and login for submission system.
You send us encoder and presets (our hardware characteristics can be found on the comparison page). Testing methodology is described in presentation.
We test the speed of your encoder. We let you know if some speedup is needed or any bugs are found. Finally, when all requirements are satisfied, you confirm the usage of presets and encoder.
We perform the main measurements on this year test video set and prepare a draft of the report. This will take about 1-3 months depending on the number of participants.
We send you a draft report with your results and the results of other public participants. You can double-check the results. If you participate privately, in this step you will need to decide whether you will publish your results.
After your confirmation, we prepare and publish the final report on the comparison page. We send you the Enterprise version of the report (with all sequences, charts, etc.).