Introduction

This challenge invites proposals which optimize VVC encoding processes in order to improve run-time performance of VVC software decoding process while maintaining or minimizing compression efficiency loss which may be incurred in this process.

The challenge is based on publicly available VVC encoder and decoder software: VVdeC [1] and VVenC [2] respectively. We invite proposals which range from optimized encoder configurations, to novel video encoding algorithms under the condition that all produced bitstreams are VVC-compliant and can be decoded by the provided VVdeC decoder. Other constraints as described below will apply.

Eligibility and prizes

The challenge is sponsored by the Media Coding Industry Forum (MC-IF) and is open to students at any level (graduate or undergraduate) who are enrolled at a recognized university or college. Corporate teams are not eligible for prizes. A first prize of 1500 Euro and a second prize of 800 Euro will be awarded. Payment of a monetary award must comply with U.S. law, so recipients in certain countries subject to U.S. sanctions may be ineligible.

Background

Complexity of video coding process and resulting energy consumption on battery-powered devices has become an important consideration in deployment of video codecs. It covers two aspects: deployment of new generation codecs on power-constrained mobile platforms and resulting energy efficiency.

The ability to execute software video decoding processes in a consistent way across different families of devices (including both premium and low-tier devices) may determine how a particular service is launched. In contrast to hardware based codec deployments which are primarily based on hardware capabilities, software video coding deployments depend on factors beyond the specifics of a particular implementation. For a given video codec, the coding tools exercised in a video bitstream, and the operating point (bitrate) may significantly impact performance of a software video decoder. This grand challenge aims to provide practical solutions and suggestions for the VVC encoding process such that the software decoding process is optimized while still achieving competitive video coding performance.

Evaluation criteria

Deliverables

Responses to the grand challenge are invited and sought to provide the following information (deliverables):

Description of the proposed solution:
- Rationale and methods of the proposed solution
- Theoretical frameworks developed (if any)
- Indication which of the challenges variants are addressed by the proposals
Resulting video bitstreams
Results according to evaluation framework and test sets.
Source code (patches) + configuration presets
- Organizers may choose to exercise provided solutions to test claims and align results between competing proposals.
- Provided code patches must be provided under the same license as VVenC, to enable potential integration into the VVenC software.

Participants should submit the written portion of their response to the organizers via email, then upload instructions for bitstreams etc. will be provided.

Further details of evaluation framework

Anyone interested in the challenge is encouraged to familiarize themselves with previous work in this area that may provide a good starting point for investigations such as, e.g., in [3].

Results using the VVenC ‘fast’ or ‘medium’ presets are of most interest, and the encoder configurations used in previous studies may be a useful starting point. However, this is not a strict requirement and participants have the freedom to select appropriate encoder configurations that best demonstrate their solution.

It is up to participants to determine the optimal trade-off between decoder runtime reduction and BD-rate loss, but in general responses which incur coding losses greater than 15% (BD-rate) are discouraged.

Performance metrics

The performance of a submission is to be measured against an ‘anchor’ or ‘reference’. Bitstreams and runtime results using VVenC with baseline presets form the ‘anchor’.

Proposals will be evaluated using VVdeC. The run-time performance of decoded video bitstreams submitted in response to the challenge will be compared with the decoder run-time performance of the anchor.
Encoder run-time performance will be included in evaluation and compared with encoder run-time for baseline presets.
Compression performance will be calculated as a BD rate compared with anchor bitstreams.

Test content and bitrates

Participants are free to provide results based on test content of their choice provided that such a test set can be accessed by the challenge organizers and is under a permissive license. One recommended test set relevant for the challenge can be found in [4] and is further described in [5].

Participants are encouraged to focus on the following video formats: 720p, 1080p or 2160p.

Submitted video bitstreams must conform to the VVC Main Profile. Test content used must be in YUV format with 8-bit or 10-bit sample depth and 4:2:0 chroma sampling. 10-bit encoding is encouraged; 8-bit bit depth is allowed for test content that itself is 8-bit.

Timeline

15 June – Grand Challenge opens
15 September – Grand Challenge submission deadline

Contact information

The challenge is sponsored and organized by the Media Coding Industry Forum (MC-IF). The contact persons for any questions are Lukasz Litwic (Ericsson, [email protected]) and Justin Ridge (Nokia, [email protected]).

References

“Fraunhofer versatile video decoder (VVdeC),” https:// github.com/fraunhoferhhi/vvdec
“Fraunhofer versatile video encoder (VVenC),” https:// github.com/fraunhoferhhi/vvenc
Kränzler et al., “Optimized Decoding-Energy-Aware Encoding In Practical VVC Implementations,” 2022 IEEE International Conference on Image Processing (ICIP), Bordeaux, France, 2022, pp. 1536-1540
https://dash-large-files.akamaized.net/WAVE/3GPP/5GVideo/ReferenceSequences/
3GPP, “3GPP TR: 26.955: Video codec characteristics for 5G-based services and applications,” 2024. Available at: https://www.3gpp.org/ftp/Specs/archive/26_series/26.955/26955-i00.zip