Enhanced Neural Video Representation Compression Across Extreme Complexity and Quality Scales

Abstract

Implicit neural representations (INRs) have recently emerged as a promising approach to video compression, delivering competitive rate-distortion performance alongside rapid decoding. However, existing neural video codecs struggle to balance complexity and scalability. Lightweight models often suffer from degraded compression performance when scaled to different bitrate/quality levels, whereas high-performance models exhibit limited scalability, as their model complexity typically increases with quality. This lack of a unified architecture capable of maintaining consistent complexity across a wide range of bitrates severely limits their diverse real-world deployment.

To address these challenges, we introduce NVRC++, a novel INR-based video codec that utilizes a lightweight INR with multiple high-resolution feature grids, providing high scalability at any given complexity level. This is paired with an optimization framework that enables efficient overfitting on high-resolution grids for long video sequences, thereby exploiting spatio-temporal redundancies without prohibitive computational or memory overhead. Additionally, an advanced entropy model is designed for efficiently compressing the high-dimensional grid parameters. As a result, NVRC++ provides four complexity levels (from 7 kMACs/pixel to 360 kMACs/pixel), each spanning wide bitrate and quality ranges while supporting real-time decoding. The experimental results show that NVRC++ offers a much faster decoding speed (up to 8.5x) compared to the SOTA INR-based video codec, NVRC, while delivering comparable performance.

Complexity-Performance Trade-off

BD-rate (vs. x265 veryslow, UVG, PSNR) against (a) synthesis-only decoding speed, (b) full decoding speed, and (c) computational complexity. NVRC++ (S1-S4) offers a superior trade-off at every complexity level from 7 to 360 kMACs/pixel, with 34-365 FPS 1080p decoding on a single consumer GPU.

Results

UVG dataset

MCL-JCV dataset

Qualitative Results

At similar or lower bitrates, NVRC++ recovers finer detail than NVRC and DCVC-RT.

Citation

@inproceedings{kwan2026enhanced,
    title={Enhanced Neural Video Representation Compression Across Extreme Complexity and Quality Scales}, 
    author={Ho Man Kwan and Tianhao Peng and Fan Zhang and Mike Nilsson and Andrew Gower and David Bull},
    booktitle={European Conference on Computer Vision (ECCV)},
    year={2026}
}