Real-Time Neural Video Compression with Integrated Super-Resolution

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Introduction:

Neural video codecs transmit content at encoded resolution, but traditional approaches
require encoding at the target display resolution to achieve acceptable quality. This creates a
fundamental bandwidth limitation - higher resolution viewing requires proportionally higher
bitrates. Autonomous vehicle teleoperation demands high-resolution video for operator
decision-making while operating under strict bandwidth constraints.

Traditional codecs separate compression and super-resolution as independent stages, preventing joint
optimization and forcing suboptimal rate-distortion trade-offs. Recent neural compression
architectures demonstrate the flexibility to integrate reconstruction directly into the decoding
process, enabling resolution-agnostic latent representations that contain information for
multiple output resolutions.

The goal of this work is to develop a neural video codec with integrated super-resolution
decoding that achieves superior rate-distortion performance compared to encoding directly at
target resolution, demonstrating that joint end-to-end training of compression and
reconstruction enables better quality at equivalent bitrates for teleoperation scenarios.

Work Packages:
- Literature survey of neural super-resolution techniques and their integration with learned
compression
- Development & Implementation of joint training strategy optimizing end-to-end rate-
distortion across compression and super-resolution stages
- Extension of a neural codec architecture with resolution-flexible decoder supporting multiple
reconstruction modes from shared latent representation
- Comparative evaluation against baseline encoding at target resolution and post-processing
super-resolution approaches on autonomous driving video data

Recommended Literature:

1. Nonlinear Transform Coding
2. End-to-End Neural Video Compression: A Review
3. DCVC-RT
4. Learned Low Bitrate Video Compression with Space-Time Super-Resolution

If you are interested or have any questions, please send me an e-mail (niklas.krauss@tum.de) with your CV and a
current transcript of your records, thank you!

Voraussetzungen

Requirements:
- Programming experience with Python and well versed withPytorch
- High personal motivation and independent working style.
- Very good language proficiency in German or English

 

 

Verwendete Technologien

Neural Data Compression, Python, Pytorch, Machine Learning, Machine Learning, Super Resolution, Compression, Autonomous Driving, Autonomous Vehicles, Teleoperation

Tags

FTM Studienarbeit, FTM Krauss, FTM AV, FTM AV Safe Operation, FTM Informatik, FTM Teleoperation

Möglicher Beginn

sofort

Kontakt

Niklas Krauß
Raum: 3507
Tel.: +49172 1736882
niklas.krausstum.de

Ausschreibung