Skip to content
Technical University of Munich

Visual Signal Processing and Communication

Visual signal processing and communication research at LMT addresses how visual information can be represented, compressed, transmitted, reconstructed, and enhanced efficiently while preserving the fidelity for human and machine consumption. This field is increasingly important as images and videos dominate digital communication, from video conferencing and streaming platforms to immersive media, remote operation, autonomous systems, and machine-to-machine communication.  

Key Topics
  • Learned and Hybrid Image and Video Compression
  • Low-latency and Energy-efficient Visual Communication Systems
  • Visual Representation Learning and Semantic Understanding
  • Generative Visual Communication
  • Visual Enhancement, Restoration, and Inverse Problems
  • Perceptual Quality Assessment and Metrics 
Our Work

Our research investigates fundamental and practical questions in visual signal processing, including how visual information can be represented efficiently, which aspects should be preserved for human perception or machine analysis, and how to balance efficiency, quality, robustness, and complexity. We study these challenges at the intersection of signal processing, information theory, compression, machine learning, and visual perception. 

One of the directions is learned and hybrid image/video compression. We study neural coding tools for efficient representation, context modeling, and entropy modeling, and investigate how learned components can complement conventional codec structures. For instance, this includes work on screen content compression, implicit neural representations, and hybrid coding approaches that combine classical and deep learning-based methods. 

In addition, we focus on standards-based video coding optimization for geometry-aware visual data, particularly gaming sequences. Our work investigates efficient encoder design and the integration of depth and camera information within conventional coding frameworks. 

Another research direction is complexity- and energy-aware visual processing, where we study how algorithmic complexity translates into runtime, power, and energy consumption across hardware platforms. These insights guide the design of energy-efficient video compression algorithms for edge devices and heterogeneous systems.   

Beyond compression, we study visual representation learning and semantic understanding. Here, the goal is to identify and encode information most relevant to perception or downstream tasks. This research direction provides a foundation for semantic and generative visual communication, enabling ultra-low-bitrate transmission and the reconstruction of visual content through compact, high-level representations.

Key Publications
  1. Eteke, Cem; Griessel, Alexander; Kellerer, Wolfgang; Steinbach, Eckehard: BIR-Adapter: A parameter-efficient diffusion adapter for blind image restoration. Pattern Recognition, 2026
  2. Dogaroglu, Hasan Burak; Wiedemann, Cari Emanuele; Steinbach, Eckehard: Multiresolution Contexts for Implicit Neural Codecs. 2025 Picture Coding Symposium (PCS 2025), 2025
  3. Eteke, Cem; Griessel, Alexander; Kellerer, Wolfgang; Steinbach, Eckehard: Lossy Coding for Spatially Adaptive Conditioning in Semantic Image Communication. 2024 IEEE Visual Communications and Image Processing (VCIP2024), 2024
  4. Koyuncu, A. Burakhan; Jia, Panqi; Boev, Atanas; Alshina, Elena; Steinbach, Eckehard: Efficient Contextformer: Spatio-Channel Window Attention for Fast Context Modeling in Learned Image Compression. IEEE Transactions on Circuits and Systems for Video Technology 34(3), 2024, 7498-7511
Contact

If you are interested in Visual Signal Processing and Communication or would like to learn more about our work, feel free to reach out to Burak Dogaroglu, M.Sc. or Serdar Caglar, M.Sc. You are also welcome to get in touch directly with the researchers working in this field:

To top