Session Abstract

High-dimensional visual data—such as event streams, light fields, hyperspectral and polarization images, depth sensing, and neural scene representations—are reshaping the landscape of multimedia research by providing richer spatial, temporal, spectral, and geometric information than conventional RGB videos. These modalities offer significant robustness in challenging conditions including fast motion, low light, adverse weather, and complex dynamics, yet they also introduce substantial challenges in sensing, reconstruction, alignment, compression, perception, and downstream reasoning. This special session brings together advances across the full pipeline of high-dimensional visual data, spanning novel sensing and computational imaging, efficient reconstruction and representation learning, multimodal scene understanding, and trustworthy reasoning in real-world systems. We welcome contributions that integrate physical priors with learning-based models, propose efficient or deployable system designs, or provide new datasets and benchmarks that deepen our understanding of how high-dimensional visual information can enhance multimedia analysis and applications.

Call for Papers

The Special Session on Advances in Imaging, Perception, and Reasoning for High-Dimensional Visual Data invites original research papers along the full pipeline from sensing and reconstruction to cross-modal perception and downstream reasoning. We welcome works on new imaging modalities, computational imaging, high-dimensional scene representations, multimodal understanding, and efficient or trustworthy reasoning systems. Accepted papers will be included in ICME 2026 and presented in the special session. Researchers from multimedia, computer vision, computational imaging, remote sensing, robotics and related communities are encouraged to submit.

Topics of interest include (but are not limited to):

• Novel imaging sensors: event, light field, hyperspectral, polarization, depth, holographic
• RGB+X fusion: RGB–LiDAR, RGB–thermal, RGB–radar, audio–visual and related settings
• Computational imaging: reconstruction, restoration and enhancement for high-dimensional data
• Neural scene representations: NeRF, Gaussian splatting and related 3D/4D modeling
• High-dimensional scene understanding: depth, flow, semantics and 3D/4D perception
• Cross-modal registration, calibration and alignment for multi-sensor systems
• Quality assessment and uncertainty estimation for high-dimensional or multimodal data
• Foundation and vision–language models for high-dimensional or multimodal visual inputs
• Efficient and edge-friendly models: compression, model–codec co-design and streaming
• Multimodal reasoning and decision making with high-dimensional visual inputs
• Benchmarks, datasets and simulators for imaging–perception–reasoning pipelines
• Applications in transportation, AR/VR/MR, robotics, remote sensing and smart cities

programme

TBD

• Papers must be no longer than 6 pages, including all text, figures, and references.
• The abstract should be about 100-150 words.
• You can obtain more detailed information through this link.