Hyperspectral Imaging
Overview
Hyperspectral imaging measures each pixel at many narrow, often contiguous wavelengths, helping distinguish materials and estimate properties that RGB images can miss. Research spans sensor design, reconstruction from compressed or sparse measurements, restoration and super-resolution, and tasks such as classification, unmixing, anomaly detection, tracking, and atmospheric or material retrieval. Methods range from physical and mixing models to optimization-based unfolding, CNNs, Transformers, state-space models, diffusion, self-supervised learning, and foundation models that combine spectral signatures with spatial, temporal, or 3D context. A central challenge is preserving spectral fidelity while improving resolution and working with noisy or incomplete data, few labels, and changing sensors or scenes, making robust benchmarks and real-world validation important.
Latest papers
139 papers10 papers in the last four weeks, down 17% on the four weeks before. 0.1% of all new papers.