Geometry-driven spectral encoding by maximizing distinguishability in encoding space
Chinese Society for Optical EngineeringPeer-Reviewed Publication
Spectral imaging allows materials to be identified without contact, but compact systems often struggle to preserve enough information for accurate reconstruction. Researchers from Zhejiang Normal University, Zhejiang University, and Westlake University have introduced a geometry-driven spectral encoding strategy based on the Euclidean Distance of Encoding space (EDE). Instead of optimizing only the statistical properties of an encoding matrix, EDE directly maximizes the separation between encoded spectral responses in measurement space. The team combined genetic algorithms, automatic differentiation, and transfer-matrix modeling to design manufacturable TiO₂/SiO₂ multilayer thin-film encoders. The resulting system improves reconstruction performance over random, correlation-optimized, and coherence-optimized benchmarks, and enables high-fidelity spectral recovery across the visible to short-wave infrared range. The work establishes a new geometric design paradigm for compact, intelligent computational optical systems.
- Journal
- PhotoniX
- Funder
- National Natural Science Foundation of China, National Key Research and Development Program of China