When
Where
Title: Multidimensional Imaging: Towards Enhanced Spectrum, Polarization, and 3D Vision
Abstract
The pursuit of richer visual information has constantly pushed beyond the limitations of conventional sensors. Multidimensional imaging, including spectrum, polarization, phase, depth, time, and other advanced dimensions, reflects this ongoing demand. Despite 2D sensor constraints, innovations in optics, computation, and machine learning have made it possible to extract and reconstruct complex data from limited inputs. This thesis presents three examples of multi-dimensional imaging, namely hyperspectral imaging, spectral-polarization 4D demosaicking and 3D endoscopic stereo vision, that integrate learning-based algorithms and bio-inspired optical designs to enhance the acquisition and reconstruction of complex multidimensional information. By combining optics and algorithms, the proposed modalities demonstrate how learning-driven optics can unlock richer, more precise representations of the visual world for scientific and clinical applications.