FiDALiS aims at the ideal and dynamic adaptation of the 5-dimensional sampling raster to the content of the captured scene. All sensors resp. sensory systems have a limited data-rate (measured in rays / second) and hence the captured information content is highly dependent on the distribution of the rays over the 4 spatial plus the temporal coordinate. The project fundamentally investigates and describes this adaptive 5-dimensional sampling.
5D Light Field SamplingRan LiMaster’s Thesis
Extending Deep Convolutional Demosaicing to Camera ArraysAlexander BlattMaster’s Thesis
Adding temporal consistency to an existing view interpolation algorithmJohannes ReuterMaster’s Thesis