Extended Data Fig. 7: Results for rolling-shutter DiffuserCam.
From: Neural space–time model for dynamic multi-shot imaging
a, The raw image measurement. b, Comparisons of the reconstruction using basic deconvolution (assumes a static scene), FISTA with anisotropic 3D Total Variation regularization (TV)27 (the original reconstruction method), and our NSTM algorithm. c, NSTM reconstruction at different timepoints, with their corresponding measurement rows indicated by colored boxes on the raw image. The colored curves show some selected motion trajectories recovered by the motion network.
