Fig. 2: Illustration of GP application to 3D spectroscopic data set.
From: Imaging mechanism for hyperspectral scanning probe microscopy via Gaussian process modelling
Original image (ground truth) in a is corrupted by removing 70% of observations (measured spectroscopic curves) as shown in b. The GP regression is then used to reconstruct the signal (c). The absolute error is shown in d. Note that the absolute error may be misleading when the ratio of signals of interest does not change (see Fig. 3). The data set dimensions are 32 × 32 × 102.
