Illustration of a pixelated mapping between source and image planes, as implemented in SLITRONOMY. Top row, left to right: nearest-neighbour interpolation of source surface brightness for two grid resolutions (same as image plane or doubled), bilinear interpolation of surface brightness for the same resolutions, and parametric “groundtruth” using LENSTRONOMY. Bottom row, left to right: source plane corresponding to the top row panels, for different data pixel size to source pixel ratio r_pix. Dark isolated pixels in source plane are not mapped to any image plane pixel, hence not constrained by imaging data. Our reconstruction technique is able to fill these “missing pixels”, through sparse regularisation and multiresolution property of wavelets (https://github.com/aymgal/SLITronomy-papers/blob/master/paper_I/visualize_lensing_operator.ipynb).