Extended Data Fig. 1. Decomposition of RL deconvolution iteration and the internal structure of RLN and RLN-a.

a) RL deconvolution can be decomposed into four parts: forward projector (FP) function, division (DV) step, back projector (BP) function and update step. b) Schematic of RLN consisting of three parts: down-scale estimation starting from the average-pooled input image, H1; original-scale estimation starting with original-scale input image, H2; and merging/fine-tuning, H3. H1 and H2 are inspired by the RL deconvolution update formula, which mimic the unmatched forward/back projection steps. c) Schematic of RLN-a, without the DV and update steps in RLN. See Methods for more detail.