Sparsity-based inverse halftoning

Proposed is the sparsity-based inverse halftoning method, in which a sparse prior that reflects the natural gradient statistics is used with the sparse representation of the lowpass-filtered halftoned patches paired with the corresponding continuous patches to solve a regularised deconvolution. The experimental results show that the proposed method that was based on the sparsity, i.e. the combination of the sparse prior and the sparse representation, can reconstruct an unknown continuous image with less noise and fine details from an input halftoned image.