Research · Inverse problems
Classical Reconstruction-Based Super-Resolution
A non-learning single-image SR pipeline posed as an inverse problem: progressive upscaling, edge-guided refinement and iterative back-projection.
A reconstruction-first super-resolution pipeline that performs progressive 2× upscaling, simulates forward degradation (blur and downsample), estimates residuals, and back-projects corrections to enforce fidelity. Edge-guided refinement and anti-ringing constraints stabilise the result without inventing detail.