Md Raihan Khan

Md Raihan Khan

Lecturer in EEE (North Western University) & B.Sc. graduate (KUET). Efficient, geometry-aware deep learning: generative modeling, optimal transport, and neural signals.

AutoLumNet: Monotone Optimal Transport for Single-Shot Exposure Correction

Published in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) — under review, 2026, 2026

AutoLumNet is a bi-branch, exposure-aware network that performs single-shot correction of both under- and over-exposed images. Rather than relying on heuristic reconstruction losses, the correction step is reformulated as monotone optimal transport, yielding a principled mapping between degraded and well-exposed intensity distributions.

The method attains an average MSEC score of 23.75, remaining competitive with or superior to state-of-the-art low-light and exposure-correction models while preserving structural consistency and natural color.

Recommended citation: Airin Akter Tania, Md Raihan Khan, and Mohiuddin Ahmad, "AutoLumNet: Monotone Optimal Transport for Single-Shot Exposure Correction," IEEE TPAMI (under review, 2026).