Abstract
In this paper, we explore how conventional image enhancement can improvemodel robustness in medical image analysis. By applying commonly usednormalization methods to images from various vendors and studying theirinfluence on model generalization in transfer learning, we show that thenonlinear characteristics of domain-specific image dynamics cannot be addressedby simple linear transforms. To tackle this issue, we reformulate the imageharmonization task as an exposure correction problem and propose a methodtermed Global Deep Curve Estimation (GDCE) to reduce domain-specific exposuremismatch. GDCE performs enhancement via a pre-defined polynomial function andis trained with the help of a ``domain discriminator'', aiming to improve modeltransparency in downstream tasks compared to existing black-box methods.