UOTIP: Unbalanced Optimal Transport Map for Unpaired Inverse Problems
Researchers have introduced UOTIP, a new method for solving unpaired image inverse problems. This technique utilizes Unbalanced Optimal Transport to learn a mapping between noisy measurements and clean target signals. UOTIP is designed to be robust to various noise levels and class imbalances in datasets, offering improved performance on both linear and nonlinear inverse problems. AI
IMPACT Introduces a novel method for image reconstruction, potentially improving performance in applications relying on inverse problem solving.