Flow Map Denoisers: Traversing the Distortion-Perception Plane for Inverse Problems
This work introduces flow maps as a continuous family of denoisers indexed by a parameter 't'. By varying 't', researchers can continuously control image restoration quality along the distortion-perception (DP) frontier. The method proves exact optimality for Gaussian noise, showing how to find the best balance between data fidelity and perceptual quality in inverse problems.