FLOW: Feature-Level Optimal Warping for Generalized Remote Physiological Measurement

Remote photoplethysmography (rPPG) enables non-contact physiological measurement but remains vulnerable to domain shifts from illumination, motion, and sensors. We propose $\mathbf{\text{FLOW}}\mathbf{\text{(Feature-Level}}\mathbf{\text{Optimal}}\mathbf{\text{Warping)}}$, an \emph{optimal transport–driven} framework for domain-generalized rPPG. FLOW integrates a $\mathbf{\text{Temporal}}\mathbf{\text{Refinement}}\mathbf{\text{Module}}\mathbf{\text{(TRM)}}$ to stabilize temporal dynamics and a $\mathbf{\text{Prototype-based}}\mathbf{\text{Cross-Temporal}}\mathbf{\text{Optimal}}\mathbf{\text{Transport}}\mathbf{\text{(PCOT)}}$ module to achieve domain-invariant alignment via learnable prototypes.Beyond feature alignment, FLOW employs soft cross-temporal correspondence modeling that aligns temporal features in a flexible manner, allowing the model to respect and preserve the intrinsic rhythmic patterns of physiological signals. Moreover, the lightweight design of our modules allows seamless integration into existing end-to-end rPPG architectures without additional preprocessing. Two regularization terms further enforce source consistency and identity preservation. Theoretically, we derive a generalization bound under conditional optimal transport. Extensive experiments across four rPPG benchmarks show that FLOW achieves state-of-the-art cross-domain performance with lightweight design and strong physiological fidelity.