Physical prior guided cooperative learning framework for joint turbulence degradation estimation and infrared video restoration

TL;DR

The paper addresses the joint problem of atmospheric turbulence degradation estimation and infrared video restoration. It proposes Physical Prior Guided Cooperative Learning (P2GCL), featuring two networks: TMNet estimates turbulence strength and outputs the refractive index structure constant $C_n^2$ as a physical prior, while TRNet restores infrared sequences using $C_n^2$ and feeds the restored frames back to TMNet in a cyclic loop. Two physics-inspired losses are introduced: a $C_n^2$-guided frequency loss and a physical constraint loss to align learning with turbulence physics. Experiments show P2GCL achieves the best performance for turbulence strength estimation (MAE reduced by 0.0156 and $R^2$ increased by 0.1065) and infrared restoration (PSNR increased by 0.2775 dB). These results demonstrate the value of incorporating a physical prior into cooperative learning to improve measurement accuracy and image quality under atmospheric turbulence.

Abstract

Infrared imaging and turbulence strength measurements are in widespread demand in many fields. This paper introduces a Physical Prior Guided Cooperative Learning (P2GCL) framework to jointly enhance atmospheric turbulence strength estimation and infrared image restoration. P2GCL involves a cyclic collaboration between two models, i.e., a TMNet measures turbulence strength and outputs the refractive index structure constant (Cn2) as a physical prior, a TRNet conducts infrared image sequence restoration based on Cn2 and feeds the restored images back to the TMNet to boost the measurement accuracy. A novel Cn2-guided frequency loss function and a physical constraint loss are introduced to align the training process with physical theories. Experiments demonstrate P2GCL achieves the best performance for both turbulence strength estimation (improving Cn2 MAE by 0.0156, enhancing R2 by 0.1065) and image restoration (enhancing PSNR by 0.2775 dB), validating the significant impact of physical prior guided cooperative learning.