LVTSR: learning visible image texture network for infrared polarization super-resolution imaging.

Area of Science:

• Optics and Photonics
• Image Processing
• Sensor Technology

Background:

• Infrared polarization (IRP) division-of-focal-plane (DoFP) imaging offers valuable information but suffers from low resolution due to sensor limitations.
• High-resolution visible light (VIS) imaging is readily available, presenting an opportunity for multi-modal image enhancement.
• Super-resolution (SR) for IRP DoFP is complex, requiring accurate polarization reconstruction beyond standard infrared SR.

Purpose of the Study:

• To develop an effective multi-modal super-resolution (SR) network for Infrared Polarization (IRP) division-of-focal-plane (DoFP) images.
• To leverage high-resolution visible light (VIS) images to enhance the resolution and polarization reconstruction of IRP DoFP images.
• To achieve end-to-end IRP DoFP SR by integrating VIS image constraints and polarization information into the network's loss function.

Main Methods:

• Proposed a novel multi-modal SR network integrating VIS image constraints for IRP DoFP image reconstruction.
• Incorporated polarization information as a key component within the loss function for end-to-end training.
• Created a benchmark dataset comprising 1559 pairs of registered multi-modal images for training and evaluation.

Main Results:

• The proposed method successfully utilized VIS images to restore polarization information in low-resolution IRP images.
• Achieved a 4x magnification factor, significantly enhancing the resolution of IRP DoFP images.
• Demonstrated superior quantitative and visual performance compared to existing state-of-the-art methods in multi-modal IRP SR.

Conclusions:

• The developed multi-modal SR network effectively enhances IRP DoFP imaging resolution by utilizing VIS image priors.
• The integration of polarization information in the loss function is crucial for accurate polarization reconstruction in SR tasks.
• The proposed approach offers a promising solution for high-resolution IRP DoFP imaging applications.