Short-wave infrared polarimetric image reconstruction using a deep convolutional neural network based on a...

Area of Science:

• Remote Sensing
• Computer Vision
• Optics

Background:

• Short-wave infrared (SWIR) imaging is crucial for remote sensing but often suffers from lower spatial resolution compared to visible imaging.
• Limited spatial resolution in SWIR images restricts the extraction of detailed information vital for various applications.
• Polarization information in SWIR can provide unique insights but is often coupled with resolution limitations.

Purpose of the Study:

• To develop a deep learning-based method for reconstructing high-resolution polarization SWIR images.
• To leverage readily available color images to enhance the spatial resolution of SWIR imagery.
• To recover potentially lost spatial and polarization details in SWIR images through advanced reconstruction techniques.

Main Methods:

• A deep learning model was trained using a dataset constructed from color images.
• The model was specifically designed for the task of SWIR image reconstruction.
• The method utilizes color image data to guide the enhancement of SWIR image resolution and polarization information.

Main Results:

• The proposed method effectively enhances the quality of polarized SWIR images.
• Significant improvements in spatial resolution were achieved for the reconstructed SWIR images.
• Experimental results demonstrate the capability to recover previously obscured spatial and polarization information.

Conclusions:

• Deep learning offers a powerful approach to overcome the spatial resolution limitations of SWIR cameras.
• The proposed method successfully reconstructs high-resolution polarization SWIR images, enhancing remote sensing data utility.
• This technique enables the recovery of valuable buried information, expanding the scope of SWIR remote sensing applications.