RGB-IR Cross Input and Sub-Pixel Upsampling Network for Infrared Image Super-Resolution.

Area of Science:

• Computer Vision
• Image Processing
• Deep Learning

Background:

• Deep learning excels at image super-resolution, enhancing image quality.
• Infrared (IR) image resolution often requires improvement for detailed analysis.

Purpose of the Study:

• To develop a deep learning network for enhancing IR image spatial resolution.
• To improve IR image quality by leveraging co-registered RGB data.

Main Methods:

• Proposed an RGB-IR cross-input and sub-pixel upsampling network.
• Fused feature maps from RGB and IR inputs for reconstruction.
• Utilized sub-pixel convolution for accurate feature extraction and upsampling.
• Incorporated a guided filter layer for IR image denoising and detail preservation.

Main Results:

• The proposed network demonstrated superior accuracy and visual improvement on IR image super-resolution.
• Experimental results validated the method's effectiveness on custom and existing datasets.
• The fusion of RGB and IR data significantly enhanced the reconstruction of high-resolution IR images.

Conclusions:

• The novel network effectively increases the spatial resolution of IR images using complementary RGB data.
• The approach offers a significant advancement in IR image super-resolution and quality enhancement.
• This method provides a robust solution for applications requiring high-resolution infrared imagery.