Improved Thermal Infrared Image Super-Resolution Reconstruction Method Base on Multimodal Sensor Fusion

Area of Science:

• Computer Vision
• Image Processing
• Sensor Fusion

Background:

• Traditional super-resolution methods struggle with poor image quality in thermal infrared imaging.
• Existing techniques fail to adequately address limitations inherent to the imaging mechanism.

Purpose of the Study:

• To develop a novel thermal infrared image super-resolution reconstruction method.
• To enhance image resolution and reconstruct high-frequency details using multimodal sensor information.
• To overcome the inherent limitations of thermal infrared imaging mechanisms.

Main Methods:

• Designed a novel super-resolution reconstruction network with primary feature encoding, reconstruction, and high-frequency detail fusion subnetworks.
• Incorporated hierarchical dilated distillation modules and a cross-attention transformation module for advanced feature extraction.
• Utilized a hybrid loss function and a specialized learning strategy for robust performance, even without reference images.

Main Results:

• The proposed method significantly enhances the resolution of thermal infrared images.
• High-frequency details are effectively reconstructed by leveraging multimodal sensor data.
• Experimental results demonstrate superior reconstruction quality compared to existing contrastive methods.

Conclusions:

• The multimodal sensor fusion approach effectively overcomes limitations in thermal infrared super-resolution.
• The novel network architecture and loss function contribute to high-quality image reconstruction.
• The method shows significant potential for improving thermal imaging applications.