Deep learning network for fusing optical and infrared images in a complex imaging environment by using the modified...

Area of Science:

• Image Processing
• Computer Vision
• Deep Learning

Background:

• Image fusion combines optical and infrared images, crucial for various applications.
• Complex environments pose challenges for achieving optimal fusion results.
• Existing methods struggle with low-quality images from challenging imaging conditions.

Purpose of the Study:

• To propose a novel deep learning network for effective image fusion.
• To address the challenges of fusing low-quality images from complex environments.
• To improve the accuracy and quality of fused images.

Main Methods:

• A modified U-Net based deep learning network with encoding and decoding structures was developed.
• Shared convolutional modules in encoding and decoding networks enhance fusion performance.
• An attention mechanism module was integrated into the decoding network to capture salient features.

Main Results:

• The proposed model demonstrates superior performance in fusing optical and infrared images.
• Subjective and objective evaluations confirm the effectiveness of the fusion method.
• The attention mechanism aids in extracting relevant features for more accurate fusion.

Conclusions:

• The developed deep learning model offers an effective solution for image fusion in complex environments.
• The integration of attention mechanisms significantly improves feature extraction and fusion accuracy.
• This approach advances the field of image fusion for low-quality and complex imaging scenarios.