Infrared and Visible Image Fusion for Highlighting Salient Targets in the Night Scene

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Existing infrared and visible image fusion methods struggle with uneven nighttime luminance.
• Effective fusion is crucial for identifying salient targets and retaining textural details in night scenes.

Purpose of the Study:

• To develop an advanced infrared and visible image fusion method specifically designed for night scenes.
• To enhance the highlighting of salient targets while preserving rich textural information.

Main Methods:

• A global attention module was designed to rescale channel weights based on global contextual information.
• The loss function was segmented into foreground and background components to balance target salience and detail retention.
• A luminance estimation function was incorporated to dynamically adjust foreground loss parameters according to nighttime luminance.

Main Results:

• The proposed method effectively highlights salient targets in fused night scene images.
• Experimental results demonstrate superior fusion performance and generalization capabilities compared to existing advanced methods.
• The fusion process successfully retains rich texture details alongside salient target information.

Conclusions:

• The developed method offers a significant improvement for infrared and visible image fusion in challenging night conditions.
• The integration of a global attention module and adaptive loss function enhances the quality and utility of fused images.
• This approach provides a robust solution for applications requiring accurate target detection and detailed scene representation in low-light environments.