Infrared and Visual Image Fusion Based on a Local-Extrema-Driven Image Filter

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Infrared and visual image fusion aims to integrate complementary information from both modalities.
• Existing methods often struggle to effectively preserve salient features from both image types.

Purpose of the Study:

• To develop a novel image fusion method that effectively amalgamates infrared and visual image features.
• To enhance the informative content of the fused image by leveraging local extrema for feature extraction.

Main Methods:

• A novel local-extrema-driven image filter is proposed for image smoothing and feature extraction.
• The filter is iteratively applied to extract multi-scale bright and dark feature maps.
• Fusion of feature maps uses elementwise-maximum and elementwise-minimum strategies, while base images are fused using structural similarity and intensity.
• The final fused image is constructed by combining fused feature maps and base images.

Main Results:

• The proposed method demonstrates superior performance in infrared and visual image fusion.
• Experimental results on the TNO dataset show the method matches or surpasses eleven state-of-the-art fusion techniques.
• Both qualitative and quantitative assessments confirm the effectiveness of the proposed approach.

Conclusions:

• The local-extrema-driven filter provides an effective mechanism for infrared and visual image fusion.
• The developed fusion strategy successfully integrates complementary features, leading to enhanced image quality.
• The method offers a robust and high-performing solution for multi-modal image fusion applications.