Layer-Based Approach for Image Pair Fusion.

Area of Science:

• Computer Vision
• Image Processing
• Computational Photography

Background:

• Image pairs (e.g., noisy/blurred, infrared/noisy) are used for high-quality low-light photography.
• Naive fusion of infrared and noisy images yields poor results due to structural and brightness differences.

Purpose of the Study:

• To propose a new layer-based image fusion method for image pairs.
• To address challenges in fusing infrared and noisy images by preserving local contrast and enhancing details.

Main Methods:

• Decomposing image pairs into base and detail layers.
• A local contrast-preserving conversion creates a compatible infrared base layer.
• An optimization framework with sparsity and redundancy priors estimates the noise-free detail layer.
• An iterative feedback loop refines the noisy image's detail layer for noise suppression.

Main Results:

• The proposed method effectively fuses infrared and noisy image pairs.
• The layer-based approach demonstrates applicability to noisy and blurred image fusion.
• Experimental results confirm the method's effectiveness in image pair fusion.

Conclusions:

• The developed layer-based fusion method significantly improves image quality in low-light conditions.
• The technique offers a robust solution for fusing diverse image pairs, including infrared/noisy and noisy/blurred.
• This approach enhances detail preservation and noise suppression in fused images.