Research on optimized multi-exposure image fusion method for improving information entropy in high-brightness region:...

Area of Science:

• Computer Vision
• Image Processing
• Computational Photography

Background:

• Overexposure in high-brightness regions causes significant information loss in images.
• Effective image analysis and measurement in challenging lighting conditions require robust preprocessing techniques.

Purpose of the Study:

• To develop a novel multi-exposure image fusion (MEF) method specifically for high-brightness environments.
• To enhance image information entropy (EN) and support image-based measurement tasks.

Main Methods:

• Simultaneous capture of low-, medium-, and high-exposure urban artificial light at night (ALAN) images.
• Development of a grayscale value weight matrix focusing on pixel value gradient information.
• An optimized MEF weighting strategy using saturation and contrast as supplementary references.
• Multi-scale fusion using a Laplacian pyramid.

Main Results:

• The proposed MEF method significantly improves the EN of fused ALAN regions by 86.49%.
• Mutual information (MI) of the fused images increased by 13.88% compared to classical MEF.
• Experimental validation across six different scenarios demonstrated superior performance.

Conclusions:

• The novel MEF method effectively enhances information in high-brightness regions of ALAN images.
• This approach offers a valuable preprocessing solution for image analysis and measurement tasks.
• The method shows significant improvements in information entropy and mutual information.