Infrared and Visible Image Fusion Based on Visual Saliency Map and Image Contrast Enhancement

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Traditional image fusion methods struggle with artifacts and preserving details, especially in low-visibility conditions.
• Existing techniques often fail to adequately enhance target saliency and texture information in fused images.
• Dark scenes and smoke significantly degrade the quality and information content of fused infrared and visible images.

Purpose of the Study:

• To develop an improved infrared and visible image fusion method addressing limitations of traditional techniques.
• To enhance target prominence and preserve rich information for better target detection and recognition.
• To overcome challenges in image fusion under adverse conditions like low contrast, dark scenes, and smoke.

Main Methods:

• Utilized an improved gamma correction and local mean method for input image contrast enhancement.
• Employed a differential rolling guidance filter (DRGF) for image decomposition into basic and detail layers, suppressing artifacts.
• Incorporated visual saliency maps to guide target weighting and control basic layer fusion, improving contrast preservation.
• Developed a pixel intensity and gradient-based method for detail layer fusion to retain edge information.

Main Results:

• The proposed method effectively suppresses artifacts commonly found in traditional image fusion.
• Enhanced contrast and preserved target saliency and texture details in fused images.
• Demonstrated superior performance over existing fusion algorithms in both subjective and objective evaluations.
• Successfully improved image fusion quality under challenging conditions such as dark scenes and smoke.

Conclusions:

• The novel infrared and visible image fusion method offers significant improvements over existing techniques.
• The approach effectively balances target prominence, detail preservation, and artifact reduction.
• This method provides a robust solution for image fusion in demanding environmental conditions, enhancing target recognition capabilities.