Multi-Spectral Fusion and Denoising of Color and Near-Infrared Images Using Multi-Scale Wavelet Analysis.

Area of Science:

• Image Processing and Computer Vision
• Multi-spectral Image Fusion
• Signal Denoising

Background:

• Integrating Red, Green, Blue (RGB) and Near-Infrared (NIR) data presents challenges due to spectral discrepancies.
• Existing fusion methods often struggle with noise reduction and detail preservation while maintaining color fidelity.
• Wavelet domain processing offers a powerful framework for multi-resolution analysis and feature extraction in image fusion.

Purpose of the Study:

• To develop an advanced multi-spectral fusion and denoising technique for RGB and NIR images.
• To address the spectral distribution differences between RGB and NIR data during the fusion process.
• To enhance image quality by preserving details, reducing noise, and improving color rendition in the fused output.

Main Methods:

• Formulated fusion and denoising as a maximum a posteriori estimation problem in the wavelet domain.
• Introduced a wavelet scale map to model and reconcile discrepancies between RGB and NIR data distributions.
• Employed contrast preservation and gradient denoising terms, utilizing local contrast, visibility, and NIR wavelet coefficient gradients.

Main Results:

• Successfully fused RGB and NIR images, demonstrating significant noise reduction and detail preservation.
• Achieved effective color enhancement in the fused images, guided by luminance variations post-fusion.
• The wavelet scale map effectively adjusted NIR data to match RGB distributions, improving fusion accuracy.

Conclusions:

• The proposed method provides a robust solution for multi-spectral image fusion and denoising.
• The technique successfully integrates complementary information from RGB and NIR spectra for superior image quality.
• This approach offers enhanced visual fidelity, including better detail, reduced noise, and improved color characteristics.