Multi-Focus Color Image Fusion Based on Quaternion Multi-Scale Singular Value Decomposition.

Area of Science:

• Image Processing
• Computer Vision
• Signal Processing

Background:

• Existing multi-focus color image fusion methods often alter color structures, leading to tonal distortion and blur.
• Separate fusion of color components in multi-scale decomposition causes loss of inherent color information.

Purpose of the Study:

• To propose a novel fusion algorithm that addresses color structure changes and tonal distortion in multi-focus color image fusion.
• To improve the visual quality of fused images by preserving color information and enhancing resolution.

Main Methods:

• Utilizing quaternion multi-scale singular value decomposition (QMSVD) for image decomposition.
• Employing activity level (local window energy) and matching level (quaternion color difference) for low-frequency sub-image fusion.
• Integrating low-frequency fusion results into high-frequency sub-image fusion using a local contrast rule.
• Reconstructing fused images via inverse QMSVD transform.

Main Results:

• The proposed QMSVD method effectively fuses multi-focus color images.
• Achieved superior visual effects with high resolution and rich colors.
• Demonstrated minimal information loss in the fused images compared to existing methods.

Conclusions:

• The QMSVD-based fusion algorithm successfully overcomes the limitations of traditional methods.
• This approach offers a significant improvement in multi-focus color image fusion quality.
• The method preserves color integrity and enhances image details effectively.