Multi-focus image fusion using fractal dimension

Area of Science:

• Computer Vision
• Digital Image Processing

Background:

• Multi-focus image fusion combines partially focused images into a fully focused one.
• Existing transform-domain methods often suffer from translation variance, impacting texture and edge preservation.
• Translation-invariant transforms simplify fusion rule design by producing consistent approximation and detail images.

Purpose of the Study:

• To develop a novel translation-invariant multi-focus image fusion method.
• To utilize the à-trous wavelet transform for enhanced fusion performance.
• To introduce fractal dimension and Otsu's threshold for improved clarity and detail fusion.

Main Methods:

• A translation-invariant approach using the à-trous wavelet transform was developed.
• Fractal dimension was employed as a clarity measure for approximation coefficients.
• Otsu's thresholding was used to fuse the detail coefficients.

Main Results:

• The proposed method demonstrated competitive and superior performance compared to nine state-of-the-art fusion techniques.
• Both subjective and objective assessments, using eight quality metrics, validated the method's effectiveness.
• The approach proved effective for both grayscale and color multi-focus image pairs.

Conclusions:

• The à-trous wavelet transform-based translation-invariant fusion method offers significant advantages.
• The integration of fractal dimension and Otsu's thresholding enhances fusion quality.
• The proposed method represents a valuable advancement in multi-focus image fusion technology.