Multiscale image fusion using the undecimated wavelet transform with spectral factorization and nonorthogonal filter...

Area of Science:

• Image processing
• Signal processing
• Computer vision

Background:

• Multiscale transforms are popular for pixel-level image fusion.
• Fusion performance degrades for images from different sensor modalities.

Purpose of the Study:

• To improve image fusion for different sensor modalities.
• To reduce artifacts and reconstruction errors in fused images.

Main Methods:

• A novel undecimated wavelet transform (UWT)-based fusion scheme.
• Splitting image decomposition into two successive filtering operations.
• Using spectral factorization of analysis filters and nonorthogonal filter banks.

Main Results:

• Minimized spreading of coefficient values around singularities.
• Reduced reconstruction errors and artifacts like ringing.
• Improved fusion performance for cross-modality images.

Conclusions:

• The UWT-based fusion framework offers significant advantages over traditional methods.
• The approach is robust to artifacts and enhances feature selection.
• This method provides clearer fused reconstructions for diverse image sources.