Research on a multi-dimensional image information fusion algorithm based on NSCT transform

Area of Science:

• Optics and Photonics
• Image Processing
• Materials Science

Background:

• Traditional inspection cameras rely solely on light intensity, limiting accuracy in complex environments.
• Polarization of light offers additional material information (roughness, texture, refractive index) for improved target recognition.

Purpose of the Study:

• To develop an image fusion algorithm for combining light intensity and polarized images.
• To enhance the accuracy of target detection and defect identification in power inspection.

Main Methods:

• Denoising and preprocessing of polarized images using noise template threshold matching.
• Image fusion using Non-Subsampled Contourlet Transform (NSCT) to merge light intensity and polarized images.

Main Results:

• The fused image demonstrated improved subjective and objective evaluation indicators compared to source images.
• Enhanced preservation of edge information in the fused image.
• Demonstrated potential for improved target recognition accuracy.

Conclusions:

• The developed image fusion technique effectively integrates multi-dimensional optical information.
• This approach offers a valuable reference for advanced optical inspection in power systems.
• Improved accuracy in target recognition is achievable through polarization and fusion techniques.