Texture characterization and classification of polarized images based on multi-angle orthogonal difference.

Area of Science:

• Computer Vision
• Image Processing
• Optics

Background:

• Local Binary Pattern (LBP) is effective for image texture extraction and classification.
• Standard LBP cannot capture texture information from polarized images at different angles.

Purpose of the Study:

• To develop a novel texture descriptor for polarized images.
• To address the limitations of existing LBP methods in characterizing polarized texture.

Main Methods:

• Proposed the multi-angle orthogonal difference polarization image texture descriptor (MODP_ITD).
• MODP_ITD integrates three patterns: MODP_LBP, MODP_LPCSP, and MODP_LDBP.
• Cascaded frequency histograms of these patterns to generate the final descriptor.

Main Results:

• The MODP_ITD accurately represents detailed texture information in polarization images.
• Experimental results demonstrated the descriptor's effectiveness on a custom polarization texture image database.
• Achieved competitive classification accuracy compared to existing LBP methods.

Conclusions:

• The MODP_ITD is a robust descriptor for polarized image texture analysis.
• This method offers significant advantages for classification tasks involving polarized image data.