Completed local ternary pattern for rotation invariant texture classification.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Existing texture descriptors like Completed Local Binary Pattern (CLBP) and Completed Local Binary Count (CLBC) show high accuracy but inherit drawbacks from Local Binary Pattern (LBP).
• Local Binary Pattern (LBP) is sensitive to noise and has limited discriminative power due to pattern classification issues.
• Local Ternary Pattern (LTP) offers improved noise robustness over LBP, but can still exhibit similar weaknesses.

Purpose of the Study:

• To propose a novel completed Local Ternary Pattern (CLTP) operator.
• To develop an associated CLTP scheme for rotation invariant texture classification.
• To overcome the noise sensitivity and discrimination limitations of LBP and LTP.

Main Methods:

• Development of a novel completed Local Ternary Pattern (CLTP) operator.
• Implementation of a CLTP scheme for rotation invariant texture classification.
• Experimental evaluation on four diverse texture databases.

Main Results:

• The proposed CLTP descriptor achieved impressive classification accuracy.
• CLTP demonstrated superior performance compared to CLBP and CLBC descriptors.
• The method effectively addresses rotation invariance in texture classification.

Conclusions:

• The novel CLTP operator and scheme provide a robust solution for rotation invariant texture classification.
• CLTP significantly enhances classification accuracy over existing methods.
• This approach offers a promising advancement in texture analysis.