Quaternionic Local Ranking Binary Pattern: A Local Descriptor of Color Images.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Traditional image descriptors often process color channels independently.
• Existing methods may not fully capture the complex relationships within color image data.
• A need exists for descriptors that leverage the full color information efficiently.

Purpose of the Study:

• To propose a novel local descriptor for color images named Quaternionic Local Ranking Binary Pattern (QLRBP).
• To develop a method that processes color information in the quaternionic domain, considering inter-channel relationships.
• To evaluate the performance of QLRBP against state-of-the-art image descriptor techniques.

Main Methods:

• Color images are converted into their quaternionic representation (QR).
• A Clifford translation is applied to the QR.
• A reference quaternion is used to rank pixel QRs, followed by local binary coding on the phase to generate descriptors.

Main Results:

• The proposed Quaternionic Local Ranking Binary Pattern (QLRBP) descriptor was successfully developed.
• QLRBP effectively encodes relationships between color channels within the quaternionic domain.
• Experimental results show QLRBP surpasses several existing state-of-the-art methods.

Conclusions:

• QLRBP offers a robust and effective approach for color image description.
• The quaternionic representation enables simultaneous handling of all color channels and their interdependencies.
• QLRBP demonstrates superior performance compared to traditional and current advanced methods.