Local Parallel Cross Pattern: A Color Texture Descriptor for Image Retrieval.

Area of Science:

• Computer Science
• Image Processing
• Pattern Recognition

Background:

• Image retrieval (IR) technology is crucial for applications like e-commerce and surveillance.
• Designing effective feature descriptors remains a key challenge in image retrieval.
• The proliferation of visual sensor equipment fuels the demand for advanced IR solutions.

Purpose of the Study:

• To propose novel feature descriptors for enhanced image retrieval.
• To address the bottleneck of effective feature descriptor design in image retrieval.
• To improve the efficiency and robustness of image retrieval systems.

Main Methods:

• Constructed a six-layer color quantizer to extract color maps.
• Designed a local parallel cross pattern (LPCP) by combining Local Binary Pattern (LBP) and color maps.
• Extended LPCP to uniform (ULPCP) and rotation-invariant (RILPCP) versions to reduce complexity and enhance robustness.

Main Results:

• The proposed LPCP, ULPCP, and RILPCP descriptors demonstrated effectiveness and efficiency on eight benchmark datasets.
• Experimental results showed superior performance compared to eight state-of-the-art color texture descriptors.
• The novel descriptors achieved competitive results against Convolutional Neural Network (CNN)-based models.

Conclusions:

• The developed local parallel cross pattern descriptors offer significant improvements in image retrieval.
• The proposed methods provide a robust and computationally efficient alternative to existing techniques.
• These findings contribute to advancing the field of image retrieval and feature descriptor design.