A Light Field FDL-HCGH Feature in Scale-Disparity Space

Area of Science:

• Computer Vision
• Image Processing
• Computational Photography

Background:

• Feature detection and description are crucial for many computer vision tasks.
• Existing 4D light field (LF) feature detectors and descriptors often lack computational efficiency and robustness.

Purpose of the Study:

• To propose a novel and efficient 4D light field feature descriptor.
• To improve feature matching accuracy and robustness in light field imaging.

Main Methods:

• A new light field feature descriptor is proposed, utilizing the Fourier disparity layer representation.
• Harris feature detection is performed in a scale-disparity space.
• A circular neighborhood is employed for feature descriptor extraction, differing from traditional square neighborhoods.

Main Results:

• The proposed descriptor demonstrates more accurate feature matching compared to the LiFF LF feature.
• It achieves lower computational complexity than existing methods.
• Experimental results on synthetic and real-world datasets show superior performance in feature detection robustness and matching accuracy.

Conclusions:

• The novel Fourier disparity layer-based feature descriptor offers a computationally efficient and robust solution for 4D light field applications.
• The method outperforms existing approaches in both feature detection and matching accuracy.
• This work contributes to advancing feature matching techniques in light field imaging.