Development and Evaluation of a Keypoint-Based Video Stabilization Pipeline for Oral Capillaroscopy

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Microcirculation Research

Background:

• Capillaroscopy of the oral mucosa is vital for microcirculation assessment but suffers from motion artifacts.
• Video instability compromises diagnostic accuracy and automated analysis of capillaries.

Purpose of the Study:

• To develop and evaluate a comprehensive video stabilization model for oral capillaroscopy.
• To improve the quality and reliability of videocapillaroscopy imaging for better diagnostic insights.

Main Methods:

• A multi-phase pipeline integrating keypoint extraction (SIFT, ORB, GFTT), optical flow estimation, and affine transformation-based frame alignment.
• Introduction of synthetic tremors using Gaussian affine transformations to simulate real-world acquisition conditions.
• Performance evaluation using Structural Similarity Index Measure (SSIM) and jitter reduction metrics.

Main Results:

• The proposed stabilization model effectively reduced jitter and improved image quality, achieving an average SSIM of 0.789.
• All tested keypoint extraction algorithms (SIFT, ORB, GFTT) demonstrated comparable stabilization performance.
• GFTT offered slightly higher structural fidelity, while ORB provided superior computational efficiency.

Conclusions:

• The implemented video stabilization model significantly enhances oral capillaroscopy image quality.
• The findings support the potential of this method for improving diagnostic accuracy and enabling automated capillary analysis.
• The model offers comparable or superior performance to existing methods in nailfold capillaroscopy while maintaining efficiency.