Physiological Information Preserving Video Compression for rPPG

Area of Science:

• Biomedical Engineering
• Computer Vision
• Signal Processing

Background:

• Remote photoplethysmography (rPPG) offers non-contact physiological monitoring but is challenged by video compression.
• Conventional compression degrades rPPG signals, limiting real-time transmission and large dataset generation.

Purpose of the Study:

• To develop a specialized video compression scheme optimized for preserving rPPG signals.
• To enable efficient storage and real-time transmission of rPPG data for healthcare and AI.

Main Methods:

• A novel compression approach integrating facial region-of-interest (ROI) reallocation, rPPG signal-preserving bit allocation, and temporal up/down-sampling coding.
• Evaluation using UBFC-rPPG, ECG-Fitness, and a self-collected dataset.

Main Results:

• The proposed method achieved near-complete preservation of physiological information.
• Video compression reduced the original size by a factor of 60.

Conclusions:

• The developed compression scheme effectively preserves rPPG signals even at extreme compression ratios.
• This innovation is expected to advance telemedicine and deep learning in rPPG measurement by facilitating large-scale datasets.