Lossless Encoding of Time-Aggregated Neuromorphic Vision Sensor Data Based on Point-Cloud Compression

Area of Science:

• Computer Vision
• Data Compression
• Sensor Technology

Background:

• Neuromorphic Vision Sensors (NVSs) offer advantages like low power and high dynamic range over traditional cameras.
• NVS data, while inherently lower in data rate, can be further compressed for efficient storage and transmission.
• Previous methods like Time Aggregation-based Lossless Video Encoding for Neuromorphic Vision Sensor Data (TALVEN) have been explored.

Purpose of the Study:

• To introduce and evaluate a novel compression strategy for NVS data.
• To compare the effectiveness of the new strategy against prior methods.
• To analyze the impact of scene complexity and time aggregation intervals on compression performance.

Main Methods:

• Leveraging time aggregation of NVS events.
• Encoding time-aggregated data using point-cloud compression techniques.
• Implementing the Time-Aggregated Lossless Encoding of Events based on Point-Cloud Compression (TALEN-PCC) strategy.

Main Results:

• TALEN-PCC demonstrates superior compression ratios compared to the TALVEN strategy on the tested dataset.
• Compression gains are most significant in low-event rate and low-complexity scenes.
• Higher compression is achieved with time aggregation intervals exceeding 5 ms, though gains diminish compared to state-of-the-art for intervals under 5 ms.

Conclusions:

• TALEN-PCC represents an effective advancement in lossless compression for NVS data.
• The method's performance is sensitive to scene characteristics and temporal aggregation settings.
• Further research may optimize TALEN-PCC for various NVS applications and data types.