Field-programmable encoding for address-event representation

Area of Science:

• Computer Engineering
• Electrical Engineering
• Image Sensor Technology

Background:

• Conventional frame-based image sensors have high bandwidth and power demands due to full-frame readout.
• Event-based sensors reduce data transmission using address-event representation (AER) but face latency issues with increased activity.
• Existing AER protocols struggle with scalability and error rates in high-event scenarios.

Purpose of the Study:

• To develop a novel image sensor readout scheme that combines the advantages of frame-based and event-based approaches.
• To introduce a field-programmable address-event representation (FP-AER) encoding scheme for flexible image sensor operation.
• To evaluate the performance of FP-AER against existing AER methods in imaging applications.

Main Methods:

• Development of a new field-programmable AER (FP-AER) encoding scheme.
• Implementation of a configurable readout design using field-programmable gate arrays (FPGAs) or similar technology.
• Comparative performance analysis of FP-AER against conventional AER protocols under various imaging conditions.

Main Results:

• The proposed FP-AER scheme successfully integrates benefits from both frame-based and event-based readout strategies.
• FP-AER demonstrates superior performance compared to existing AER-based approaches in both scanning and event-driven readout modes.
• The field-programmable nature allows for on-the-fly configuration, enhancing adaptability.

Conclusions:

• FP-AER offers a significant advancement in image sensor technology by overcoming limitations of existing methods.
• The proposed scheme provides a flexible and efficient solution for reducing bandwidth and power consumption in image sensors.
• FP-AER represents a promising direction for next-generation intelligent imaging systems.