Improving Fast Auto-Focus with Event Polarity

TL;DR

This paper introduces a novel event-based auto-focus algorithm leveraging event polarity symmetry, achieving high-speed and precise focusing in challenging conditions with robustness validated on public datasets.

Contribution

The paper proposes a new focus evaluation function based on event polarity symmetry, significantly improving speed and accuracy in event-based auto-focus systems.

Findings

Achieves focus within 0.004 seconds on a high-speed platform

Demonstrates robustness on public event-based autofocus dataset

Provides publicly available dataset and code

Abstract

Fast and accurate auto-focus in adverse conditions remains an arduous task. The emergence of event cameras has opened up new possibilities for addressing the challenge. This paper presents a new high-speed and accurate event-based focusing algorithm. Specifically, the symmetrical relationship between the event polarities in focusing is investigated, and the event-based focus evaluation function is proposed based on the principles of the event cameras and the imaging model in the focusing process. Comprehensive experiments on the public event-based autofocus dataset (EAD) show the robustness of the model. Furthermore, precise focus with less than one depth of focus is achieved within 0.004 seconds on our self-built high-speed focusing platform. The dataset and code will be made publicly available.