EAD

焦点控制 (FC) 对于相机在具有挑战性的现实场景中捕获清晰图像至关重要。自动对焦 (AF) 通过自动调整对焦设置来促进FC。然而，由于最近推出的事件相机缺乏有效的AF方法，它们的FC仍然依赖于像手动对焦调整一样的幼稚AF，导致在具有挑战性的现实世界条件下适应性差。特别是，事件数据和帧数据之间在感测模态，噪声和时间分辨率方面的固有差异在为事件摄像机设计有效的AF方法方面带来了许多挑战。为了应对这些挑战，我们开发了一种新颖的基于事件的自动聚焦框架，该框架由特定于事件的焦点度量 (称为事件率 (ER)) 和强大的搜索策略 (称为基于事件的黄金搜索 (EGS)) 组成。为了验证我们方法的性能，我们收集了一个基于事件的自动对焦数据集 (EAD)，其中包含在具有严重照明和运动条件的各种具有挑战性的场景中的良好同步的帧，事件和焦点位置。在此数据集和其他现实场景上的实验证明了我们的方法在效率和准确性方面优于最新方法。Focus Control (FC) is critical for cameras to capture sharp images in challenging real-world scenarios. Auto-Focus (AF) facilitates FC by automatically adjusting focus settings. However, due to the lack of effective AF methods for recently launched event cameras, their FC still relies on naive AF similar to manual focus adjustment, leading to poor adaptability under challenging real-world conditions. Particularly, the inherent differences between event data and frame data in terms of sensing modalities, noise, and temporal resolution pose numerous challenges to designing effective AF methods for event cameras. To address these challenges, we develop a novel event-based auto-focus framework, which consists of an event-specific focus metric called Event Rate (ER) and a robust search strategy named Event-based Golden Search (EGS). To validate the performance of our method, we collect an event-based auto-focus dataset (EAD) that contains well-synchronized frames, events, and focus positions across various challenging scenarios with severe lighting and motion conditions. Experiments on this dataset and other real-world scenarios demonstrate that our method outperforms state-of-the-art methods in terms of both efficiency and accuracy.

搜集汇总

背景与挑战

背景概述

EAD是一个专为事件相机自动对焦研究设计的数据集，包含在多种挑战性现实场景（如光照不足和快速运动）中同步采集的帧、事件和焦点位置数据。该数据集用于验证一种基于事件的新型自动对焦方法，包括事件率焦点度量和搜索策略，以提高对焦效率和准确性，由多所高校于2022年联合发布并支持CVPR 2022相关论文。