Peiqi Duan†, Boyu Li†, Yixin Yang, Hanyue Lou, Minggui Teng, Xinyu Zhou, Yi Ma, Boxin Shi‡
National Key Laboratory for Multimedia Information Processing and National Engineering Research Center of Visual Technology,
School of Computer Science, Peking University.
（† Contributed equally to this work as first authors, ‡ Corresponding author）
Event cameras are emerging imaging technology that offers advantages over conventional frame-based imaging sensors in dynamic range and sensing speed. Complementing the rich texture and color perception of traditional image frames, the hybrid camera system of event and frame-based cameras enables high-performance imaging. With the assistance of event cameras, high-quality image/video enhancement methods make it possible to break the limits of traditional frame-based cameras, especially exposure time, resolution, dynamic range, and frame rate limits. This paper focuses on five event-aided image and video enhancement tasks (i.e., event-based video reconstruction, event-aided high frame rate video reconstruction, image deblurring, image super-resolution, and high dynamic range image reconstruction), provides an analysis of the effects of different event properties, a real-captured and ground truth labeled benchmark dataset, a unified benchmarking of state-of-the-art methods, and an evaluation for two mainstream event simulators. In detail, this paper collects a real-captured evaluation dataset EVENTAID for five event-aided image/video enhancement tasks, by using “Event-RGB” multi-camera hybrid system, taking into account scene diversity and spatiotemporal synchronization. We further perform quantitative and visual comparisons for state-of-the-art algorithms, provide a controlled experiment to analyze the performance limit of event-aided image deblurring methods, and discuss open problems to inspire future research.
Link: Download from Dropbox
The equipment setup we used to collect the proposed dataset
(a) For the EventAid-B, we use one RGB camera to capture long-exposure blur images as the input, and another RGB camera to capture short-exposure clear images as the ground truth, the corresponding events are captured by one event camera. This setup can also collect EventAid-D where two different exposure images can be merged into one HDR image. (b) For the EventAid-R/-F/-S/-D dataset, we collocate an event camera and an RGB camera by mounting a 50:50 split ratio beam splitter in front of them. For each task, an RGB camera with corresponding attributes is selected to ensure that effective ground truths are captured. (c) We use a signal generator to simultaneously send square-wave signals to all cameras to achieve synchronized shooting.
Overview of the real-world scenes on the proposed datasets
The frequency of associations between 8 types of scene attributes is visualized in the chord diagram. Each attribute is subdivided into detailed categories, and one sample scene of each category is showcased on the left or right sides. The colored bands below the sample scenes serve to indicate the positions of categories under the same attribute.
@article{duan2025eventaid,
title={EventAid: Benchmarking Event-aided Image/Video Enhancement Algorithms with Real-captured Hybrid Dataset},
author={Duan, Peiqi and Li, Boyu and Yang, Yixin and Lou, Hanyue and Teng, Minggui and Ma, Yi and Shi, Boxin},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2025},
publisher={IEEE}
}