Hybrid Event–Frame Imaging • Noise Modeling • Simulation

Hybrid Event–Frame Sensors: Modeling, Calibration, and Simulation

Yunfan LU1,2, Nico Messikommer2, Xiaogang Xu3, Liming Chen4, Yuhan Chen1, Nikola Zubic2, Davide Scaramuzza2, Hui Xiong1

1AI Thrust HKUST(GZ) · 2Robotics and Perception Group, University of Zurich · 3CUHK · 4AlpsenTek

Hybrid sensor layout and unified modeling–calibration–simulation framework
Overview: interleaved APS/EVS pixels (Quad-Bayer example), unified noise model, joint calibration, H-ESIM simulator, and downstream tasks such as frame interpolation and deblurring.
Placeholder for: images/hesim/1-CoverFigure.jpg

Abstract

Hybrid event–frame sensors place an APS and an EVS on the same chip, combining high dynamic range and low latency events with dense intensity frames. However, the mixed circuit introduces complex, poorly modeled noise in both modalities.

We propose a unified statistical noise model that jointly describes APS and EVS, covering photon shot, dark current, fixed-pattern and quantization noise, and linking event noise to illumination. Based on this model, we build a calibration pipeline on real hybrid sensors and introduce H-ESIM, a simulator that generates RAW frames and events with calibrated noise.

Experiments on two industrial hybrid sensors show that our model fits measured noise well and that networks trained or fine-tuned on H-ESIM generalize to real data for video interpolation and deblurring.

Method

Unified Noise Model

APS and EVS share the same latent signal \(I_c(t)\). APS integrates \(I_c\) over exposure to produce RAW intensity, while EVS reads log-intensity differences and fires events when the change exceeds a threshold.

Hybrid sensor imaging pipeline and pixel circuits
Imaging pipeline and pixel circuits: APS path with shot, dark-current, row, black-level and quantization noise; EVS path with logarithmic front-end and comparator driven by the same photodiode.
Placeholder for: images/hesim/2-APS-Noise-Calibration-Pixel-Circuit.jpg

Calibration from Real Sensors

We collect dark and static illuminated scenes under varied exposure times on two hybrid sensors (GEN2 and Eiger). APS calibration recovers clean intensity and per-position variance; EVS calibration uses APS intensity as a proxy for brightness to fit an intensity-dependent noise model.

Key steps of the H-ESIM simulator
H-ESIM: inverse color mapping from 3200 fps sRGB, APS noise injection with calibrated variance, and EVS simulation by mapping intensity to voltage, computing event probabilities and sampling events.
Placeholder for: images/hesim/7-HESIM.jpg

H-ESIM Simulator

Given high-frame-rate video, H-ESIM synthesizes noisy RAW frames and events. APS synthesis uses an sRGB→RAW inverse pipeline plus calibrated noise and quantization. EVS synthesis maps intensity to voltage and uses calibrated thresholds and noise parameters to sample ON/OFF events.

Results

Noise Calibration

APS calibration shows that hybrid layouts amplify row and fixed-pattern noise and that variance depends on brightness, exposure and CFA position. A second-order variance model matches measured noise across GEN2 and Eiger.

APS noise calibration on GEN2 and Eiger
APS noise calibration on GEN2 and Eiger: fixed components, variance maps, and model vs. measurement comparison.
Placeholder for: images/hesim/3-APS-Noise-Calibration-Light.jpg

EVS calibration confirms that noise event probability grows with brightness and follows the Q-function prediction, with polarity asymmetry explained by threshold–brightness coupling.

Noise events on GEN2 and Eiger
EVS noise analysis: event-probability maps, log-scale histograms and ON/OFF balance in dark and illuminated conditions.
Placeholder for: images/hesim/6-EVS-Noise-Calibration-Shot.jpg

Simulation Quality

From high-speed RGB videos, H-ESIM generates RAW frames under different exposures and matching events, with events aligning to motion edges and staying sparse in static regions.

Simulation examples of RAW frames and events
Simulation examples: sharp and motion-blurred RAW frames and their corresponding event streams under varying exposure.
Placeholder for: images/hesim/10-SimulateExamples.jpg

Downstream Tasks

We evaluate temporal and spatial tasks on real hybrid data: video frame interpolation (TimeLens, CBMNet, TimeLens-XL, HR-INR) and RAW deblurring (eSL, MADE, EFNet). Models fine-tuned on H-ESIM show sharper details and better perceptual scores.

Frame interpolation on hybrid inputs
Frame interpolation: HR-INR fine-tuned on H-ESIM reduces ghosting and recovers sharper structures on Eiger/GEN2 sequences.
Placeholder for: images/hesim/11-VFI-VIS.jpg

Citation

If you find this work useful, please cite:

@article{lu2025hybrid,
  title   = {Hybrid Event--Frame Sensors: Modeling, Calibration, and Simulation},
  author  = {Yunfan Lu and Nico Messikommer and Xiaogang Xu and Liming Chen
             and Yuhan Chen and Nikola Zubic and Davide Scaramuzza and Hui Xiong},
  note    = {Preprint},
  year    = {2025}
}