Research map / 3D vision / Depth and stereo
Intensity-Robust Autofocus for Spike Camera
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a spike-stream autofocus framework with a light-intensity-robust focus measure (spike dispersion) and a fast golden-section-style search (SGFS), evaluated on a new spike autofocus dataset.
Problem
Autofocus control for spike cameras is lacking; existing frame- and event-based focus measures fail under scene light intensity fluctuations and spike streams are noisy and high-throughput, making traversal-based search slow.
Approach
Spike dispersion normalizes the AC power of spike integrals by the mean power, cancelling light-intensity effects, and SGFS adapts the event-based golden search with real-time focus feedback so it need not traverse all focal positions.
Key results
- SGFS converges typically about twice as fast as EGS on real macroscopic and microscopic spike data.
- Spike dispersion keeps a usable focus curve under continuous and sudden light changes where gradient, MF-DCT and ER fail (qualitative curves on simulated and USAF data).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use spike dispersion (SD) instead of gradient-based, MF-DCT, or extended event-rate focus measures | Spike-based autofocus · Synthetic autofocus data with continuous or fluctuating illumination | focus-measure robustness | Helped |
| Use spike dispersion (SD) instead of gradient-based, MF-DCT, or extended event-rate focus measures | Spike-based autofocus · USAF real-world data with continuous or fluctuating illumination | focus-measure robustness | Helped |
| Use spike dispersion (SD) instead of gradient-based, MF-DCT, or extended event-rate focus measures | Spike-based autofocus · Lily anther section and fan real-world data | focus-measure robustness | Helped |
| Use spike-based golden fast search (SGFS) instead of event-based golden search (EGS) | Spike-based autofocus with SD · Bottle and cardiac muscle section real-world data | search speed: more than twice as fast | Helped |
Surprising findings
- Frame-based measures and event rate fail under light fluctuations even when computed on TFP reconstructions, while SD remains effective.
Limitations
- Performance on scenes with multiple targets is currently subpar, requiring detection/recognition integration.
- Evaluation is largely qualitative focus-curve comparisons rather than extensive quantitative metrics in the main text.
Experimental setup
Datasets: camel dataset crossroad dataset USAF target captures
Benchmarks: spike-based autofocus dataset (SAD)
Metrics: focus score curves convergence speed
Baselines: gradient-based focus measure MF-DCT event rate (ER) EGS TFP reconstruction
Builds on: event-based golden search (EGS) texture from playback (TFP) MF-DCT
Closest papers by meaning
- SpikeNeRF: Learning Neural Radiance Fields from Continuous Spike Stream (CVPR 2024 · similarity 0.88)
- Boosting Spike Camera Image Reconstruction from a Perspective of Dealing with Spike Fluctuations (CVPR 2024 · similarity 0.86)
- Finding Visual Saliency in Continuous Spike Stream (AAAI 2024 · similarity 0.86)
- Rethinking High-speed Image Reconstruction Framework with Spike Camera (AAAI 2025 · similarity 0.85)
- Recognizing Ultra-High-Speed Moving Objects with Bio-Inspired Spike Camera (AAAI 2024 · similarity 0.85)
- Optical Flow for Spike Camera with Hierarchical Spatial-Temporal Spike Fusion (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.