Event Camera Patents

Event camera (dynamic vision sensor) patents cover event-pixel and change-detection innovations; readout-architecture and arbitration innovations; stacked-sensor and integration innovations; and event-processing and application innovations — with IP held by the two pure-play event-sensor companies, image-sensor giants, and academic foundational labs. MAJOR EVENT-CAMERA PATENT HOLDERS: PROPHESEE: the Metavision event-based vision sensor platform and a large event-sensor estate, plus a partnership with Sony to co-develop stacked event sensors (Prophesee is a leading commercial event-camera IP holder). iniVATION: the Dynamic Vision Sensor DVS and DAVIS (combining a conventional frame output with asynchronous events), tracing to the foundational ETH Zurich / University of Zurich work. SONY: stacked back-illuminated event sensors (combining a photodiode wafer and a logic/event-processing wafer) co-developed with Prophesee. SAMSUNG: DVS sensors (Gen-3/Gen-4) and event-based vision research. OTHERS: OmniVision/CelePixel, and the ACADEMIC FOUNDATIONAL HOLDERS — Tobi Delbrück (ETH/UZH, the DVS/DAVIS pixel), Christoph Posch (ATIS, asynchronous time-based image sensor, now Prophesee), and Misha Mahowald/Carver Mead neuromorphic lineage. The event-pixel change-detector circuit, readout/arbitration, and stacked-sensor integration are the core event-camera patent domains — and the field competes with conventional frame cameras on latency, dynamic range, and power.

Event-pixel circuit innovations; change-detection and thresholding innovations; readout, arbitration, and timestamping innovations; and pixel-performance innovations represent core event-camera patent domains — and the asynchronous change-detecting pixel is the defining invention that makes event cameras fundamentally different from frame cameras. EVENT-PIXEL PATENTS: the pixel circuit that detects a LOG-intensity (brightness) change and asynchronously emits an 'event' (ON/OFF for brightness increase/decrease) only when the change crosses a threshold — no global shutter, no frames; including the photoreceptor, the change-amplifier/comparator, and pixel layout. Intensity-readout variants (ATIS — emitting an event plus an exposure measurement; DAVIS — combining frames and events) are key architectural patents. CHANGE-DETECTION / THRESHOLD PATENTS: adjustable contrast thresholds, noise (background-activity) filtering at the pixel/array level, and bandwidth control. READOUT / ARBITRATION PATENTS: address-event-representation AER readout (arbitrating which pixels report when, in order), timestamping with microsecond resolution, handling event bursts/saturation, and data-rate management (event cameras produce sparse, asynchronous data, not frames). PIXEL-PERFORMANCE PATENTS: high dynamic range (>120 dB — the log response handles extreme lighting), low latency (microseconds), low power (only active pixels consume), small pixel pitch, and low noise. The change-detecting event pixel and the AER readout/arbitration are the highest-value, most-foundational event-camera IP.

Stacked-sensor and 3D-integration innovations; pixel-pitch and resolution innovations; event-processing and representation innovations; and application and fusion innovations represent additional event-camera patent domains — though event-processing algorithm claims face §101 scrutiny and pair with the sensor. STACKED-SENSOR PATENTS: stacking a back-illuminated photodiode wafer on a logic/event-processing wafer (Sony/Prophesee) to shrink pixel pitch and add on-chip processing — wafer bonding, per-pixel/per-region logic, and on-sensor event filtering/compression; stacking is the key to making event sensors small and high-resolution enough for consumer/automotive use. EVENT-PROCESSING / REPRESENTATION PATENTS: converting sparse asynchronous events into useful outputs — event-to-frame/voxel representations, optical flow, feature tracking, depth/SLAM, object detection, and spiking-neural-network processing of events — these algorithm-heavy claims are most defensible claimed with the specific sensor and a concrete result (a bare 'process events with a neural network' claim is §101-vulnerable). FUSION PATENTS: fusing event data with frame cameras, IMU, or lidar (leveraging events' low latency to deblur or interpolate frames). APPLICATION PATENTS: high-speed machine vision, automotive (fast obstacle detection), AR/VR eye tracking and SLAM (events' low power/latency fit head-mounted use), industrial monitoring, and scientific imaging. Stacked-sensor integration (for small, high-resolution sensors) and sensor-coupled event processing are the highest-value applied event-camera IP.

Event camera startup IP strategy must navigate Prophesee's and iniVation's substantial event-sensor estates, Sony/Samsung image-sensor patents, the foundational academic DVS/ATIS patents (ETH/UZH Delbrück, Posch — some licensable, some underlying the incumbents), conventional CMOS image-sensor prior art, foundry dependence, a §101 constraint on event-processing algorithms, and a market where adoption hinges on cost, resolution, and the software ecosystem; understand that the core event-pixel and AER-readout concepts are foundationally patented (FTO/licensing needed), that the durable startup IP is in novel pixel circuits, stacked-sensor integration, low-noise/high-resolution designs, sensor-coupled processing, and applications, and that the algorithmic/software ecosystem and demonstrated benefit matter as much as patents; identify whitespace in high-resolution stacked pixels, low-noise pixels, sensor-side processing, fusion, and specific applications. EVENT-CAMERA STARTUP IP STRATEGY: CORE PIXEL/READOUT IS FOUNDATIONALLY PATENTED — NOVEL PIXELS, STACKING, AND PROCESSING ARE THE IP: the change-detecting pixel and AER readout are foundationally held (Prophesee/iniVation/academic) — patent a novel pixel circuit, stacked-sensor integration, low-noise/high-resolution design, or sensor-coupled processing; STACKED-SENSOR INTEGRATION IS HIGHEST-VALUE WHITESPACE: stacking photodiode + logic wafers to get small pixels, high resolution, and on-chip processing is what makes event cameras viable for consumer/automotive — the most commercially decisive patent terrain (Sony/Prophesee lead, but room exists); LOW NOISE, HIGH RESOLUTION, AND HDR ARE COMPETITIVE AXES: reducing background-activity noise and increasing resolution/dynamic range are patentable improvements that drive adoption; EVENT PROCESSING MUST BE TIED TO THE SENSOR (§101): claim optical-flow/SLAM/detection processing with the specific event sensor, not as a bare algorithm; APPLICATIONS AND FUSION ARE PRACTICAL WHITESPACE: AR/VR eye tracking, automotive, high-speed industrial, and event+frame fusion are growing application areas with patentable system IP; ECOSYSTEM/SOFTWARE IS A MOAT: event cameras need tooling and algorithms to be useful — the software ecosystem matters as much as hardware patents; WHEN TO PATENT: NOVEL SENSOR/SYSTEM WITH MEASURED PERFORMANCE: file once a sensor shows measured results (latency µs + dynamic range dB + resolution + power mW + noise/background-activity + pixel pitch) vs. DVS/ATIS/frame baselines — measured latency, dynamic range, resolution, power, and noise are the critical event-camera IP metrics; KEY FTO CHECKLIST: Prophesee Metavision event sensor + ATIS (Posch), Sony stacked partnership; iniVation DVS/DAVIS frame+event (Delbrück ETH/UZH); Sony/Samsung stacked BSI event sensor; change-detecting log-intensity event pixel, photoreceptor/comparator; AER address-event-representation readout/arbitration/timestamp; background-activity noise filtering; stacked photodiode+logic wafer bonding on-chip processing; HDR/low-latency/low-power; event-to-frame/optical-flow/SLAM processing (§101-tied-to-sensor); event+frame/IMU/lidar fusion; AR/VR/automotive application; CMOS image-sensor prior art.