Event-based Vision: A Survey

Abstract

Event cameras are bio-inspired sensors that differ from conventional frame cameras: Instead of capturing images at a fixed rate, they asynchronously measure per-pixel brightness changes, and output a stream of events that encode the time, location and sign of the brightness changes. Event cameras offer attractive properties compared to traditional cameras: high temporal resolution (in the order of µs), very high dynamic range (140 dB vs. 60 dB), low power consumption, and high pixel bandwidth (on the order of kHz) resulting in reduced motion blur. Hence, event cameras have a large potential for robotics and computer vision in challenging scenarios for traditional cameras, such as low-latency, high speed, and high dynamic range. However, novel methods are required to process the unconventional output of these sensors in order to unlock their potential. This paper provides a comprehensive overview of the emerging field of event-based vision, with a focus on the applications and the algorithms developed to unlock the outstanding properties of event cameras. We present event cameras from their working principle, the actual sensors that are available and the tasks that they have been used for, from low-level vision (feature detection and tracking, optic flow, etc.) to high-level vision (reconstruction, segmentation, recognition). We also discuss the techniques developed to process events, including learning-based techniques, as well as specialized processors for these novel sensors, such as spiking neural networks. Additionally, we highlight the challenges that remain to be tackled and the opportunities that lie ahead in the search for a more efficient, bio-inspired way for machines to perceive and interact with the world.

Real-Time Gaze Tracking with Event-Driven Eye Segmentation

“…Event Camera Event cameras operate each pixel independently and asynchronously [35]. Each pixel gets activated when its intensity change surpasses a predefined threshold.…”

“…The response is called an "event", which includes the pixel coordinates, a timestamp, and a polarity value. Because of the increased hardware complexity, the resolution of event cameras is typically lower than that of conventional cameras [35], but event cameras have a much higher frame rate (upward of tens of thousands of Hz) since they produce only (sparse) events occasionally rather than (dense) pixels regularly.…”

“…We generate the event map by emulating the high-level process of an event camera [35]. We first calculate the pixel-wise absolute difference, ∆F t+1 , between two consecutive frames, F t and F t+1 .…”

“…While our algorithm relies on events, the two additional cues of ROI and edge map from the previous frame are critical to the accuracy. , where the pixel values are naturally in the log scale [35].…”

Instantaneous Stereo Depth Estimation of Real-World Stimuli with a Neuromorphic Stereo-Vision Setup

“…While efficiently solved in biological systems, classical machine vision approaches require significant computational resources: Indeed, by sampling all pixels at regular time intervals, frame-based cameras suffer from data redundancy and temporal information loss. By contrast, biologically inspired neuromorphic event cameras, such as the Dynamic Vision Sensor (DVS) [1], transmit asynchronous streams of events generated by individual pixels in response to perceived brightness changes [2]- [4]. Leveraging this sparse yet continuous encoding of visual stimuli allows to deeply simplify the stereo-matching problem.…”