Image Reconstruction from an Event Camera
This repository contains code for brightness image reconstruction from a rotating event camera. For simplicity, we assume that the orientation of the camera is given, e.g., it is provided by a pose-tracking algorithm or by ground truth camera poses. The algorithm uses a per-pixel Extended Kalman Filter (EKF) approach to estimate the brightness image or gradient map that caused the events.
Disclaimer and License
This code has been tested with MATLAB R2017a on Ubuntu 16.04. This is research code, expect that it changes often and any fitness for a particular purpose is disclaimed. The source code is released under a GNU General Public License (GPL).
Please run the file
This script reads a file of events and a file of camera rotations and produces a panoramic image with the reconstructed brightness that caused the events. See the example provided.
When running the script, a figure will emerge showing the evolution of the reconstructed brightness map as events are being processed:
The above reconstructed map is displayed in logarithmic scale since the event camera known as Dynamic Vision Sensor (DVS) operates on logarithmic brightness.
Some details on the EKF approach
Two possible measurement functions are provided for the EKF correction step:
- the event rate (the reciprocal of the time between two consecutive events within the same pixel), which gives an explicit EKF. See references  and  below.
- the brightness contrast (the quantity thresholded by the event camera to generate events), which gives an implicit EKF. See reference  below.
A sample of the output produced by the algorithm can be found in this folder.
EKF output: estimated gradient map, which is "integrated" using Poisson image reconstruction to yield the above brightness map. The following image shows the magnitude and direction of the gradient map, combined: color represents direction, whereas saturation represents magnitude.
If you use this code in an academic context, please cite the following references:
H. Kim, A. Handa, R. Benosman, S.-H. Ieng, A.J. Davison, Simultaneous Mosaicing and Tracking with an Event Camera. British Machine Vision Conference, 2014.
H. Rebecq, T. Horstschaefer, G. Gallego, D. Scaramuzza, EVO: A Geometric Approach to Event-based 6-DOF Parallel Tracking and Mapping in Real-time. IEEE Robotics and Automation Letters (RA-L), Vol. 2, Issue 2, pp. 593-600, Apr. 2017.
G. Gallego, C. Forster, E. Mueggler, D. Scaramuzza, Event-based Camera Pose Tracking using a Generative Event Model. arXiv:1510.01972, 2015.