ICIP 2022 Paper #1301. Special Session "Neuromorphic and perception-based image acquisition and analysis"
UTILITY AND FEASIBILITY OF A CENTER SURROUND EVENT CAMERA
Tobi Delbruck, Chenghan Li, Rui Graca, Brian Mcreynolds
Inst. of Neuroinformatics, UZH-ETH ZUrich
Zurich, Switzerland.
This paper shows how to build a practical small-pixel precise center surround dynamic vision sensor event camera. The key insight of the paper is that we can use compact poly resistors for the lateral surround network as long as the surround is many pixels.
The addition of the biologically-motivated surround suppresses redundant and noninformative DVS responses to low spatial frequencies and responses to flickering illumination from uniform areas of the scene.
This site accompanies the above accepted and presented 2022 IEEE International Conference in Image Processing paper, posted to https://arxiv.org/abs/2202.13076
Primary Contact: tobi@ini.uzh.ch
Developed by the Sensors Group of the Inst. of Neuroinformatics, Univ. of Zurich and ETH Zurich.
Information about other datasets and tools are on the Sensors Group Datasets.
We would appreciate citing the conference paper when and if it appears. In the meantime, the citation should be
Delbruck, Tobi, Chenghan Li, Rui Graca, and Brian Mcreynolds. “Utility and Feasibility of a Center Surround Event Camera.” In 2022 IEEE International Conference on Image Processing (ICIP), 381–85. IEEE, 2022. https://doi.org/10.1109/ICIP46576.2022.9897354 .
http://arxiv.org/abs/2202.13076
@INPROCEEDINGS{Delbruck2022-hs, title = "Utility and Feasibility of a Center Surround Event Camera", booktitle = "2022 {IEEE} International Conference on Image Processing ({ICIP})", author = "Delbruck, Tobi and Li, Chenghan and Graca, Rui and Mcreynolds, Brian", publisher = "IEEE", pages = "381--385", month = oct, year = 2022, url = "http://dx.doi.org/10.1109/ICIP46576.2022.9897354", keywords = "Resistors;Switches;Vision sensors;Retina;Cameras;Transconductors;Biology;retina;pixel;neuromorphic", issn = "2381-8549, 1522-4880", isbn = "9781665496209, 9781665496216", doi = "10.1109/ICIP46576.2022.9897354"}
Center Surround Dynamic Vision Sensor (CSDVS) pixel circuit that produces spatially filtered brightness change events. It is tiled to connect the lateral resistors to neighboring pixels.
The videos here show results of Figures 3 and 4 in the paper. They were generated using v2e https://github.com/SensorsINI/v2e. See below for more details to reproduce these results.
Fig. 3: Moving and flashing spots
Generated using the spots.py synthetic input module.
The DVS responds over the entire spot and the CSDVS responds only at the edges.
Fig. 4A: Moving gradients
Generated from the gradients.py synthetic input module.
The DVS responds to all temporal changes and the CSDVS responds only to the abrupt edges of the vertical bar.
Fig. 4B: Cloudy sky
Generated from this cloudy-sky source video.
The DVS responds to all changes, but the CSDVS blocks the low-spatial frequency gradients from the clouds.
Fig. 4C: Flickering illumination
Generated from this flicker source video.
The DVS strongly responds to the flicker but the CSDVS blocks the flicker from uniform regions of the scene.