| 講演抄録／キーワード |
| 講演名 |
2022-12-12 10:10
［招待講演］Coded Two-Bucket Sensors for Active and Passive Imaging
○Kyros Kutulakos・Rahul Gulve・Navid Sarhangnejad・Roman Genov（U of T） |
| 抄録 |
（和） |
I will discuss our recent progress in designing coded two-bucket sensors, a new family of programmable CMOS image sensors that combine the advantages of coded-exposure and multi-tap imaging. Each pixel in a coded two-bucket (C2B) sensor consists of two charge collection sites?buckets?whose activity can be programmed independently at each pixel and toggled many times during a single exposure. A key performance metric in this family of sensors is their subexposure rate, i.e., the maximum speed at which a pixel?s active bucket can be toggled. We have achieved orders-of-magnitude improvement in this rate over the past six years, with our latest sensor supporting more than 1000 subexposures per video frame readout when operating at 30 fps, as well as the ability to control bucket activities adaptively. I will show our initial experiments with this sensor in a variety of computational imaging applications?simultaneous direct/indirect imaging, snapshot structured-light imaging, snapshot multi-spectral imaging, high-dynamic range imaging?and discuss the broader opportunities this class of sensors opens for active 2D and 3D imaging: conferring robustness to indirect light; enabling operation in strong sunlight; offering improved visibility through skin and other scattering media; and revealing the scattering and refractive properties of everyday objects and materials. |
| （英） |
I will discuss our recent progress in designing coded two-bucket sensors, a new family of programmable CMOS image sensors that combine the advantages of coded-exposure and multi-tap imaging. Each pixel in a coded two-bucket (C2B) sensor consists of two charge collection sites?buckets?whose activity can be programmed independently at each pixel and toggled many times during a single exposure. A key performance metric in this family of sensors is their subexposure rate, i.e., the maximum speed at which a pixel?s active bucket can be toggled. We have achieved orders-of-magnitude improvement in this rate over the past six years, with our latest sensor supporting more than 1000 subexposures per video frame readout when operating at 30 fps, as well as the ability to control bucket activities adaptively. I will show our initial experiments with this sensor in a variety of computational imaging applications?simultaneous direct/indirect imaging, snapshot structured-light imaging, snapshot multi-spectral imaging, high-dynamic range imaging?and discuss the broader opportunities this class of sensors opens for active 2D and 3D imaging: conferring robustness to indirect light; enabling operation in strong sunlight; offering improved visibility through skin and other scattering media; and revealing the scattering and refractive properties of everyday objects and materials. |
| キーワード |
（和） |
coded-exposure imaging / multi-tap sensors / snapshot HDR imaging / snapshot video acquisition / structured-light triangulation / epipolar-only imaging / / |
| （英） |
coded-exposure imaging / multi-tap sensors / snapshot HDR imaging / snapshot video acquisition / structured-light triangulation / epipolar-only imaging / / |
| 文献情報 |
映情学技報, vol. 46, no. 41, IST2022-40, pp. 1-4, 2022年12月. |
| 資料番号 |
IST2022-40 |
| 発行日 |
2022-12-05 (IST) |
| ISSN |
Print edition: ISSN 1342-6893 Online edition: ISSN 2424-1970 |
| PDFダウンロード |
|
|