13 May 2015 Hybrid approach to mean-variance and photon transfer measurement
Author Affiliations +
Abstract
This paper presents a hybrid technique for measuring conversion gain that blends spatial and temporal information, allowing users to calculate an accurate conversion gain with little knowledge of sensor defects. It blends a single pixel method with multiple pixel methods. We present measured data from a visible CMOS image sensor using two multiple pixel methods and the hybrid method. Additionally, we provide arguments for validity of the hybrid method. To our knowledge, this is the first report of this technique. Conversion gain (e-/DN) directly relates measured digital numbers (DN) to input-referred electrons (e-) for an image sensor. Conversion gain can be directly measured by considering the sensor under varying illumination states in coordination with Poisson statistics. Typically, there are two approaches: measure a single pixel over time or measure a group of pixels at one point in time after correcting for gain non-uniformity. The plotted statistics from these measurements are called either mean-variance or photon-transfer curves. The measurement of a single pixel is relatively straightforward and requires collection of many consecutive frames to get meaningful statistics not dominated by thermal noise. The data volume for an accurate single-pixel measurement can become unwieldy in terms of number of frames required. This is especially true for large format image sensors. In contrast, the measurement of a group of pixels requires fewer consecutive frames, but needs non-uniformity adjustments to correctly calculate statistics.
© (2015) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Blake C. Jacquot, Brett M. Bolla, Sean Maguire, "Hybrid approach to mean-variance and photon transfer measurement", Proc. SPIE 9481, Image Sensing Technologies: Materials, Devices, Systems, and Applications II, 94810D (13 May 2015); doi: 10.1117/12.2176115; https://doi.org/10.1117/12.2176115
PROCEEDINGS
9 PAGES


SHARE
RELATED CONTENT

Readout circuit design of the retina-like CMOS image sensor
Proceedings of SPIE (February 19 2015)
Towards one trillion positions
Proceedings of SPIE (June 22 2015)
A Local Spatial Correlation Length For Background Modeling
Proceedings of SPIE (January 19 1984)
Hyperspectral image sensor for weed-selective spraying
Proceedings of SPIE (November 11 1999)
Geometric Preprocessing Of Sensor Data Used For Image Matching
Proceedings of SPIE (September 04 1979)

Back to Top