Abstract
Application of a conventional image sensor to ITS, security monitoring, and so on is hampered by its small dynamic range, 60-70dB, due to a single fixed exposure time for all pixels. To accommodate large contrast of brightness existed in an image for such applications, sensors with over 100 dB dynamic range are necessary. In this paper we propose a technique that extends dynamic range of CMOS image sensors by adjusting both the exposure time and readout circuit gain on a pixel-by-pixel basis, and expressing the input illumination in floating-point representation. From the HSPICE simulation we obtained 90 dB dynamic range and 9-bit linearity for 6-step exposure time and 3-step readout circuit gain. The result suggests that further splitting the exposure time into 8 steps we are able to achieve dynamic range beyond 100 dB.
