An adaptive multiple-reset CMOS wide dynamic range imager for automotive vision applications

Abstract

When vision sensing is part of automotive active safety systems, image information detection performance is critical. Scene elements such as obstacles, lane markings, and signal lights have to be detected reliably under all conditions. Photospace models quantifying scene conditions show that traffic scenes can have extremely high luminance ratios requiring a wide dynamic range (WDR) of more than 120dB.Of the various available WDR technologies, multiple reset CMOS technology is ideally suited for automotive vision due to its high sensitivity, effective highlight compression, adaptability, and low cost. Adaptive highlight compression is achieved by partially resetting the pixel. To preserve mid-tone information, the shape of the imager's response curve is adjusted by applying multiple partial resets of differing magnitudes throughout the exposure period. To reliably capture spatial and color information, the signal-to-noise ratio must exceed a detectability threshold over the entire sensor response curve. Lab measurements and model simulations of multiple-slope CMOS WDR imagers show that the number, timing, and height of reset barriers are critical. Too few barriers results in zones of insufficient local contrast, severely impacting detection capability, and limiting the effective dynamic range.This technology's potential can be best utilized by deploying at least five reset barriers, giving a dynamic range of up to 160dB. Optimum adaptation to the dynamic range of the scene is ensured by model-based algorithms for setting barrier time and voltage. System comparisons with multiple exposure and logarithmic response technologies show the advantages of multiple reset technologies with respect to detection reliability.

A 151 dB High Dynamic Range CMOS Image Sensor Chip Architecture With Tone Mapping Compression Embedded In-Pixel

“…[1], [2]. There are two basic alternatives for acquiring HDR scenes depending on how the pixel data are acquired and coded, namely, either linearly [3], [4] or using some compressive function [5].…”