Key elements for machine vision are the intra-scene dynamic range of the optical front-end, and a data representation that is as independent as possible from the illumination level. Furthermore, combining an optical front-end and a processor on the same chip enables a single-chip vision system to perform image acquisition, analysis and decision-making. Approaches that enable high dynamic range are logarithmic imagers [1] and lin-log imagers [2], but they suffer from poor fixed-pattern noise (FPN) performance [1,2] and non-uniform transfer functions [2] due to the combination of linear and logarithmic domains. While multiple-exposure imagers [3] solve the FPN problem, they require post-processing to combine several frame captures. Finally the dynamic range of time-domain logarithmic imagers with fixed reference voltages [4] is limited by the maximum allowable exposure time.

No abstract

No abstract

No abstract

No abstract