High-dynamic-range binary pixel processing using non-destructive reads and variable oversampling and thresholds

We present a binary image sensor architecture that employs spatial and temporal oversampling and novel signal processing. Our method is based on a mathematical theory of repeated conditional sampling from Poisson distributions. Previous work on binary sampling requires either a pixel reset after each sampling, or feedback proportional to the sampled signal to linearize the sensor response. Our method allows design of sensor response by variation in sampling period length, thresholds and conditional or non-conditional reset. Our sensor architecture supports wide dynamic range, and a low-light sensitivity limited only by the fundamental silicon-process-dependent pixel sensitivity.