CMOS Image Sensor Enhancements for Low-Light Imaging

Abstract

Low-light imaging poses significant challenges across domains like surveillance, mobile photography, and scientific instrumentation. CMOS image sensors have seen major enhancements to address low-light performance, particularly through innovations up to 2017. Key advancements include backside illumination (BSI) and stacked CMOS architectures that improve quantum efficiency and light capture; pixel binning, which aggregates signals across adjacent pixels to increase sensitivity; advanced readout techniques such as digital correlated double sampling (CDS) and charge-domain CDS to suppress noise; high dynamic range (HDR) pixel designs like LOFIC for preserving detail in dim scenes; low-noise pixel circuits featuring in-pixel amplification and bootstrapping for reduced read noise; and novel pixel constructs such as Geiger-mode avalanche photodiodes (GM-APDs) for photon-level detection. This paper synthesizes these pre-2018 innovations, comparing their impact on signal-to-noise ratio, sensitivity, and dynamic range. We analyze trade-offs including complexity, area, and power, and propose an integrated roadmap for next-generation low-light CMOS sensors. The study provides practical design guidance for applications requiring both high sensitivity and performance in minimal illumination.