RTS Noise Reduction in CMOS Image Sensors Using Correlated Extrema Cancellation

As the CMOS image sensor technology scales, random telegraph signal (RTS) noise from the in-pixel source follower limits performance in low light conditions. This work presents a column-level noise reduction algorithm compatible with a 4T pixel rolling shutter readout. The proposed system can be implemented in conjunction with correlated double sampling (CDS) to address the noisiest RTS pixels in the array. The algorithm exploits order statistics of oversampled noise instead of linear statistics, as used in conventional methodologies. The maxima and minima of noise samples have a significantly higher correlation, achieving superior noise reduction compared to the correlated multiple sampling (CMS) technique. Simulation results indicate up to 12-dB attenuation of output noise compared to CMS and more than 30 dB compared to standard CDS.