Fixed pattern noise removal in cmos imagers across various operational conditions

Abstract

CMOS imagers can possess higher levels of imager noise than their predecessors, CCDs. This noise can be of the form of temporal variation and fixed pattern. The fixed pattern component of this noise can be removed, which is known already in the art. The invention in this disclosure is that proper correction can be developed for all imager conditions (imager integration time and imager temperature) using a single FPN (fixed pattern noise) dark map, a single FPN PRNU (pixel response nonuniformity) map, imager integration time and imager temperature. Without this invention, a dark frame capture and a flat field capture (integrating sphere), are required before every image capture, a practical impossibility in typical picture taking. Further, the estimates of both FPN maps (dark and PRNU) in this invention are improved estimates relative to such captured directly preceding image capture since such have be formed with multiple frame averaging at calibration time, thus removing any temporal noise from these map estimates. These dark FPN and PRNU FPN maps are modified by a scaling and biasing functional with the measured values of integration time and of imager temperature. A second approach is to make the biasing and scaling functions dependant only on mean dark response taken from the imager's dark pixels, at time of capture.