| An important factor limiting the performance of sensor arrays is the nonuniform response of detectors. Fixed pattern noise caused by the nonuniform response of the sensors gives the uncorrected images a white-noise-degraded appearance. With integrated VLSI sensor arrays, nonuniform response becomes an even more challenging problem since all the detectors are fabricated collectively on a single substrate and individual adjustment of detectors is no longer possible. Nonuniformities currently limit performance of infrared (IR) sensors. Nonuniformity correction techniques have been developed and implemented to perform the necessary calibration for most infrared sensing applications. These correction techniques can be divided into two primary categories: (1) reference-based correction using calibrated images on start-up and (2) scene-based techniques that continually recalibrate the sensor for parameter drifts. The problems with the reference-based methods have been well-documented in the literature (parameter drift, expense, and so on). Existing scene-based techniques require complex implementations and suffer from ghosting artifacts in the corrected images.; This research develops and presents several scene-based nonuniformity correction algorithms for both analog and digital implementations. Using clues from neurobiological adaptation, we have developed the constant-statistics (CS) algorithms for nonuniformity correction of imaging arrays. With very low computational complexity, the CS algorithms dramatically reduce the nonuniformity noise of sensor arrays. However, as seen in all scene-based methods, a "ghosting" artifact occurs in the CS calibrated image when an object that does not move enough tends to "burn in" and can remain visible for thousands of images after the object has left the field of view. By modifying the correction update during ghosting situations, we are able to significantly remove the ghosting artifact and improve the overall accuracy of the correction procedure. The digital implementation of the CS algorithms has been applied to both synthetic and real IR images to significantly reduce nonuniformity noise in corrected images. The CS algorithm has also been mapped to analog hardware and designed and fabricated with a 2{dollar}mu{dollar}m CMOS technology. Measured results from the chip show that the analog chip achieves invariance to gain and offset variations of the input signal. |
