A comparative evaluation of image background subtraction techniques
Ankit K. Jain, Daniel V. Rabinkin · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2007
Many applications generate digital image sequences using a lens system, a light-transducing pixel array, and sample-and-hold A/D electronics. Non-uniformity Correction (NUC) is an image processing operation that is often required in such systems. The NUC operation subtracts the background from a temporally evolving image on a pixel-by-pixel basis. Background estimation must be robust against image features, electronic glitches, and sudden changes in background. The NUC is often applied in conjunction with a mechanical image dithering that enables separation of the image from the static additive background. In real-time applications with large pixe-count focal planes, the NUC must process large data bandwidths. The NUC algorithm must be operationally efficient to process data in real time. This paper examines a number of NUC algorithms. It defines a set of performance metrics and evaluates algorithm performance in terms of trades between these metrics and processing cost. It provides a guide for selecting an appropriate NUC algorithm based on operating conditions and available compute resources.