The effects of sampling variation on image recognition systems

Presents the results of preliminary studies on the effects of sampling variation on the performance of image recognition systems. The authors show that random perturbations in the sampling process can result in substantial deviations in measured features from an input image. Provided certain conditions are met, this variation can be exploited to improve recognition accuracy. The authors describe three experiments that demonstrate this approach in the context of step edge detection, simple pattern recognition, and OCR