Over the past 30 years, with the advent of computers and digital imaging, many automated systems have been introduced for the purpose of air-void characterization. The majority of the systems employs a contrast-enhancement procedure where a polished cross-section of concrete is darkened with paint, and white powder is forced into the depressions left by air-voids. The system described here follows the same approach and uses a flatbed scanner to collect a single digital image of the entire sample. For all of the systems based on contrast enhancement, the first step is to select a threshold level. Image pixels brighter than the threshold level represent air and image pixels darker than the threshold level represent non-air (i.e., paste or aggregate). Further digital processing steps may be employed but the initial selection of threshold level exerts a strong influence on whether a pixel in the final data set is classified as air or non-air. A systematic approach for threshold determination has been proposed based on an iterative procedure that compares automatically determined air-void parameters to manually determined air-void parameters from a set of training specimens. The calibration procedure finds a single optimum threshold level for the automated system that is to be used for all subsequent analyses. The approach was tested on a population of 88 specimens with manually determined air-void parameters, with the goal of determining an appropriate value for the number of training specimens.