A novel computer vision algorithm is presented which is designed for measurement of skeletal growth. The overall system was fully developed on an IBM-PC. It has the advantages of eliminating subjective variations resulting from the instability of human visual judgment and detecting abnormal skeletal growth in children in a short period of time. Two standard radiographs, taken from the same child at two time instants, were used in these analyses. The system consists of three stages: preprocessing, segmentation, and measurement. In the first stage, histogram equalization and autothresholding techniques were utilized to normalize the input radiograph and stretch the contrast. The segmentation determines the contour of each bone of interest. The measurement stage provides the necessary information for clinical diagnosis.