A Psychometric Approach to Edge Detector Calibration in Grey-scale Images

An edge detection algorithm is a filter which significantly reduces the amount of information present in an image such that only high frequency changes in either range or intensity are visible in the resulting image. In order to perform effective edge detection the user must have a clear idea of the frequency above which an edge will be identified. In grey-scale images, edges represent sudden or high-frequency changes in the grey-scale, also referred to as intensity or luminence, level of an image. In practice, what is considered a "high-frequency change" is dependent upon the purpose for which the edge detector has been selected. In this paper it is proposed that to determine the minimum grey-level threshold and per-pixel intensity change at which the user deems a "true" step edge exists, psychometric testing is required. This information is then used to calibrate common edge detection methods which are subsequently used to filter a series of common grey-scale images