Abstract

The detection of luminance discontinuities often produces edge contours that are perceptually unimportant for the analysis of the scene. To try to remove from the image those "irrelevant" edges is not a simple task due to the fact that this procedure usually involves the choice of rather arbitrary thresholds to be applied to certain measures of the local luminance discontinuity. This threshold is only rarely "image indipendent" and very often its optimal choice is linked to the particular image being processed. For this reason, the most effective thresholding algorithms relay on some information about the image which is either known a-priori or measured from the image itself. Whenever the threshold is not fixed a-priori but it is determined in some manner by the algorithm, the thresholding technique is called adaptive.

© 1985 Optical Society of America