Abstract
In the enhancement/thresholding method of edge detection, the gradient values of pixels exceeding a certain threshold are designated as edge pixels. However, selecting a threshold has commonly been performed through ad hoc measures. This paper describes a method for automatically selecting a threshold using a 5-parameter model. The model is based on the weighted sum of two gamma density functions corresponding to edge and ncln-edge pixels. A variety of statistical and fitting methods for finding the model parameters were evaluated by comparing their computed thresholds to perceptual thresholds determined by subjects for 16 different images. The performance of the model was also analyzed under different noise levels. Index Terms: Automatic thresholding, edge detection, gradient, Sobel operator, statistical classification.
Keywords
Automatic thresholding, edge detection, gradient, Sobel operator, statistical
Date of this Version
August 1996