The use of variance and entropic thresholding methods for image segmentation

摘要

In this paper, we investigate methods for the automatic thresholding of X-ray images of food packets in order to determine their suitability for the detection of contaminants (or foreign objects) that may be present within. Single and dual threshold schemes based on variance and entropic measures of the image histogram are applied. A modified entropic measure is proposed which decides whether a single or dual threshold is appropriate dependent on the histogram data. The performance of each of the methods is evaluated and compared using images containing real and computer generated foreign object samples.