Local Estimation of the Uniform Error Threshold

The theory behind selection of the ``uniform error'' threshold which equalizes the probability of misclassification in an image containing two classes is presented. It is shown how this threshold can be estimated using local operations. Some examples and possible extensions are considered.