Accuracy Assessment
Note that the previously presented classification algorithms are neither “right” nor “wrong”. For example, someone might object to your use of the simple minimum distance to means classifier by pointing out that your classes have different variability, or a person might object to your use of a Maximum Likelihood Classifier because your data is not normally distributed. These algorithms are neither “right” nor “wrong”, but rather they do a better or worse job in classifying pixels in an image. Accuracy assessment tries to quantify how good of a job was done by the classifier.