Performance indicators for the statistical evaluation of digital image classifications

A statistical procedure is proposed to evaluate the algorithms for the numerical classification of images. The approach is based on the derivation of performance indicators from measurements of signature separability and thresholding analysis. Although these measurements are not new in image processing techniques, they are used in this study in an original way for the comparison of outputs resulting from different classification criteria. The theoretical description of the method suggested is followed by its practical application to a case-study for mapping crop coefficients in an irrigation district