Class contrast: A robust method for classification of heterogeneous zones in solids applied to thermal NDT images

Contrast methods are based on evaluation of the "distance" between the values of a representative parameter of the phenomenon affecting the zones to be classified. The selection of a particular value of the reference by means of an operator is based on the fact that the operator has an a priori knowledge about the specimen. The Class Contrast method presented in this paper overcomes these limitations and works on the basis of 2 guidelines: 1) contrast methods are differential methods, therefore final result should not depend on the arbitrary selection of the reference points; 2) classes are represented by a certain number of points, all affected by the same noise, and all these points should contribute to increase the final SNR. Class Contrast has been successfully applied to thermal NDT on composites. Results are reported together with comparison with results from application of classical thermal contrast method.