Defects Detection by Approximation, Multilevel Segmentation and Comparison of Samples

In this paper the approximation algorithm was used twice: to obtain the approximated image and to approximate cumulative histogram. The usage of the comparison formula to transformed images allows to detect coordinates of defects and to measure their intensity marked proportionally to the difference between the etalon and defective images. Approximation by piecewise linear function gives threshold values for multilevel segmentation and defect detection.