Defect inspection system.

摘要

A defect inspection system in which defect-containing regions are extracted from the binarized image of the inspected material and the kind, pattern and degree of each defect are precisely discriminated. The system comprises an address detector for determining the address of the defect, a binarizer for binarizing the differentiated image and a defect region extractor for extracting a region containing the address of the defect from the binarized image to transform the values of the picture elements in the defect-containing region into the value which indicates the defect. The binarized image is obtained by the binarizer while comparing each picture element of the differentiated image with the third threshold (TH3), which is determined by referring to the first threshold (TH2) calculated from the histogram of the densities of all differentiated image and the second threshold (TH2) calculated from the average density of the picture elements surrounding each target picture element. Also provided is a system in which the binarized image is segmented into plural defect-containing regions so as to make it possible to discriminate the feature of respective defects. A neural network may be used for the discrimination of the feature of the detected defect.