| 摘要 |
PURPOSE:To accurately detect the feature amount of a defect by extracting a signal component by allowing an operator matrix to operate on image data, performing labeling by binary-coding and geometrical processing, and then setting a defect presence estimation region. CONSTITUTION:The image of the surface of a disk 1 to be inspected is picked up by one-dimensional CCD camera and then image data are stored in a frame memory 12. An operator matrix is allowed to operate on the original image data to extract a high-frequency signal component. After it is binary-coded by a specific threshold value, an isolated point and a hole are eliminated by inflation and degeneration algorithms. The image is labeled by a geometrical processing and then a region obtained by adding a specific margin width to the fillet diameter of the labeled region is dealt with as a defect presence estimation region D. A defect feature amount is calculated from the original image corresponding to the region D. The feature amount corresponds to the dispersion amount of the image brightness distribution within the region D or the number of picture elements with a brightness value exceeding a specific threshold. The specific part of a signal which changes in analog and the feature amount can be detected accurately. |
