摘要 |
Detecting defects in reference images used for optical inspections reduces false defect detections in the test image. Reference images are presumed perfect, but in practice contain defects. Defects in the reference image are detected by measuring the symmetry or randomness of pixels in the area of the suspected defect in both images. Measurements of the pixel intensity ranges, edge smoothness, and total edge slope in the two images are compared to determine if a suspect defect is actually in the reference image.
|