Abstract

Digital image processing is increasingly influential in the inspection of industrial products. For microelectronic industry, the pattern examination required for photolithographic masks can be facilitated by the image segmentation. This work demonstrates that arrays of rectangular chromium films on photolithographic masks are effectively detected by the gradient-based edge detection. The RGB micrographs are successively converted to grayscale and binary images. The closing morphology algorithm is then applied to reduce noise in the images before the application of the Canny edge detector. After subsequent steps of fill area and remove small objects, the area of each rectangle can then be computed. This image processing procedure gives rise to less than 0.8% difference from the one-by-one inspection and the ratios of rectangular areas confirmed the high accuracy. The standard deviations in number of pixels averaged from 10-17 rectangles in three images are 0.81-1.89% indicating the precision of photolithography and image processing. The brightness used in this optical microscopy has no apparent effect but the precision and accuracy are significantly reduced in the image with low magnification and different illumination.