New techniques for patterned wafer inspection based on a model of human preattentive vision

Two new techniques for detecting defects on patterned wafers are presented. The techniques are based on a model of human preattentive visual detection of pattern anomalies. Defect detection is based on comparisons of local to global first order statistics of edge orientation and contrast. The model takes advantage of the fact that preattentive vision operates on the lower frequency components of the visual scene. zthis allows us to sample the image bringing about a significant reduction of data.