Adaptive and robust detection of large defects and image misalignment

A non-transitory computer-readable medium, storing program instructions executable on a computer system for performing a computer-implemented method for detecting defects on a specimen, wherein the computer-implemented method comprises:

• generating one or more statistics for each of multiple tiles in difference image frames generated for a job of images for a specimen generated by an inspection subsystem, wherein the one or more statistics comprise a mean for each of the multiple tiles;
• identifying outlier tiles in the multiple tiles based on the one or more statistics generated for each of the multiple tiles, wherein identifying the outlier tiles comprises generating a histogram of the mean for each of the multiple tiles and identifying the outlier tiles based on the histogram of the mean;
• designating the multiple tiles not identified as the outlier tiles based on the histogram of the mean as remaining tiles, wherein the one or more statistics further comprise range and standard deviation for each of the remaining tiles in a first of the difference image frames, and wherein identifying the outlier tiles further comprises generating range and standard deviation histograms for the remaining tiles in the first of the difference image frames and identifying outlier remaining tiles in the first of the difference image frames based on the range and standard deviation histograms as additional outlier tiles;
• determining one or more defect detection parameters of a defect detection method for each of the difference image frames based on the identified outlier tiles;
• and
• detecting defects on the specimen by applying the defect detection method to the difference image frames with the determined one or more defect detection parameters.

A computer-implemented method for detecting defects on a specimen, comprising:

• generating one or more statistics for each of multiple tiles in difference image frames generated for a job of images for a specimen generated by an inspection subsystem, wherein the one or more statistics comprise a mean for each of the multiple tiles;
• identifying outlier tiles in the multiple tiles based on the one or more statistics generated for each of the multiple tiles, wherein identifying the outlier tiles comprises generating a histogram of the mean for each of the multiple tiles and identifying the outlier tiles based on the histogram of the mean;
• designating the multiple tiles not identified as the outlier tiles based on the histogram of the mean as remaining tiles, wherein the one or more statistics further comprise range and standard deviation for each of the remaining tiles in a first of the difference image frames, and wherein identifying the outlier tiles further comprises generating range and standard deviation histograms for the remaining tiles in the first of the difference image frames and identifying outlier remaining tiles in the first of the difference image frames based on the range and standard deviation histograms as additional outlier tiles;
• determining one or more defect detection parameters of a defect detection method for each of the difference image frames based on the identified outlier tiles;
• and
• detecting defects on the specimen by applying the defect detection method to the difference image frames with the determined one or more defect detection parameters, wherein said generating, identifying, designating, determining, and detecting are performed by a computer subsystem.