Image alignment method, comparative inspection method, and comparative inspection device for comparative inspections

An apparatus for inspecting a specimen, comprising:

• an image detection unit to detect an image of a specimen on which patterns are formed;
• a processor to determine an offset between said detected image and a reference image, and align said detected image and said reference image using a data of said calculated offset, and compare said aligned inspection image and said reference image to detect a defect candidate, and extract a feature of said defect candidate;
• and
• an output unit to output information on said extracted feature of said defect candidate,
• wherein said processor determines said offset between said detected image and a reference image by dividing said detected image and said reference image into plural divisional images, calculating offsets between said plural divisional images of said detected image and said reference image and determining said offset among said calculated offsets by the directions of said patterns.

A method for inspecting a specimen, comprising:

• detecting an image of a specimen on which plural patterns are formed;
• determining offsets between said detected image and a reference image stored in a memory by the directions of said plural patterns;
• aligning said detected image and said reference image using said offsets determined by the direction of said plural patterns;
• and
• comparing said aligned inspection image and said reference image to detect a defect candidate;
• and
• extracting a feature of said defect candidate.

A method for inspecting a specimen, comprising:

• detecting an image of a specimen on which patterns are formed;
• processing said detected image for detecting defect candidate by determining an offset between said detected image and a reference image, aligning said detected image and said reference image using a data of said calculated offset, comparing said aligned inspection image and said reference image to detect a defect candidate, and extracting a feature of said defect candidate;
• and
• outputting information on said extracted feature of said defect candidate,
• wherein in the step of processing, said offset is determined by dividing said detected image and said reference image into plural divisional images, calculating offsets between said plural divided images and reference image and determining said offset among said calculated offsets by the directions of said patterns.