PURPOSE: To generate no fatigue of an eye without dropping the inspection accuracy, by combining a visual inspection and an optical inspection.

CONSTITUTION: When executing the learning, at the time of an inspection for correcting a total value of addition of an adding circuit 12 by collating with the naked eye an image of a printed character part drawn on a video surface of a monitor 19 and an image of a printed character part in a standard pattern, a data signal corresponding to a pattern to be inspected is binary-coded, based as a reference on a set digital value in a digital value setting circuit 16 in case when an added value of an adding circuit 13 has coincided with the total value of addition of the adding circuit 12, an equivalent area is corrected automatically in accordance with variable density of the printed character. Further, binary-coding signals which have been read in a memory 18 and 20 are compared successively, by which the pattern to be inspected is inspected by position information. In this way, a result of inspection being extremely similar to an inspection by the naked eye can be obtained.