| 摘要 |
PROBLEM TO BE SOLVED: To suppress the occurrence of recognition error by unnecessitating the test of correlative value evaluation due to a sample image by dynamically changing a correlative threshold value. SOLUTION: A correlation search is performed as shown by a broken like by turning registered image data 16 to 18 through a template and calculating a correlative value relating two images while successively moving this template 18. At such a time, the correlative value is operated by a matching engine. In this case, the correlative threshold value (prescribed reference value) is dynamically changed. Namely, concerning the moving amount to template 18, it is not moved every time for one pixel but simple correlative operation is performed by rough search at the interval of several pixels. Then, the rough search is performed again with the threshold value of (1/2), for example, to the used correlative threshold value and when no error occurs, precise search is performed with the threshold value of [(1/2)+(1/2<n> )]. Then, the distribution figure of provided correlative values is retrieved so that the optimum correlative threshold value showing the degree of coincidence between images can be determined. |
