摘要 |
PROBLEM TO BE SOLVED: To enable accurate object recognition even with input image data whose magnification is unknown. SOLUTION: A rectangle is extracted from the input image data (S2) and a rectangle having short- and long-side lengths within permissible ranges is extracted as a candidate rectangle for an object (S3). The magnification R is calculated (S4) from the short- and long-side lengths of the candidate rectangle and the short- and long-side lengths of the rectangle of the object and used to reversely vary the magnification of the image data in the candidate rectangle (S5), and the reversely mangificaton-varied image data are used to perform an object recognizing process (S6).
|