摘要 |
<p>The method involves performing geometrical transformation, for each of test pairs of positions of eyes, of a digital image (11) into a normalised image (12N) to coincide the eye positions with a pair of preset reference positions. A part of each normalised image covering a part of a face is compared with a corresponding part of a preset image (13) to obtain a score (SN) representing a rate of resemblance between the parts of the images. The positions of eyes of the test pair associated with better score are declared as effective positions of eyes of the face. An independent claim is also included for a method for progressive correction of position of eyes of a face in a digital image.</p> |