| 摘要 |
<P>PROBLEM TO BE SOLVED: To enhance the precision and effect of recognition of an object in an image to be recognized. <P>SOLUTION: A handwritten input image is compared as a template image with an image to be recognized. Namely, handwritten image data from a coordinate indication input device are inputted and normalized and the handwriting input data or normalized handwriting image data are processed to a gray scale or into a luminance signal and quantized to obtain a template image, which is compared with the image to be recognized which is separately inputted. Even when a user draws spectacles with a mouse etc., a person wearing spectacles in an image can be specified based upon the spectacles as feature points through the above processing. <P>COPYRIGHT: (C)2004,JPO |
