| 摘要 |
<P>PROBLEM TO BE SOLVED: To improve red-eye area detection accuracy, in red-eye detection processing, in accordance with a pattern of red-eye areas frequently generated in an actual photographing environment. <P>SOLUTION: Red-eye candidate areas in a digital photograph image received by an image input receiving section 12 of a red-eye detecting device 10 are identified by a red-eye candidate area finding section 16, using reference data in a memory 14. A display section 20 displays the image with the red-eye candidate areas marked and provisionally corrected. With a specifying section 22, a user can specify an unidentified red-eye area and an erroneously identified area. By learning a characteristic of the unidentified red-eye area and erroneously identified area that are specified, an updating section 24 updates the reference data so as to increase a probability of an area with a characteristic similar to that of the unidentified red-eye area being identified as a red-eye candidate area and decrease the probability of an area with a characteristic similar to the erroneously identified area being identified as a red-eye candidate area. <P>COPYRIGHT: (C)2006,JPO&NCIPI |
