| 摘要 |
A face-image area is decided with comparatively good accuracy. An (N−1)th frame of a subject image α1 and an Nth frame of a subject image α2 are obtained by sensing the image of a subject successively. Face-image detection processing is applied to the frames of the subject images α1 and α2 to detect face-image areas C1, C2 and C3. The face-image area C2 is construed to be linked to the face-image area C1, which is the corresponding face-image area. Since a face-image area that corresponds to the face-image area C3 does not exist, the face-image area C3 is construed as not being linked. The linked face-image area C2 in the Nth frame of the subject image α2 is decided upon as a face-image area. Thus, face areas can be decided with comparatively good accuracy.
|
