Database Reference
In-Depth Information
Fig. 7.14
Segmentation of Foreman sequence, (
a
) Frame no. 1 with the user mark, (
b
)the
segmentation result at Frame no. 10
Fig. 7.15
Segmentation of MPEG sequence, (
a
) Frame no. 1 with the user mark, (
b
)the
segmentation result at Frame no. 10
7.5
Segmentation of Face Object Under Illumination
Variations
The video segmentation for face objects is more complex than other pattern
detection problems. This is not only because the faces are non-rigid and have
a high degree of variability in location, color and pose [
202
,
203
], but also due
to the illumination variations. The illumination component of the video frame
varies a great deal, often more than the reflection component. Occlusion and
lighting distortion can change the overall appearance of face objects. A non-uniform
illumination will change the values of gray level distribution of a human face,
and the edge of face will be blurred. Such change will cause the large intra-class
Search WWH ::
Custom Search