Graphics Reference
In-Depth Information
Figure 8.69
Overview of capturing process. At each position of the turntable, the camera is shifted
so the plane of focus shifts through the image. The images labeled (a) are captured pho-
tographs; those labeled (b) are the result of combing in the row marked in black from all
camera shifts. (From [Jakob et al. 09]
c
2009 ACM, Inc. Included here by permission.)
their depth precisely. Both issues can be resolved by using multiple sweeps at
different orientations of the turntable, which provides multiple observations of
the clearly focused hair fibers from different camera angles.
Figure 8.69 illustrates two different camera positions and turntable orienta-
tions and the resulting images. The hair-fiber geometry is represented in a global
(
-coordinate system chosen such that the
y
-axis coincides with the rotational
axis. The plane of focus in a captured image is a plane parallel to
y
but rotated
by the angle
x
,
y
,
z
)
of the turntable and displaced by a distance
d
according to the
translation of the camera. A pixel in a captured image is given by coordinates
(
θ
u
,
y
)
,where
u
is the horizontal position in the plane of focus. The global
(
x
,
y
,
z
)
-
coordinates of a focused pixel at position
(
u
,
y
)
in an image can be obtained by
rotating
and translating by
d
. Figure 8.69(a) shows a pair of captured
images. Part (b) of the figure contains images constructed by stacking
u
-slices for
afixed
y
(shown at the left) for varying values of the camera distance
d
.
Figure 8.69 illustrates two different camera positions and turntable orienta-
tions and the resulting images. The images in part (a) are the captured images for
the geometry shown at the left. Each captured image is processed using a
ridge
detection
algorithm to locate the focused hair fibers in the image.
18
This process
takes a large number of filters (the paper uses rotated oblong filters known as
Ga-
(
u
,
y
)
by
θ
18
Similar feature-detection algorithms are common in CG applications, and are fundamental in
computer vision. One example of ridge detection is the matching of fingerprints.
