Graphics Reference
In-Depth Information
(Post-deformation)
(Pre-deformation)
(Pre-deformation)
→
→
'
Multiply the
pixels at the same
(
u
,
v
) position
θ
'
i
θ
i
×
P
P
(Post-deformation)
equivalent
Sliding the pixel in the
reflectance function
(method in the paper)
×
→
θ
'
i
equivalent
Rendering by
virtually changing
the light position
Sliding the pixel in the
environment map
(used
Spiderman 2
)
×
P
(a)
(b)
Figure 8.35
Rendering a reflectance field with an environment map amounts to multiplying the envi-
ronment map with the appropriate slice of the reflectance function. A surface deformation
has the effect of rotating the reflectance field, or, equivalently, rotating the environment
map. In producing
Spider Man 2
, this rotation was approximated by a translation.
possible expression, and even if it were possible, the storage requirements would
have been extreme. The method in the “Animatable Facial Reflectance Fields”
paper employs a method of reusing sections of the face that do not change with
certain deformations. The forehead, neck, and eyes are examples. The method,
called “coherent expression culling” in the paper, not only reduces the storage
requirements, it also reduces possible artifacts that may occur as the same nearly
unchanged parts of the face are resampled.
The basic method used in
Spider Man 2
followed the method in the paper,
but there were some differences. To render an image, the incident illumination is
multiplied by the appropriate reflectance function, and the results are summed to
produce the exitant radiance. The incident illumination is contained in an envi-
ronment map. Rotating the environment map is equivalent to rotating a surface
element (this is essentially what is done i
n
Figure 8.32(c)
)
. If the angle of rota-
tion is not too large, the rotation can be approximated by simply translating or
“sliding” the pixels in the environment map (
Figure 8.35
)
. This is faster than
performing the rotation, and better suited to hardware rendering.
Rendering a point from a particular viewing direction on the postdeformation
face is performed as follows. First, the reflectance map corresponding to the
viewing direction with respect to the predeformation surface normal is located.
Rendering is performed using this reflectance function with the environment map
