Game Development Reference
In-Depth Information
Figure 15.4 On the left, surface rendered by diffuse texture. On the right, the same material rendered
by BTF
However, the step of compression (which is optional in 2D textures approach) becomes
mandatory in case of BTFs. In fact, the volume of data to be processed is too high to
expect manipulation of a non-compressed BTF in an acceptable calculation time.
Acquisition of BTF
Acquiring a BTF consists of measuring an interval of 6 dimensions of the transfer
function of the light on a surface S :
BTF λ ( x , θ i , φ i , θ r , φ r ):
=
BSSRDF λ ( x i , x r , θ i , φ i , θ r , φ r ) dx i
S
One of the methods for acquiring BTF consists of capturing the geometry of a small
object and its reflection properties, parameterised on the surface, i.e. on spatially
varying BRDF. Furukawa et al. (2002) and Lensch et al. (2003) use a laser to acquire the
geometry of the object and capture several images by changing the directions of vision
and light. Furukawa et al. map the images on the triangles of the model and compress
the data re-parameterised by using a tensor. Lensch et al. use a grouping procedure to
compensate for the insufficient samples. However, such approaches provide a model
of the appearance that depends on the captured geometry which cannot be re-used for
other objects.
To compensate this limitation, an alternative to the previously mentioned
approaches is capturing the appearance of an opaque material, regardless of its geo-
metry, by capturing its BTF on a plane sample. The BTF is acquired by capturing
several images of the sample, as per the different directions of view and light. The
images acquired in this manner can be used instead of 2D textures and mapped on
arbitrary geometries.
Compression of BTF
A measured BTF is a collection of certain images that are difficult to use for the
synthesis of texture and rendering (a BTF of 256
256 pixels can require several
hundred Mb memory). From the perspective of a real time rendering, it is essential to
×
Search Nedrilad ::




Custom Search