A
virtual endoscopy system based on PC (continued) |
We have explored
three primary acceleration techniques listed as follows: |
1. Object and image
space coherences |
There are coherences between
adjoining rays cast from screen and one between adjacent sampled points
along the same ray. So we have presented the algorithm, sampled
points decomposing based ray casting for virtual endoscopy, to
explore the coherences. |
2. Combination of volume rendering
and surface rendering |
Surface rendering
can obtain fast rendering speed using 3D hardware accelerator, but
it loses large detailed information about the 3D datasets after extracting
triangles of iso-surfaces. So it can not acquire high quality images
and can not be applied in VE. However, we can make use of the depths
information produced by surface rendering to accelerate the ray casting
algorithm. So we presented an interactive ray
casting algorithm based on the w-buffer of PC graphics card. |
3. Using Intel IA-32 SIMD Instructions |
Intel IA-32 SIMD
technologies can execute several single floating-point computations
in parallel, and it is very useful for volume rendering. We presented
a novel imaging acceleration method with SIMD technologies, which
can improve rendering speed several times without any specialized
purpose hardware and time-consuming preprocessing. Combining SIMD
technologies with threshold segmentation to reduce large empty samplings,
the rendering speed is further more improved, more than 10 times faster
than brute force ray casting. So we presented a
fast volume rendering algorithm based on Intel SIMD and segmentation
technologies. |
|
Experiments
results |
|
Return
to previous page |