10.11.2014 Views

Master's Thesis - Studierstube Augmented Reality Project - Graz ...

Master's Thesis - Studierstube Augmented Reality Project - Graz ...

Master's Thesis - Studierstube Augmented Reality Project - Graz ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

2.1 Technical Visualization<br />

relatively easy.<br />

The term ”iso”, from the greek word ισoζ which means ”similar”, is used in this<br />

context to underline what the visualiztion is actually doing: It tries to produce a<br />

triangle mesh by computing an iso-surface with a certain iso-value representing the<br />

same property of the volume. A very popular method to extract a certain property in<br />

a given volume is the so called marching cubes algorithm [Lorensen1987].<br />

The underlying voxel grid is divided into small cubes where always four corners are<br />

built up of four pixels in adjacent slices. The goal is to determine how this cube is<br />

intersected by the desired object’s surface and thus the adjacent slices are intersected.<br />

Choosing an iso-value will result rarely in one of the cubes corners directly; in most cases<br />

the positions of the desired value will not be found without an interpolation along the<br />

intersected edges of the cube. In the majority of cases a surface will intersect three or<br />

more edges of these cubes consisting of vertices with values greater and smaller than the<br />

iso-value. A connection of these intersections will lead to one or at most three triangles<br />

of the desired iso-surface and subsequent processing of all cubes in the volumetric<br />

dataset (”marching”) consequently unfolds the whole surface. Then the algorithm tries<br />

to find the edges which are intersected. This reduces the intersection problem for eight<br />

vertices to 2 8 = 256 cases which can be further generalized by rotations and symmetries<br />

to 15 cases. Finding these intersected edges is rather easy. Every cube’s corner with a<br />

value greater or equal the iso-value will be assigned to one and the others to zero. In<br />

the first case the vertex is behind or on the surface and in the latter it is in front of it.<br />

The creation of a lookup table for these cases enables an efficient build-up; the table<br />

can be accessed via an adopted notation referring to each case by an index created from<br />

a binary eight bit interpretation of the corner’s state. The exact surface intersection on<br />

the edges can then be linear interpolated with the real image values where the vertices<br />

are placed.<br />

Finally the vertex normals for correct shading have to be investigated. These components<br />

can be estimated along the tree coordinate axis by the gradient vectors at the<br />

cube vertices and linearly interpolated depending on the light model.<br />

Two ambiguous cases and certain incompatible triangulated cube-types in the above<br />

described approach have been resolved by introducing new types of intersecting families<br />

[Nielson1991; Montani1994; Chernyaev1995, and others]. Based on this idea many<br />

18

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!