Changeset 272 for trunk/VUT/doc/SciReport/mutual.tex
- Timestamp:
- 09/15/05 18:17:46 (19 years ago)
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
trunk/VUT/doc/SciReport/mutual.tex
r269 r272 16 16 This boundary it is based on the current view cell visibility 17 17 classifications. We assume that there is a defined connectivity 18 between the view cells which is the case for both BSP-tree or kD-tree19 based view cells. Then given a PVS consisting of visible view cells18 between the view cells which is the case for both BSP-tree or kD-tree 19 based view cells. Then given a PVS consisting of visible view cells 20 20 (with respect to an object) we can find all the neighbors of the 21 visible view cells which are invisible. In order words a view cell21 visible view cells which are invisible. In order words a view cell 22 22 belongs to this boundary if it is classified invisible and has at 23 23 least one visible neighbor. If all view cells from this boundary are 24 proven invisible, no other view cell (behind this boundary) can be24 proven invisible, no other view cell (behind this boundary) can be 25 25 visible. If the verification determines some view cells as visible, 26 26 the current invisible boundary is extended and the verification is … … 378 378 379 379 For the computation of the maximal error due to the current shaft we 380 assume that one tested region is a view cell, whereas the other is an380 assume that one tested region is a view cell, whereas the other is an 381 381 object bounding box or cell of the spatial hierarchy. The threshold is 382 382 computed as follows: We first triangulate the farthest intersection 383 383 points in the shaft as seen from the view cell side of the shaft. Then 384 for each computed triangle we calculate a point in the view cell which384 for each computed triangle we calculate a point in the view cell which 385 385 maximizes the projected area of the triangle. The conservative 386 386 estimate of the maximal error is then given by a sum of the computed … … 396 396 % The basic idea is the following: 397 397 398 % Use 2 plane parametrization for describing all lines intersecting the object and the view cell. The line sets will be described by aligned rectangles on the two planes which allows to construct convex shafts bounded always by only 4 lines. After determing the initial rectangles bounding the whole line set perform recursive subdivsion as follows:398 % Use 2 plane parametrization for describing all lines intersecting the object and the view cell. The line sets will be described by aligned rectangles on the two planes which allows to construct convex shafts bounded always by only 4 lines. After determing the initial rectangles bounding the whole line set perform recursive subdivsion as follows: 399 399 400 400 % 1. Cast the 4 corner rays and deterine ALL intersections with the occluding objects 401 401 402 % 2. If any ray is unoccluded termite the whole test with VISIBLE and extend the PVS of the object with the new vie cell (and possibly more if the rays goes further). Start the verification for the new viewcells in the occluded layer.402 % 2. If any ray is unoccluded termite the whole test with VISIBLE and extend the PVS of the object with the new view cell (and possibly more if the rays goes further). Start the verification for the new view cells in the occluded layer. 403 403 404 404 % 3. If all 4 rays pierce the same convex mesh, (or mesh triangle for non convex meshes) terminate this branch with INVISIBLE. Note the terminating mesh need not be the front most one. … … 407 407 % Note that here it would be the best to selected those intersections for the rays which have the most similar depths to better estimate the maximal error, i.e. the "depth triangles" need not be formed by the first visible layer. 408 408 409 % 5. If 4 does not terminate subdivide either the view cell or the object rectangle depending on the sample depths and error computed.409 % 5. If 4 does not terminate subdivide either the view cell or the object rectangle depending on the sample depths and error computed. 410 410 411 411 % It is actually quite simple to create a conservative variant of the algorithm: subdivide only to a given depth and avoid test 4. Then the samples can only be terminated by a "STRONG occluder" which hides the whole shaft. This is similar to Dannys method, but here a hierarchical shaft subdivision is performed which should deliver much more accuracy.
Note: See TracChangeset
for help on using the changeset viewer.