Massive changes to the organisation of PolyVoxCore and PolyVoxYtil.
Also added start of logging capability.
This commit is contained in:
543
library/PolyVoxCore/source/PolyVoxImpl/FastSurfaceExtractor.cpp
Normal file
543
library/PolyVoxCore/source/PolyVoxImpl/FastSurfaceExtractor.cpp
Normal file
@ -0,0 +1,543 @@
|
||||
#pragma region License
|
||||
/******************************************************************************
|
||||
This file is part of the PolyVox library
|
||||
Copyright (C) 2006 David Williams
|
||||
|
||||
This program is free software; you can redistribute it and/or
|
||||
modify it under the terms of the GNU General Public License
|
||||
as published by the Free Software Foundation; either version 2
|
||||
of the License, or (at your option) any later version.
|
||||
|
||||
This program is distributed in the hope that it will be useful,
|
||||
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
GNU General Public License for more details.
|
||||
|
||||
You should have received a copy of the GNU General Public License
|
||||
along with this program; if not, write to the Free Software
|
||||
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
||||
******************************************************************************/
|
||||
#pragma endregion
|
||||
|
||||
#include "PolyVoxImpl/FastSurfaceExtractor.h"
|
||||
|
||||
#include "VolumeIterator.h"
|
||||
#include "IndexedSurfacePatch.h"
|
||||
#include "PolyVoxImpl/MarchingCubesTables.h"
|
||||
#include "SurfaceVertex.h"
|
||||
|
||||
namespace PolyVox
|
||||
{
|
||||
|
||||
void extractFastSurfaceImpl(Volume<uint8_t>* volumeData, Region region, IndexedSurfacePatch* singleMaterialPatch)
|
||||
{
|
||||
singleMaterialPatch->clear();
|
||||
|
||||
//For edge indices
|
||||
int32_t* vertexIndicesX0 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
int32_t* vertexIndicesY0 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
int32_t* vertexIndicesZ0 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
int32_t* vertexIndicesX1 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
int32_t* vertexIndicesY1 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
int32_t* vertexIndicesZ1 = new int32_t[(region.width()+2) * (region.height()+2)];
|
||||
|
||||
//Cell bitmasks
|
||||
uint8_t* bitmask0 = new uint8_t[(region.width()+2) * (region.height()+2)];
|
||||
uint8_t* bitmask1 = new uint8_t[(region.width()+2) * (region.height()+2)];
|
||||
|
||||
//When generating the mesh for a region we actually look one voxel outside it in the
|
||||
// back, bottom, right direction. Protect against access violations by cropping region here
|
||||
Region regVolume = volumeData->getEnclosingRegion();
|
||||
regVolume.setUpperCorner(regVolume.getUpperCorner() - Vector3DInt32(1,1,1));
|
||||
region.cropTo(regVolume);
|
||||
|
||||
//Offset from volume corner
|
||||
const Vector3DFloat offset = static_cast<Vector3DFloat>(region.getLowerCorner());
|
||||
|
||||
//Create a region corresponding to the first slice
|
||||
Region regSlice0(region);
|
||||
regSlice0.setUpperCorner(Vector3DInt32(regSlice0.getUpperCorner().getX(),regSlice0.getUpperCorner().getY(),regSlice0.getLowerCorner().getZ()));
|
||||
|
||||
//Iterator to access the volume data
|
||||
VolumeIterator<uint8_t> volIter(*volumeData);
|
||||
|
||||
//Compute bitmask for initial slice
|
||||
uint32_t uNoOfNonEmptyCellsForSlice0 = computeInitialRoughBitmaskForSlice(volIter, regSlice0, offset, bitmask0);
|
||||
if(uNoOfNonEmptyCellsForSlice0 != 0)
|
||||
{
|
||||
//If there were some non-empty cells then generate initial slice vertices for them
|
||||
generateRoughVerticesForSlice(volIter,regSlice0, offset, bitmask0, singleMaterialPatch, vertexIndicesX0, vertexIndicesY0, vertexIndicesZ0);
|
||||
}
|
||||
|
||||
for(uint32_t uSlice = 0; ((uSlice <= region.depth()-1) && (uSlice + offset.getZ() < region.getUpperCorner().getZ())); ++uSlice)
|
||||
{
|
||||
Region regSlice1(regSlice0);
|
||||
regSlice1.shift(Vector3DInt32(0,0,1));
|
||||
|
||||
uint32_t uNoOfNonEmptyCellsForSlice1 = computeRoughBitmaskForSliceFromPrevious(volIter, regSlice1, offset, bitmask1, bitmask0);
|
||||
|
||||
if(uNoOfNonEmptyCellsForSlice1 != 0)
|
||||
{
|
||||
generateRoughVerticesForSlice(volIter,regSlice1, offset, bitmask1, singleMaterialPatch, vertexIndicesX1, vertexIndicesY1, vertexIndicesZ1);
|
||||
}
|
||||
|
||||
if((uNoOfNonEmptyCellsForSlice0 != 0) || (uNoOfNonEmptyCellsForSlice1 != 0))
|
||||
{
|
||||
generateRoughIndicesForSlice(volIter, regSlice0, singleMaterialPatch, offset, bitmask0, bitmask1, vertexIndicesX0, vertexIndicesY0, vertexIndicesZ0, vertexIndicesX1, vertexIndicesY1, vertexIndicesZ1);
|
||||
}
|
||||
|
||||
std::swap(uNoOfNonEmptyCellsForSlice0, uNoOfNonEmptyCellsForSlice1);
|
||||
std::swap(bitmask0, bitmask1);
|
||||
std::swap(vertexIndicesX0, vertexIndicesX1);
|
||||
std::swap(vertexIndicesY0, vertexIndicesY1);
|
||||
std::swap(vertexIndicesZ0, vertexIndicesZ1);
|
||||
|
||||
regSlice0 = regSlice1;
|
||||
}
|
||||
|
||||
delete[] bitmask0;
|
||||
delete[] bitmask1;
|
||||
delete[] vertexIndicesX0;
|
||||
delete[] vertexIndicesX1;
|
||||
delete[] vertexIndicesY0;
|
||||
delete[] vertexIndicesY1;
|
||||
delete[] vertexIndicesZ0;
|
||||
delete[] vertexIndicesZ1;
|
||||
}
|
||||
|
||||
uint32_t getIndex(uint32_t x, uint32_t y, uint32_t regionWidth)
|
||||
{
|
||||
return x + (y * (regionWidth+1));
|
||||
}
|
||||
|
||||
uint32_t computeInitialRoughBitmaskForSlice(VolumeIterator<uint8_t>& volIter, const Region& regSlice, const Vector3DFloat& offset, uint8_t* bitmask)
|
||||
{
|
||||
uint32_t uNoOfNonEmptyCells = 0;
|
||||
|
||||
//Iterate over each cell in the region
|
||||
volIter.setPosition(regSlice.getLowerCorner().getX(),regSlice.getLowerCorner().getY(), regSlice.getLowerCorner().getZ());
|
||||
volIter.setValidRegion(regSlice);
|
||||
do
|
||||
{
|
||||
//Current position
|
||||
const uint16_t x = volIter.getPosX() - offset.getX();
|
||||
const uint16_t y = volIter.getPosY() - offset.getY();
|
||||
|
||||
//Determine the index into the edge table which tells us which vertices are inside of the surface
|
||||
uint8_t iCubeIndex = 0;
|
||||
|
||||
if((x==0) && (y==0))
|
||||
{
|
||||
const uint8_t v000 = volIter.getVoxel();
|
||||
const uint8_t v100 = volIter.peekVoxel1px0py0pz();
|
||||
const uint8_t v010 = volIter.peekVoxel0px1py0pz();
|
||||
const uint8_t v110 = volIter.peekVoxel1px1py0pz();
|
||||
|
||||
const uint8_t v001 = volIter.peekVoxel0px0py1pz();
|
||||
const uint8_t v101 = volIter.peekVoxel1px0py1pz();
|
||||
const uint8_t v011 = volIter.peekVoxel0px1py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
if (v000 == 0) iCubeIndex |= 1;
|
||||
if (v100 == 0) iCubeIndex |= 2;
|
||||
if (v110 == 0) iCubeIndex |= 4;
|
||||
if (v010 == 0) iCubeIndex |= 8;
|
||||
if (v001 == 0) iCubeIndex |= 16;
|
||||
if (v101 == 0) iCubeIndex |= 32;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
if (v011 == 0) iCubeIndex |= 128;
|
||||
}
|
||||
else if((x>0) && y==0)
|
||||
{
|
||||
const uint8_t v100 = volIter.peekVoxel1px0py0pz();
|
||||
const uint8_t v110 = volIter.peekVoxel1px1py0pz();
|
||||
|
||||
const uint8_t v101 = volIter.peekVoxel1px0py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//x
|
||||
uint8_t iPreviousCubeIndexX = bitmask[getIndex(x-1,y, regSlice.width()+1)];
|
||||
uint8_t srcBit6 = iPreviousCubeIndexX & 64;
|
||||
uint8_t destBit7 = srcBit6 << 1;
|
||||
|
||||
uint8_t srcBit5 = iPreviousCubeIndexX & 32;
|
||||
uint8_t destBit4 = srcBit5 >> 1;
|
||||
|
||||
uint8_t srcBit2 = iPreviousCubeIndexX & 4;
|
||||
uint8_t destBit3 = srcBit2 << 1;
|
||||
|
||||
uint8_t srcBit1 = iPreviousCubeIndexX & 2;
|
||||
uint8_t destBit0 = srcBit1 >> 1;
|
||||
|
||||
iCubeIndex |= destBit0;
|
||||
if (v100 == 0) iCubeIndex |= 2;
|
||||
if (v110 == 0) iCubeIndex |= 4;
|
||||
iCubeIndex |= destBit3;
|
||||
iCubeIndex |= destBit4;
|
||||
if (v101 == 0) iCubeIndex |= 32;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
iCubeIndex |= destBit7;
|
||||
}
|
||||
else if((x==0) && (y>0))
|
||||
{
|
||||
const uint8_t v010 = volIter.peekVoxel0px1py0pz();
|
||||
const uint8_t v110 = volIter.peekVoxel1px1py0pz();
|
||||
|
||||
const uint8_t v011 = volIter.peekVoxel0px1py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//y
|
||||
uint8_t iPreviousCubeIndexY = bitmask[getIndex(x,y-1, regSlice.width()+1)];
|
||||
uint8_t srcBit7 = iPreviousCubeIndexY & 128;
|
||||
uint8_t destBit4 = srcBit7 >> 3;
|
||||
|
||||
uint8_t srcBit6 = iPreviousCubeIndexY & 64;
|
||||
uint8_t destBit5 = srcBit6 >> 1;
|
||||
|
||||
uint8_t srcBit3 = iPreviousCubeIndexY & 8;
|
||||
uint8_t destBit0 = srcBit3 >> 3;
|
||||
|
||||
uint8_t srcBit2 = iPreviousCubeIndexY & 4;
|
||||
uint8_t destBit1 = srcBit2 >> 1;
|
||||
|
||||
iCubeIndex |= destBit0;
|
||||
iCubeIndex |= destBit1;
|
||||
if (v110 == 0) iCubeIndex |= 4;
|
||||
if (v010 == 0) iCubeIndex |= 8;
|
||||
iCubeIndex |= destBit4;
|
||||
iCubeIndex |= destBit5;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
if (v011 == 0) iCubeIndex |= 128;
|
||||
}
|
||||
else
|
||||
{
|
||||
const uint8_t v110 = volIter.peekVoxel1px1py0pz();
|
||||
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//y
|
||||
uint8_t iPreviousCubeIndexY = bitmask[getIndex(x,y-1, regSlice.width()+1)];
|
||||
uint8_t srcBit7 = iPreviousCubeIndexY & 128;
|
||||
uint8_t destBit4 = srcBit7 >> 3;
|
||||
|
||||
uint8_t srcBit6 = iPreviousCubeIndexY & 64;
|
||||
uint8_t destBit5 = srcBit6 >> 1;
|
||||
|
||||
uint8_t srcBit3 = iPreviousCubeIndexY & 8;
|
||||
uint8_t destBit0 = srcBit3 >> 3;
|
||||
|
||||
uint8_t srcBit2 = iPreviousCubeIndexY & 4;
|
||||
uint8_t destBit1 = srcBit2 >> 1;
|
||||
|
||||
//x
|
||||
uint8_t iPreviousCubeIndexX = bitmask[getIndex(x-1,y, regSlice.width()+1)];
|
||||
srcBit6 = iPreviousCubeIndexX & 64;
|
||||
uint8_t destBit7 = srcBit6 << 1;
|
||||
|
||||
srcBit2 = iPreviousCubeIndexX & 4;
|
||||
uint8_t destBit3 = srcBit2 << 1;
|
||||
|
||||
iCubeIndex |= destBit0;
|
||||
iCubeIndex |= destBit1;
|
||||
if (v110 == 0) iCubeIndex |= 4;
|
||||
iCubeIndex |= destBit3;
|
||||
iCubeIndex |= destBit4;
|
||||
iCubeIndex |= destBit5;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
iCubeIndex |= destBit7;
|
||||
}
|
||||
|
||||
//Save the bitmask
|
||||
bitmask[getIndex(x,y, regSlice.width()+1)] = iCubeIndex;
|
||||
|
||||
if(edgeTable[iCubeIndex] != 0)
|
||||
{
|
||||
++uNoOfNonEmptyCells;
|
||||
}
|
||||
|
||||
}while(volIter.moveForwardInRegionXYZ());//For each cell
|
||||
|
||||
return uNoOfNonEmptyCells;
|
||||
}
|
||||
|
||||
uint32_t computeRoughBitmaskForSliceFromPrevious(VolumeIterator<uint8_t>& volIter, const Region& regSlice, const Vector3DFloat& offset, uint8_t* bitmask, uint8_t* previousBitmask)
|
||||
{
|
||||
uint32_t uNoOfNonEmptyCells = 0;
|
||||
|
||||
//Iterate over each cell in the region
|
||||
volIter.setPosition(regSlice.getLowerCorner().getX(),regSlice.getLowerCorner().getY(), regSlice.getLowerCorner().getZ());
|
||||
volIter.setValidRegion(regSlice);
|
||||
do
|
||||
{
|
||||
//Current position
|
||||
const uint16_t x = volIter.getPosX() - offset.getX();
|
||||
const uint16_t y = volIter.getPosY() - offset.getY();
|
||||
|
||||
//Determine the index into the edge table which tells us which vertices are inside of the surface
|
||||
uint8_t iCubeIndex = 0;
|
||||
|
||||
if((x==0) && (y==0))
|
||||
{
|
||||
const uint8_t v001 = volIter.peekVoxel0px0py1pz();
|
||||
const uint8_t v101 = volIter.peekVoxel1px0py1pz();
|
||||
const uint8_t v011 = volIter.peekVoxel0px1py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//z
|
||||
uint8_t iPreviousCubeIndexZ = previousBitmask[getIndex(x,y, regSlice.width()+1)];
|
||||
iCubeIndex = iPreviousCubeIndexZ >> 4;
|
||||
|
||||
if (v001 == 0) iCubeIndex |= 16;
|
||||
if (v101 == 0) iCubeIndex |= 32;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
if (v011 == 0) iCubeIndex |= 128;
|
||||
}
|
||||
else if((x>0) && y==0)
|
||||
{
|
||||
const uint8_t v101 = volIter.peekVoxel1px0py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//z
|
||||
uint8_t iPreviousCubeIndexZ = previousBitmask[getIndex(x,y, regSlice.width()+1)];
|
||||
iCubeIndex = iPreviousCubeIndexZ >> 4;
|
||||
|
||||
//x
|
||||
uint8_t iPreviousCubeIndexX = bitmask[getIndex(x-1,y, regSlice.width()+1)];
|
||||
uint8_t srcBit6 = iPreviousCubeIndexX & 64;
|
||||
uint8_t destBit7 = srcBit6 << 1;
|
||||
|
||||
uint8_t srcBit5 = iPreviousCubeIndexX & 32;
|
||||
uint8_t destBit4 = srcBit5 >> 1;
|
||||
|
||||
iCubeIndex |= destBit4;
|
||||
if (v101 == 0) iCubeIndex |= 32;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
iCubeIndex |= destBit7;
|
||||
}
|
||||
else if((x==0) && (y>0))
|
||||
{
|
||||
const uint8_t v011 = volIter.peekVoxel0px1py1pz();
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//z
|
||||
uint8_t iPreviousCubeIndexZ = previousBitmask[getIndex(x,y, regSlice.width()+1)];
|
||||
iCubeIndex = iPreviousCubeIndexZ >> 4;
|
||||
|
||||
//y
|
||||
uint8_t iPreviousCubeIndexY = bitmask[getIndex(x,y-1, regSlice.width()+1)];
|
||||
uint8_t srcBit7 = iPreviousCubeIndexY & 128;
|
||||
uint8_t destBit4 = srcBit7 >> 3;
|
||||
|
||||
uint8_t srcBit6 = iPreviousCubeIndexY & 64;
|
||||
uint8_t destBit5 = srcBit6 >> 1;
|
||||
|
||||
iCubeIndex |= destBit4;
|
||||
iCubeIndex |= destBit5;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
if (v011 == 0) iCubeIndex |= 128;
|
||||
}
|
||||
else
|
||||
{
|
||||
const uint8_t v111 = volIter.peekVoxel1px1py1pz();
|
||||
|
||||
//z
|
||||
uint8_t iPreviousCubeIndexZ = previousBitmask[getIndex(x,y, regSlice.width()+1)];
|
||||
iCubeIndex = iPreviousCubeIndexZ >> 4;
|
||||
|
||||
//y
|
||||
uint8_t iPreviousCubeIndexY = bitmask[getIndex(x,y-1, regSlice.width()+1)];
|
||||
uint8_t srcBit7 = iPreviousCubeIndexY & 128;
|
||||
uint8_t destBit4 = srcBit7 >> 3;
|
||||
|
||||
uint8_t srcBit6 = iPreviousCubeIndexY & 64;
|
||||
uint8_t destBit5 = srcBit6 >> 1;
|
||||
|
||||
//x
|
||||
uint8_t iPreviousCubeIndexX = bitmask[getIndex(x-1,y, regSlice.width()+1)];
|
||||
srcBit6 = iPreviousCubeIndexX & 64;
|
||||
uint8_t destBit7 = srcBit6 << 1;
|
||||
|
||||
iCubeIndex |= destBit4;
|
||||
iCubeIndex |= destBit5;
|
||||
if (v111 == 0) iCubeIndex |= 64;
|
||||
iCubeIndex |= destBit7;
|
||||
}
|
||||
|
||||
//Save the bitmask
|
||||
bitmask[getIndex(x,y, regSlice.width()+1)] = iCubeIndex;
|
||||
|
||||
if(edgeTable[iCubeIndex] != 0)
|
||||
{
|
||||
++uNoOfNonEmptyCells;
|
||||
}
|
||||
|
||||
}while(volIter.moveForwardInRegionXYZ());//For each cell
|
||||
|
||||
return uNoOfNonEmptyCells;
|
||||
}
|
||||
|
||||
void generateRoughVerticesForSlice(VolumeIterator<uint8_t>& volIter, Region& regSlice, const Vector3DFloat& offset, uint8_t* bitmask, IndexedSurfacePatch* singleMaterialPatch,int32_t vertexIndicesX[],int32_t vertexIndicesY[],int32_t vertexIndicesZ[])
|
||||
{
|
||||
//Iterate over each cell in the region
|
||||
volIter.setPosition(regSlice.getLowerCorner().getX(),regSlice.getLowerCorner().getY(), regSlice.getLowerCorner().getZ());
|
||||
volIter.setValidRegion(regSlice);
|
||||
//while(volIter.moveForwardInRegionXYZ())
|
||||
do
|
||||
{
|
||||
//Current position
|
||||
const uint16_t x = volIter.getPosX() - offset.getX();
|
||||
const uint16_t y = volIter.getPosY() - offset.getY();
|
||||
const uint16_t z = volIter.getPosZ() - offset.getZ();
|
||||
|
||||
const uint8_t v000 = volIter.getVoxel();
|
||||
|
||||
//Determine the index into the edge table which tells us which vertices are inside of the surface
|
||||
uint8_t iCubeIndex = bitmask[getIndex(x,y, regSlice.width()+1)];
|
||||
|
||||
/* Cube is entirely in/out of the surface */
|
||||
if (edgeTable[iCubeIndex] == 0)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Find the vertices where the surface intersects the cube */
|
||||
if (edgeTable[iCubeIndex] & 1)
|
||||
{
|
||||
if((x + offset.getX()) != regSlice.getUpperCorner().getX())
|
||||
{
|
||||
const uint8_t v100 = volIter.peekVoxel1px0py0pz();
|
||||
const Vector3DFloat v3dPosition(x + 0.5f, y, z);
|
||||
const Vector3DFloat v3dNormal(v000 > v100 ? 1.0f : -1.0f, 0.0f, 0.0f);
|
||||
const uint8_t uMaterial = v000 | v100; //Because one of these is 0, the or operation takes the max.
|
||||
const SurfaceVertex surfaceVertex(v3dPosition, v3dNormal, uMaterial);
|
||||
uint32_t uLastVertexIndex = singleMaterialPatch->addVertex(surfaceVertex);
|
||||
vertexIndicesX[getIndex(x,y, regSlice.width()+1)] = uLastVertexIndex;
|
||||
}
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 8)
|
||||
{
|
||||
if((y + offset.getY()) != regSlice.getUpperCorner().getY())
|
||||
{
|
||||
const uint8_t v010 = volIter.peekVoxel0px1py0pz();
|
||||
const Vector3DFloat v3dPosition(x, y + 0.5f, z);
|
||||
const Vector3DFloat v3dNormal(0.0f, v000 > v010 ? 1.0f : -1.0f, 0.0f);
|
||||
const uint8_t uMaterial = v000 | v010;
|
||||
SurfaceVertex surfaceVertex(v3dPosition, v3dNormal, uMaterial);
|
||||
uint32_t uLastVertexIndex = singleMaterialPatch->addVertex(surfaceVertex);
|
||||
vertexIndicesY[getIndex(x,y, regSlice.width()+1)] = uLastVertexIndex;
|
||||
}
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 256)
|
||||
{
|
||||
//if((z + offset.getZ()) != upperCorner.getZ())
|
||||
{
|
||||
const uint8_t v001 = volIter.peekVoxel0px0py1pz();
|
||||
const Vector3DFloat v3dPosition(x, y, z + 0.5f);
|
||||
const Vector3DFloat v3dNormal(0.0f, 0.0f, v000 > v001 ? 1.0f : -1.0f);
|
||||
const uint8_t uMaterial = v000 | v001;
|
||||
SurfaceVertex surfaceVertex(v3dPosition, v3dNormal, uMaterial);
|
||||
uint32_t uLastVertexIndex = singleMaterialPatch->addVertex(surfaceVertex);
|
||||
vertexIndicesZ[getIndex(x,y, regSlice.width()+1)] = uLastVertexIndex;
|
||||
}
|
||||
}
|
||||
}while(volIter.moveForwardInRegionXYZ());//For each cell
|
||||
}
|
||||
|
||||
void generateRoughIndicesForSlice(VolumeIterator<uint8_t>& volIter, const Region& regSlice, IndexedSurfacePatch* singleMaterialPatch, const Vector3DFloat& offset, uint8_t* bitmask0, uint8_t* bitmask1, int32_t vertexIndicesX0[],int32_t vertexIndicesY0[],int32_t vertexIndicesZ0[], int32_t vertexIndicesX1[],int32_t vertexIndicesY1[],int32_t vertexIndicesZ1[])
|
||||
{
|
||||
uint32_t indlist[12];
|
||||
|
||||
Region regCroppedSlice(regSlice);
|
||||
regCroppedSlice.setUpperCorner(regCroppedSlice.getUpperCorner() - Vector3DInt32(1,1,0));
|
||||
|
||||
volIter.setPosition(regCroppedSlice.getLowerCorner().getX(),regCroppedSlice.getLowerCorner().getY(), regCroppedSlice.getLowerCorner().getZ());
|
||||
volIter.setValidRegion(regCroppedSlice);
|
||||
do
|
||||
{
|
||||
//Current position
|
||||
const uint16_t x = volIter.getPosX() - offset.getX();
|
||||
const uint16_t y = volIter.getPosY() - offset.getY();
|
||||
const uint16_t z = volIter.getPosZ() - offset.getZ();
|
||||
|
||||
//Determine the index into the edge table which tells us which vertices are inside of the surface
|
||||
uint8_t iCubeIndex = bitmask0[getIndex(x,y, regSlice.width()+1)];
|
||||
|
||||
/* Cube is entirely in/out of the surface */
|
||||
if (edgeTable[iCubeIndex] == 0)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
/* Find the vertices where the surface intersects the cube */
|
||||
if (edgeTable[iCubeIndex] & 1)
|
||||
{
|
||||
indlist[0] = vertexIndicesX0[getIndex(x,y, regSlice.width()+1)];
|
||||
assert(indlist[0] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 2)
|
||||
{
|
||||
indlist[1] = vertexIndicesY0[getIndex(x+1,y, regSlice.width()+1)];
|
||||
assert(indlist[1] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 4)
|
||||
{
|
||||
indlist[2] = vertexIndicesX0[getIndex(x,y+1, regSlice.width()+1)];
|
||||
assert(indlist[2] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 8)
|
||||
{
|
||||
indlist[3] = vertexIndicesY0[getIndex(x,y, regSlice.width()+1)];
|
||||
assert(indlist[3] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 16)
|
||||
{
|
||||
indlist[4] = vertexIndicesX1[getIndex(x,y, regSlice.width()+1)];
|
||||
assert(indlist[4] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 32)
|
||||
{
|
||||
indlist[5] = vertexIndicesY1[getIndex(x+1,y, regSlice.width()+1)];
|
||||
assert(indlist[5] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 64)
|
||||
{
|
||||
indlist[6] = vertexIndicesX1[getIndex(x,y+1, regSlice.width()+1)];
|
||||
assert(indlist[6] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 128)
|
||||
{
|
||||
indlist[7] = vertexIndicesY1[getIndex(x,y, regSlice.width()+1)];
|
||||
assert(indlist[7] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 256)
|
||||
{
|
||||
indlist[8] = vertexIndicesZ0[getIndex(x,y, regSlice.width()+1)];
|
||||
assert(indlist[8] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 512)
|
||||
{
|
||||
indlist[9] = vertexIndicesZ0[getIndex(x+1,y, regSlice.width()+1)];
|
||||
assert(indlist[9] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 1024)
|
||||
{
|
||||
indlist[10] = vertexIndicesZ0[getIndex(x+1,y+1, regSlice.width()+1)];
|
||||
assert(indlist[10] != -1);
|
||||
}
|
||||
if (edgeTable[iCubeIndex] & 2048)
|
||||
{
|
||||
indlist[11] = vertexIndicesZ0[getIndex(x,y+1, regSlice.width()+1)];
|
||||
assert(indlist[11] != -1);
|
||||
}
|
||||
|
||||
for (int i=0;triTable[iCubeIndex][i]!=-1;i+=3)
|
||||
{
|
||||
uint32_t ind0 = indlist[triTable[iCubeIndex][i ]];
|
||||
uint32_t ind1 = indlist[triTable[iCubeIndex][i+1]];
|
||||
uint32_t ind2 = indlist[triTable[iCubeIndex][i+2]];
|
||||
|
||||
singleMaterialPatch->addTriangle(ind0, ind1, ind2);
|
||||
}//For each triangle
|
||||
}while(volIter.moveForwardInRegionXYZ());//For each cell
|
||||
}
|
||||
}
|
Reference in New Issue
Block a user