Fixed ansisotropic scaling problem with pixel scaling/small feature culling.

This commit is contained in:
Robert Osfield
2002-11-12 15:50:39 +00:00
parent f07dc35976
commit b468d2f939

View File

@@ -65,22 +65,31 @@ void CullStack::pushCullingSet()
// here we do it in short hand with the knowledge of how the window matrix is formed
// note P23,P33 are multiplied by an implicit 1 which would come from the window matrix.
// Robert Osfield, June 2002.
// scaling for horizontal pixels
float P00 = P(0,0)*W.width()*0.5f;
float P20 = P(2,0)*W.width()*0.5f + P(2,3)*W.width()*0.5f;
osg::Vec3 scale(M(0,0)*P00 + M(0,2)*P20,
M(1,0)*P00 + M(1,2)*P20,
M(2,0)*P00 + M(2,2)*P20);
float P20_00 = P(2,0)*W.width()*0.5f + P(2,3)*W.width()*0.5f;
osg::Vec3 scale_00(M(0,0)*P00 + M(0,2)*P20_00,
M(1,0)*P00 + M(1,2)*P20_00,
M(2,0)*P00 + M(2,2)*P20_00);
// scaling for vertical pixels
float P10 = P(1,1)*W.width()*0.5f;
float P20_10 = P(2,1)*W.width()*0.5f + P(2,3)*W.width()*0.5f;
osg::Vec3 scale_10(M(0,1)*P10 + M(0,2)*P20_10,
M(1,1)*P10 + M(1,2)*P20_10,
M(2,1)*P10 + M(2,2)*P20_10);
float P23 = P(2,3);
float P33 = P(3,3);
osg::Vec4 pixelSizeVector(M(0,2)*P23,
M(1,2)*P23,
M(2,2)*P23,
M(3,2)*P23 + M(3,3)*P33);
M(1,2)*P23,
M(2,2)*P23,
M(3,2)*P23 + M(3,3)*P33);
pixelSizeVector /= scale.length();
//cout << "pixelSizeVector = "<<pixelSizeVector<<" pixelSizeVector2="<<pixelSizeVector2<<endl;
float scaleRatio = 1.0f/sqrtf(scale_00.length2()+scale_10.length2());
pixelSizeVector *= scaleRatio;
_modelviewCullingStack.push_back(osgNew osg::CullingSet(*_projectionCullingStack.back(),*_modelviewStack.back(),pixelSizeVector));