Vector Displacement Deformer

Recently I ran into an interesting problem which provoked me to create a node, I needed to create a particle simulation on top of a surface which had an animated vector displacement map assigned to it. The problem was that the original mesh didn’t have any of this deformation, so I had to come up with a way to create an animated mesh representation of the displacement which the renderer was creating. The requirements for this would be something which could gather points, normals, tangents and uvs of the mesh, supply them to SampleShadingNetwork function and then displace with the returned values.

Here’s a sample video of a mesh with vector displacement exported from MudBox:

Here’s a demo of whats possible just by tapping into the SamplerInfo node:

Here’s a walk through the code I’ve ended up with:

The first step is to get access to the name of the shading node and attribute plugged into the inColor, this requires the retrieval of the Mplug associated with InColor and then querying it’s connections. As a backup in case nothing is plugged in, I’m storing the color value as queried from the data-block.

	MString shadingNodeAttribute = "";
        MPlugArray connectedAttrArray;
        MPlug inColorPlug = MPlug(thisMObject(),aInColor);
        inColorPlug.connectedTo(connectedAttrArray,1,0);
        if (connectedAttrArray.length())
		{
			shadingNodeAttribute = connectedAttrArray[0].name();
		}
		else
		{
                shadingNodeAttribute = "";
		}

		// get incolor in case nothing is plugged into color//

  		MDataHandle inColorDataHandle = data.inputValue( aInColor);
		MFloatVector inColor = inColorDataHandle.asFloatVector();

I’m querying all of the uv, point, tangent and normal data at once to save processing time, however the uv, tangent and normal lists are poly centric meaning there is separate vector or value for each polygon surrounding a vertex. This means a loop is required to retrieve the proper tangent Id’s in order to pluck them from the array. I chose to do this per polygon and then average the polygons around each point in order to provide the smoothest information, especially in the case that the user provided a triangulated mesh.

			// arrays for sample shading network
			MFloatPointArray points;
			MFloatVectorArray normals;
			MFloatVectorArray binormals;
			MFloatVectorArray tangents;
			MFloatArray uCoords;
			MFloatArray vCoords;

			// face arrays for tangents

			MFloatVectorArray faceBinormals;
			MFloatVectorArray faceTangents;

			// arrays which come in poly centric and must be pruned

			MFloatArray dirtyUCoords;
			MFloatArray dirtyVCoords;
			MFloatVectorArray dirtyBinormals;
			MFloatVectorArray dirtyTangents;

			// fill arrays

			transMeshFn.getPoints(points,MSpace::kObject);
			transMeshFn.getUVs(dirtyUCoords,dirtyVCoords);
			transMeshFn.getTangents(dirtyTangents,MSpace::kObject);
			transMeshFn.getVertexNormals(false,normals,MSpace::kObject);

			//// loop through polys, retrieve tangent ids and populate face tangent array  ///

			int numpolys = transMeshFn.numPolygons();

			for (int i=0; i<numpolys;i++)
			{
					MFloatVector avgTangent;

					MIntArray polyVerts;
					transMeshFn.getPolygonVertices(i,polyVerts);
					int polyVertsLen = polyVerts.length();

					for (int d=0; d<polyVertsLen;d++)
					{
						int vertId = polyVerts[d];
						int tangentId = transMeshFn.getTangentId(i,vertId);
						avgTangent+= dirtyTangents[tangentId];
					}
					faceTangents.append(avgTangent);
			}

In the case of the UVs which will be supplied to SampleShadingNetwork, I’m querying the uv indices for the point and then using the first one the list, as averaging these doesn’t make much sense. In order to get the binormal, I calculate the cross product of the normal and the tangent and multiply it by -1.

			while (transMeshIt.isDone() == false)
			{
				int index = transMeshIt.index();
				MIntArray normInd,uvInd,faceInd;

				////// get uv and normal indices assosciated with  this point

				transMeshIt.getConnectedFaces(faceInd);
				int faceLen = faceInd.length();
				MVector avgTangent;

				//// average tangents

				for (int i=0; i<faceLen;i++)
				{
						avgTangent += faceTangents[faceInd[i]];
				}
				avgTangent.normalize();
				tangents.append(avgTangent);

				//cross tangent and normal to retrieve binormal

				MVector avgBinormal = avgTangent^normals[index];
				binormals.append(-avgBinormal);

				//// use first uv index

				transMeshIt.getUVIndices(uvInd);
				uCoords.append(dirtyUCoords[uvInd[0]]);
				vCoords.append(dirtyVCoords[uvInd[0]]);

				transMeshIt.next();
			}

With all of this data now in vertex centric arrays I’m ready to perform the sample, returning me an array of colors to work with. To perform the tangent based displacement, I use the same arrays constructed for SampleShadingNetwork in order to construct a matrix for each point. Using this matrix, the values from the color are then transformed by the matrix as a point, or in option that world space is chosen, the values are simply added to the current point.

			if  (shadingNodeAttribute != "")
			{
				MRenderUtil::sampleShadingNetwork(
							shadingNodeAttribute,
							points.length(),
							false,
							false,
							cameraMatrix,
							&points,
							&uCoords,
							&vCoords,
							&normals,
							&points,
							&tangents,
							&binormals,
							 NULL,
							colors,
							transps );
			}
			while (transMeshIt.isDone() == false)
			{
				int index = transMeshIt.index();

				MFloatVector normal = normals[index];
				MFloatVector tangent = tangents[index];
				MFloatVector binormal = binormals[index];

				normal.normalize();
				binormal.normalize();
				tangent.normalize();

				// construct matrix from normals trangents and binormals

				double matrix[4][4] = { {tangent.x,tangent.y,tangent.z,0}, {normal.x,normal.y,normal.z,0},{binormal.x,binormal.y,binormal.z,0},{points[index].x,points[index].y,points[index].z,1}};
				MMatrix instMatrix(matrix);
				MPoint colorPoint = inColor;

				// set color if color was sampled

				if (shadingNodeAttribute != "")
				{
					colorPoint = colors[index];
				}

				// set point based off of displacement type chosen

				if (inDisplacementType == 0)
				{
					post_trans_points.set(MFloatPoint(((colorPoint*inScale)*instMatrix)*inverseMatrix),index);
				}
				else if (inDisplacementType == 1)
				{
					post_trans_points.set(MFloatPoint(((colorPoint*inScale)+points[index])*inverseMatrix),index);
				}
				else
				{
					post_trans_points.set(MFloatPoint((colorPoint*inScale)),index);
				}
				transMeshIt.next();

After testing this I realized that from inside the shading network going into my node you could gain access to all of the variables provided to SampleShadingNetwork through the samplerInfo node, so I chose to add the option to have the retrieved color simply become the new position of the point. This way you can perform more complex operations with the points using the nodes of the shading network and not be limited to a relative transformation.

 

• May 4, 2013
• Uncategorized
• 1 Comment