多边形体素化的精确纹理映射

Question

我最近实施了体素锥跟踪算法，第一步是对多边形进行体素化...

我使用稀疏八叉树来转换多边形，除纹理映射外，一切都很好

mipmap = 6（64x64x64）

// We find its smallest AABB
AABB aabb = FindAABB(in);
// better aabb with same width, height and depth
AABB betterAABB = FixAABB(aabb);
// Create Octree and marked the grids that are touched by triangles in the mesh
// dfunc is a callable object which return true when triangle and grid are overlapped
// cb is a callable object, it is called when octree reaches its leaf..
Octree<uint8> tree(betterAABB, mipmap);
for(auto &f: in.faces)
    tree.findTouchedLeaves(f, dfunc, cb);

回调功能......

struct Callback
{
    void operator()(OctreeNode<uint8> *node, const Face &obj)
    {
        // Check if the node is already marked.
        if(!node->data.value)
        {
            const glm::vec3 &center=node->aabb.getCenter();
            out.push_back(center);
            // Problem here!! how to calculate precise uv for texture mapping
            color.push_back(text->getPixel((obj.o.texCoord+obj.a.texCoord+obj.b.texCoord)/3.0f));
        }
        node->data.value=1;
    }

    std::vector<glm::vec3> out;
    std::vector<glm::vec4> color;
    const ImageTexture *text;
} cb;

你可以看到...我直接平均了o，a，b的三个紫外线（可怕吧？） 当三角形大于网格时，它更不精确。 我是否需要使用基于栅格化的算法？

我的问题是：如何精确地为网格着色？