Htykuut

I'm trying to compute UV coordinates in cylindrical coordinates from a set of vertices from a mesh but it's not wrapping properly.

So here what I have so far. Assuming the variable height and minY are already computed outside the for loop.

for (int i = 0; i < mesh.numVertices; ++i)

{

  float x = mesh.vertexBuffer[i].pos.x;

  float y = mesh.vertexBuffer[i].pos.z;



  float v = (mesh.vertexBuffer[i].pos.y - minY) / height; // get the ratio



  Vec2 circle = glm::normalize(Vec2(x, y));



  x = circle.x;

  y = circle.y;



  float theta = atan2(y, x); // its range is -pi < theta < pi

  theta += PI; // make it such that its 0 < theta < 2pi



  float u = theta / TWO_PI; // Get ratio;



  mesh.vertexBuffer[i].uv = Vec2(u, v);



}

As you can see, the texture image is not wrapping properly as there's this weird strip on it.

Can anyone tell if my derivation for cylinder is correct?

You have to duplicate the start vertex of each disc and to reuse it at the end of the circumference. The start vertex is associated to the texture coordinate U=0 and at the last vertex has to be associated with the texture coordinate U=1.0.

If you don`t to so you would get a transition from the last point on the circumference to the first point, where the last point has a U coordinate near to 1 and ant the first point a coordinate of 0. This cause that the entire texture is squeezed in this small area (flipped along the y axis). This is the "weird strip" you can see on your image.

Create one vertex at the end of the circumference which is equal the start - focus on <=:

for (int i = 0; i <= mesh.numVertices; ++i)

{

   ....

}

The contour where the texture wraps around onto itself represents a UV-space discontinuity. Vertices on it belonging to triangles on one side have U = 0, and on the other U = 1:

This raises the need for multiple UVs per vertex (imagine if the red and orange vertices were in the same position). It is common for 3D libraries to use a separate UV buffer, with its own set of indices.

Also, it is insufficient to process each vertex individually – there would be no way to differentiate between the U = 0 and U = 1 cases.

To resolve this, any vertex must instead be viewed as part of the triangle it belongs to.

e.g. in the figure above:

• 
  Yellow: the other vertices are on the "RHS" (theta < PI) of the discontinuity contour (gray), so the highlighted vertex has U = 0.


• 
  Red: the other vertices are on the "LHS" (theta > PI), so the highlighted vertex has U = 1.


• 
  Blue: crosses the contour, so must either be split along the contour or be discarded as invalid (unless you set up texture tiling which is always a source of grief).


• 
  Yellow: represents all other cases, and should be processed in the usual manner.

[Code coming soon]

^[hopefully]