Texture mapping to a sphere

[KDarshan20]

Hello,
I found PyVista while surfing the internet for a library that would help me in my project, a 3D solar system animation. So while working on the project I came across a problem: I couldn't get the textures mapped correctly. For instance,

import pyvista
sphere = pyvista.Sphere()
tex = pyvista.read_texture("Earth.png")
sphere.texture_map_to_plane(inplace=True)
sphere.plot(texture=tex)

Gave the following output:

While I want it to look like the in-built globe example:

Earth.png:

[akaszynski]

@banesullivan has done a lot of the implementation of texture mapping, so he'll probably correct me if I'm wrong, but I think the issue is that there's no way for pyvista/vtk to know the mapping between your earth png and the sphere. The texture_map_to_plane method works great on planes:

import pyvista
mesh = pyvista.Plane()
tex = pyvista.read_texture("Earth.png")
mesh.texture_map_to_plane(inplace=True)
mesh.plot(texture=tex)

What the texture_map_to_plane is doing is taking each coordinate of the image and interpolating it onto each coordinate of the plane mesh. This is easy to do in 2 dimensions, but in 3 dimensions, there's too many possible solutions. If you plot the mesh of the sphere, you can see that there's a non-uniform distribution of points:

So, what you'd have to do if you wanted it properly mapped is to get the right relationship of the points on the sphere to a plane (effectively, UV Mapping), and then use that projection to apply your texture.

If you do choose to project to a sphere, please post your solution here so others can benefit from it.

[KDarshan20]

Thanks, that helped me understand why the texture gets mapped so.
How do I implement uv mapping?

[KDarshan20]

I searched the documentation site and found out that texture coordinates can be defined using the t_coords property. So I have to implement UV mapping by defining t_coords by determining a relation between the XYZ coordinates and texture coordinates?

[banesullivan]

So I have to implement UV mapping by defining t_coords by determining a relation between the XYZ coordinates and texture coordinates?

That's correct! Keep your UV coordinates scaled between 0 and 1 and map each vertex to its coordinates in the UV image space.

FYI, there is a vtkTextureMapToSphere that we could implement much like the texture_map_to_plane filter

[banesullivan]

I created the globe example manually with the code in this PVGeo filter

Basically, I reproject the sphere's coordinates to a Lat/Lon space (thus a 2D plane) and essentially map the UV coordinates linearly there.

On a sphere, you HAVE to make sure there is an open edge along one longitude or else you will get artifacts of the texture folding back on itself.

I outlined all of this in this ParaView discourse thread

[KDarshan20]

I used this slightly modified form of the relation given in Wikipedia:

And also applied the solution you had given to remove the distortion at the seam, and came up with this code:

import pyvista
import math
import numpy
sphere = pyvista.Sphere(radius=1, theta_resolution=120, phi_resolution=120, start_theta=270.001, end_theta=270)
sphere.t_coords = numpy.zeros((sphere.points.shape[0], 2))
for i in range(sphere.points.shape[0]):
    sphere.t_coords[i] = [0.5 + math.atan2(-sphere.points[i, 0], sphere.points[i, 1])/(2 * math.pi), 0.5 + math.asin(sphere.points[i, 2])/math.pi]
tex = pyvista.read_texture("Earth.png")
sphere.plot(texture=tex)

(I've used a sphere of radius 1, so I don't have to explicitly calculate the unit vector)
And got this satisfactory output:

Thanks for the help! I'll close this issue now.

[banesullivan]

Awesome work, @KDarshan20! Thanks for sharing this

[KDarshan20]

Thanks to you as well!

[bluetyson]

Nice, now need to find a Mars!

[akaszynski]

Found Mars.

import pyvista
import math
import numpy
import numpy as np

sphere = pyvista.Sphere(radius=1, theta_resolution=120, phi_resolution=120,
                        start_theta=270.001, end_theta=270)
sphere.t_coords = numpy.zeros((sphere.points.shape[0], 2))


sphere.t_coords[:, 0] = 0.5 + np.arctan2(-sphere.points[:, 0], sphere.points[:, 1])/(2 * math.pi)
sphere.t_coords[:, 1] = 0.5 + np.arcsin(sphere.points[:, 2]) / math.pi


tex = pyvista.read_texture("mars.jpg")
# sphere.plot(texture=tex, smooth_shading=False, background='black', show_axes=False)

# with stars
pl = pyvista.Plotter()
pl.add_background_image('stars.jpg')
pl.add_mesh(sphere, texture=tex, smooth_shading=False)
pl.show()

---

Mars

http://www.solarviews.com/raw/mars/marscyl1l.jpg

Stars

https://pngio.com/PNG/a44278-space-stars-png.html

[bluetyson]

Great, thanks Alex!

[banesullivan]

Nice! @akaszynski, let's add that to pyvista/pyvista#940

[KDarshan20]

A source of textures for all planets, the sun, etc.: https://www.solarsystemscope.com/textures/

[banesullivan]

Awesome! I think I have to make a solar system example then!

[BrandonEscamilla]

@akaszynsk If the radius needs to be defined as 1738 km, what I need to do? I already tried your code but it doesn't work.

[akaszynski]

Hello,

Please post the full code you used.

[BrandonEscamilla]

Hi @akaszynski,

I am sorry for my late response but here is the code and the output. I am creating a sphere with the real radius of the moon. (1738 km)

`
p = pv.Plotter()
sphere = pv.Sphere(radius=1738)

sphere.t_coords = np.zeros((sphere.points.shape[0], 2))

sphere.t_coords[:, 0] = 0.5 + np.arctan2(-sphere.points[:, 0], sphere.points[:, 1])/(2 * math.pi)
sphere.t_coords[:, 1] = 0.5 + np.arcsin(sphere.points[:, 2]) / math.pi
print(sphere.t_coords)

tex = pv.read_texture("2k_moon.jpg")

p.add_mesh(sphere, texture=tex)
p.add_axes()
p.show_grid()
p.show()
`

[akaszynski]

BTW, if anyone is using this with pyvista>=0.33:

import pyvista
import math
import numpy
import numpy as np

from pyvista import examples

sphere = pyvista.Sphere(radius=1, theta_resolution=120, phi_resolution=120,
                        start_theta=270.001, end_theta=270)
sphere.active_t_coords = numpy.zeros((sphere.points.shape[0], 2))


sphere.active_t_coords[:, 0] = 0.5 + np.arctan2(-sphere.points[:, 0], sphere.points[:, 1])/(2 * math.pi)
sphere.active_t_coords[:, 1] = 0.5 + np.arcsin(sphere.points[:, 2]) / math.pi

mars = pyvista.Texture(examples.download_mars_jpg())
stars = examples.download_stars_jpg()

# with stars
pl = pyvista.Plotter()
pl.add_background_image(stars)
pl.add_mesh(sphere, texture=mars, smooth_shading=False)
pl.show()

[RichardScottOZ]

Definitely will try that one!

[RichardScottOZ]

[f123-maker]

Are there any methods to control the mapping position with coordinates? Many of the methods found are full package methods. I want a method to use coordinates for local mapping