PDA

View Full Version : UV Mapping Brain Teaser



JGary
11-06-2008, 01:00 PM
Need some second opinions on this mapping I'm trying to create for vacuform printing. Image will be printing on a half hemisphere...printing service requested a top down image for printing purposes. I used a planar mapping on the y axis and then relaxed the UV's. The following image shows test results using the red lines for reference to make sure distortion was at a minimum. If you compare the red lines to the blue ruler ticks I created, you can see it's pretty close.
65640

I sent my test image off to the printer, but sent a reply back saying they needed the golf dimples to all be the same size! This seems like a mathematical impossibility. They sent me a sample showing how they would like to see the mapping. Take a look at the following image. My mapping is the top half, theirs is the bottom. Notice that on their version, the dimples become more narrow near the equator.
65639

I don't know...maybe I'm losing my mind, but I think the dimples have to either get wider at the equator or more narrow at the pole (depending on how it will be printed) in order for them to print onto the object as perfect circles. Am I nuts?

daforum
11-06-2008, 05:37 PM
You're not nuts, this is a difficult object/ surface to create.
Someone once said to me that wikipedia is your friend and they were right.

http://en.wikipedia.org/wiki/Golfball look at the image as it shows the pattern in a simple way.

Model the ball using a tesselated sphere and increase the number of divisions. This will create pentagons and each has an equal amount of dots in each pantagon (like the diagram in the wikipedia link)

Hope this helps.

daforum
11-06-2008, 05:42 PM
Sorry, your thread is about UV mapping not modeling :bangwall:

Could you use a spherical UV map instead as this will flatten the surface better than using the planar mapping technique?

JGary
11-06-2008, 06:14 PM
Sorry, your thread is about UV mapping not modeling :bangwall:

No problem, I actually looked at that page when I was researching creating the golf ball. Turns out a golf ball can have any dimple pattern, with the dimples varying size in endless configurations. Many pro golfers have their own individual dimple patterns tailored for their specific needs...seems it's a very scientific thing. In the end, I just bought a golf ball model off of Turbosquid.


Could you use a spherical UV map instead as this will flatten the surface better than using the planar mapping technique?

Yes, a spherical map would be great and that's how I originally created these, however this particular print service requires the polar coordinates. I just have to convince them that there is no way I can lay out the dimples using polar mapping and keep them all perfect circles...it just doesn't work.

hrgiger
11-06-2008, 07:04 PM
Yes, I just tried flattening half a sphere using UVLayout and I got the same distortion near the wide end of the sphere.

JGary
11-07-2008, 03:41 PM
Yes, I just tried flattening half a sphere using UVLayout and I got the same distortion near the wide end of the sphere.

Good to see the same result...thanks for the test. I just finished writing up a two page document with example pictures explaining why this won't work. Time to do battle...

Revanto
11-07-2008, 07:57 PM
There are a number of ways of trying to get the dimples correct. But is the dimples are fine on one UV map then you can always bake an image from one UV map to another using the surface baker plugin.

Don't give up. You'll get what you are looking for in the end.

Cheers,
Revanto :p

toby
11-08-2008, 02:54 PM
It's quite possible to have undistorted circles map without distortion.

Looks like your uv's per polygon are getting wider, although not longer, the further they are from the center.

Try an atlas map, or try this triangular pattern, from the wiki
http://en.wikipedia.org/wiki/Image:Two_similar_icosahedron_golf_ball_designs.jp g
It might be easier on a full sphere made from sub-patched tri's.

toby
11-08-2008, 03:02 PM
Oh I see, it *has* to be polar
Yea, tell 'em to forget it!

geothefaust
11-08-2008, 03:26 PM
Not that I'm a master or anything, but, I'm decent at UV stuff...

Unfortunately, with that type of projection you will never be able to achieve your goal.

LW_Will
11-10-2008, 11:38 AM
Hmm... I'm thinking change the UV map to a normal map and project it using displacement nodes. If you use an info node and select normal mapping, then attach the "normal mapped" shape, it might work.

Theoretically, of course.

JGary
12-12-2008, 10:04 AM
To follow up this thread...yes, it's impossible to create a polar mapping and have perfect circles that stay the same size. I had to spend a lot of time disproving this...good to see it finally resolved and at least I did get paid to argue about it. The final projection is close to the original I made.

Ivan D. Young
12-12-2008, 10:27 AM
This would be where texturing with Texels would be nice!

Revanto
12-14-2008, 11:08 PM
Hold on. If all you are trying to do is get perfect circles then you really need to consider texture baking like I suggested to you before. Basically, you texture your model in whatever way you wish then transfer the nice perfect texturing to your polar mapping. The texture baking process takes care of the actual sizing and warping issues when transferring across. It's just a matter of lateral thinking.

There may be other ways of doing this but this may be your best bet. Even baking textures across can be imperfect but you might not need alot of work to fix the issue.

Good luck.
Revanto :p

PS: Click on the link in my signature and find the video for Lightwave Projection Texturing. This should give you a clue to how it can be done.

JGary
12-15-2008, 02:44 PM
Hold on. If all you are trying to do is get perfect circles then you really need to consider texture baking like I suggested to you before. Basically, you texture your model in whatever way you wish then transfer the nice perfect texturing to your polar mapping. The texture baking process takes care of the actual sizing and warping issues when transferring across. It's just a matter of lateral thinking.

There may be other ways of doing this but this may be your best bet. Even baking textures across can be imperfect but you might not need alot of work to fix the issue.

Good luck.
Revanto :p

PS: Click on the link in my signature and find the video for Lightwave Projection Texturing. This should give you a clue to how it can be done.

Well, I want perfect circles when the map is applied to the model, but don't care if they are perfect circles on the texture image. The print facility was trying to tell me the texture image had to be a polar map and have perfect circles that were all the same size...there's just no way to make a polar projection like that and have it map correctly back onto a sphere.

I am using texture baking for all this stuff. I make the source object in 3d (basketball, golfball, etc.), then light the object and bake it using the proper coordinates. For some of these I bake a spherical normal map from the high-rez source, then apply that to a flat surface so I can evenly light the object. I then bake that out, apply it to sphere and bake again into the desired polar coordinates.

Attached is a image showing a section of the hopefully final polar coordinates and how it looks when mapped onto a sphere.

67218