PDA

View Full Version : Non-rectangular pixels and adaptive sampled images

Goon
06-02-2005, 02:16 AM
k, this may be one of the stupidest ideas you've heard in a while, but I thought I'd get it out of my system to find that out.

Now there is absolutely no way (or at least, difficult and quite annoying) to produce a continuous, distortion free uv map of a complex surface, barring something like zbrush's uv tile system, using the standard square pixel image. So what I'm wondering is why there aren't image formats and painting systems, that allow you to locally increase pixel resolution, kind of like maya's hierarchichal subDs. The image is interpolated anyway when you zoom in or out, so I can't help but think that a system like MR's adaptive sampling for anti-aliasing could be used to add increased detail, without upping the resolution of the entire image.

Furthermore, if a painting program were available that could read both uv, and surface information, besides being kickass, I would think it could do a better job of analyzing the surface, measuring edge length, and surface area, than a straight-up standard UV snapshot. So if the system could not only localize resolution densities to achieve an even pixel resolution, but also use parrallelogram shaped pixels to account for distortions in the UV face (with adaptive resolution, I'm really not sure if the parrallelogram pixels are neccessary), then you could take something like a LSCM unwrap that has a bit of distortion in it, and just start painting.

So yeah. Think this would work?
dans
06-02-2005, 03:42 AM
Pixels have no shape, so speaking of parallelogram shaped pixels makes no sense. Here is a nice Memo from Alvy Ray Smith along the lines of that. ftp://ftp.alvyray.com/Acrobat/6_Pixel.pdf

A Pixel Is Not A Little Square, A Pixel Is Not A Little Square, A Pixel Is Not A Little Square!

Of course the traditional 2D-rasterimage is not the only way to store color information for a surface. So your idea is not stupid at all.
Goon
06-03-2005, 12:48 AM
thx for the reply dans. I can't really look at that link, so I'm not sure about the details of the whole "A Pixel is not a square" deal. Is the gist more that a pixel is merely a point on a cartesian grid of x resolution, given a color value?
dans
06-03-2005, 02:12 AM
Yes, just a piece of data at a infinite small sample point.
jeremybirn
06-03-2005, 03:00 AM
that allow you to locally increase pixel resolution, kind of like maya's hierarchichal subDs.

You can locally increase the pixel resolution of any region or subregion of a textured surface by scaling up those polygons in the UV texture editor, so they cover and use more of the map. You can even have different maps at different resolutions covering the same area, with multiple UV sets. So, final-product-wise, I think you're all set. Of course, there always could be more elegant interfaces for setting this up, like painting with a resolution brush that when applied to a part of the surface made the corresponding UV's scale apart while warped the existing texture to match, prior to painting in small detail. But just about everything about UV's could always have a nicer interface...

Also, file formats like JPEG can have vastly different levels of compression and detail in different parts of the image. A program saving out a JPEG file could allow you to paint a mask of high detail vs. low detail areas and then smooth out and compress the image accordingly.

-jeremy
CGTalk Moderation
06-03-2005, 03:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.