Forum Coordinators: RedPhantom
Poser - OFFICIAL F.A.Q (Last Updated: 2024 Nov 29 7:57 am)
Drop your Shading rate to 0.1 and increase Pixel samples to 4, if you haven't already done so. I'd honestly forget about Post Filter; it tends to slow things down and screws up details.
Coppula eam se non posit acceptera jocularum.
Attached Link: http://en.wikipedia.org/wiki/Anti-aliasing
Antialiasing is a topic much more complex than one would want it to be, and unfortunately there is no one-size-fits all solution (even worse, there is no perfect solution). If you do the math (see the attached link), you'll find out that the (default) 1-pixel box filter is only a quick-and-dirty approach and that higher quality AA requires higher oder filters that span several pixels. I suggest you try the combination of 2 pixel gauss filter and 3 pixel sinc filter for a start (for the sinc filter it may be necessary to increase the displacement bounds to 0.2 or so). Use between 3 and 5 pixel samples. For the scene in the render you posted, you shouldn't need more pixel samples than that, and a lower shading rate will only make it slower and use more RAM, but not make any difference in looks.That's a great improvement. I'd been having the same problem with
my 'hard-edged' architectural scenes. Will try those same settings!
My python page
My ShareCG freebies
That... was one of the best tip I've had for some time, it really improves my renderings. I was always a bit annoyed about the jagged lines in poserrenderings, but this really solves it. And as I look back it make sense, the more samples you make the better you anti-aliasing works. I also tried it with the samplingrate of the ambient-occlusion settings and that also makes improvements. It slows down my rendering, but I get razorsharp images. Only, setting the shadingrate to 2 makes it worse in my experience, 0,2 works better for me. Not only does it improve the "hard-edged" lines, but somehow it also softens curved lines.
Thanks for bringing this up Ptrope.
best regards,
Bopperthijs
-How can you improve things when you don't make mistakes?
Quote - absolutely never use post filter IMVHO. leave it at default. save as tiff file, not jpeg.
yes filters are only good for some sort of special effect that does not seem very special ;o)
I actually save all my renders as a png file. they are smaller in size than a Tiff and I havent noticed any loss in detail. posers jpg setting is horrible never ever use it.
One word of warning and advice: increasing the sampling rate, also increases the rendertime considerably (like I wrote earlier), making my bucketsize smaller (32 instead of 64) seems to approved the rendertime dramatically.
happy rendering,
Bopperthijs
-How can you improve things when you don't make mistakes?
Attached Link: http://graphics.stanford.edu/~mmp/chapters/pbrt_chapter7.pdf
> Quote - > Quote - absolutely never use post filter IMVHO. leave it at default. save as tiff file, not jpeg. > > > > yes filters are only good for some sort of special effect that does not seem very specialDon't confuse rendering filters with Photoshop filters. They're not the same! The post filter is what's combining the pixel samples into pixels, so it's operating at a different level than any filters you apply to a final image. This is an essential part of the antialiasing process in a 3d renderer. As I wrote earlier, there is no 'perfect' antialiasing (in theory, it would be an infinite number of samples with an infinitely large sinc filter and inifnite render time). You can get close, but you'll have to live with tradeoffs in one or the other direction. A box filter is simple and fast because it gives all samples equal importance. Gauss and sinc give the samples in the center of a pixel more weight than the others. The differences are especially visible when it comes to details smaller than a pixel, for example thin strands of hair. Many simple renderers (like the Poser 4 renderer) simply pick a filter and leave you with that. If you want the nitty-gritty details, the attached link is the 89 page chapter "Sampling and reconstruction" from the book "Physically based rendering". The part regarding filter functions (7.6) starts at PDF page 72.
Click the image to see it at the correct size.
(">10 times" because the pixel samples are in a nn square - a value of 6 results in 66=36 samples, 20*20=400 samples.)
That's our "more crazy technical rendering setting details than I wanted to know"-session for today. If you haven't had enough yet, you can take a look at discussion boards and tutorials for other renderers that allow a choice of pixel filter - mental ray, the Maya renderer or any RenderMan engine. As interesting facts on the side, the "mental ray production" preset in Maya or Softimage XSI defaults to 3 pixel gauss, the Maya renderer goes to 2.2 pixel triangle in production mode and DAZ|Studio appears to set the 3Delight engine to 4 pixel sinc, (3Delight itself defaults to 2 pixel box). C4D does not allow you to pick the filter size, but gives you also the choice between several filter types.
To conclude with a bad pun: If you care about good antialiasing, think outside the box filter. :tongue1:
Stupid answer: depends on your definition of "work". The big rule of computer graphics, "whatever looks right, is right", let your eyes be the judge. The math may say a 1 pixel box filter does not result in optimal antialising, but if said filter looks exceptionally good for your scene, then yes, by all means use it! At 1 pixel size, there probably won't be that much of a difference between the three filter types. It's just my personal experience that the gauss filter looks best at size 2, the sinc filter at size 4 and the box filter at size 1. But again, try out things for yourself.
This site uses cookies to deliver the best experience. Our own cookies make user accounts and other features possible. Third-party cookies are used to display relevant ads and to analyze how Renderosity is used. By using our site, you acknowledge that you have read and understood our Terms of Service, including our Cookie Policy and our Privacy Policy.
What am I missing, here?