Aliasing and Filtering explained

Great job, Wil. Enough detail to understand what's going on without being just too complicated for the mortal man to comprehend (Sorry Tim :D) - I'll definitely be showing this to people rather than trying to explain myself

Nice, but am I supposed to be able to see something in this shot?

http://www.bit-tech.net/content_images/aliasing_filtering/hl2noaf.jpg

It's a bit on the dark side! :o

Pretty good stuff :D The basics explained without going too overboard. A good read while I have a tea-break, cheers!

Originally Posted by mrplow
Nice, but am I supposed to be able to see something in this shot?

[ img]http://www.bit-tech.net/content_images/aliasing_filtering/hl2noaf.jpg[/ img]

It's a bit on the dark side! :o

lol thats what i was thinking, what bricks? :p

also those comparison screenies would probably work better if they were a bit bigger, had to zoom in to see the jaggies and nojaggies properly ;)

Excellent Read, it’s certainly helped me understand AA an AF better.

Thought the comparison screenshots where fine though, perhaps you guys need to tweak your monitor settings :)

Very good read, one of the best articles yet.
Explained everything in a way I could understand.

Very good article, i don't agree about the pictures though - the bricks example worked really well for me! :)

oh ok, i can see the bricks on this monitor, secondary monitor is set too dark :o

Hmmp.. JPG? Horrible quality... why dont you use png or gif.. much smaller and better quality with pic that have only a few colors in it.
http://www.bit-tech.net/content_images/aliasing_filtering/sampling.jpg

Good job. ;)

Originally Posted by mrplow
Nice, but am I supposed to be able to see something in this shot?

http://www.bit-tech.net/content_images/aliasing_filtering/hl2noaf.jpg

It's a bit on the dark side! :o

I thought the same, but after I tweaked the gamma and contrast in my display settings I could see it OK.

Good article, BTW...

I think most of us can agree that a brighter spot would have made a better example in the dark photo. But, either way, excellent article - you've made AA and AF easy to understand ;)

Fantastic article :)

Originally Posted by RotoSequence
I think most of us can agree that a brighter spot would have made a better example in the dark photo. But, either way, excellent article timmers - you've made AA and AF easy to understand ;)

I didn't write it, but I had an input. :)

Originally Posted by bigz
I didn't write it, but I had an input. :)

Oops

Originally Posted by RTT
Fantastic article :)

Indeed! /buttkiss :o

Originally Posted by mrplow
Nice, but am I supposed to be able to see something in this shot?

Well, it shows my monitor's set up right, unlike some. :p

Perhaps links to some freeware aids should be added to the article? :) :) :)

I understood all that and my head doesn't hurt too much, ;)

I'm pretty sure rotated grid 4xAA is _not_ rotated by 45 degrees... I think something more like 26.565 degrees would work better as then the four points have X and Y-ordinates regularly spaced along each axis, so for near-vertical and horizontal lines (where aliasing is arguably worst) it is of the best quality you can do.

OTOH NVidia and ATI might have decided to use 45 degrees for some perverse reason I am too stupid to comprehend... Or maybe I should patent the angle "26.565 degrees" and sell it to them :D

[edit] Found an in depth article on this http://www.3dcenter.org/artikel/multisampling_anti-aliasing/index6_e.php with some pictures:
http://www.3dcenter.org/images/multisampling_anti-aliasing/spnvati.gif
http://www.3dcenter.org/images/multisampling_anti-aliasing/ogrg.gif

An example of a difference between ATI and Nvidia is the fact that ATI implement gamma-corrected AA on all their 9700+ cards, a technique which samples alpha particles inside triangles as well as gamma particles on the outside of triangles. This is only implemented by Nvidia on the 7800 GTX.

Gamma particles? Is this some sort of inside joke?

I think you have perhaps simplified too much ...

Multisampling does not require samples to be shared between pixels. Its an approach some vendors use to try to get better image quality.

It's misleading to say mipmaps are lower-quality distance textures. They are multi-resolution texture image sets (pyramids) used to approximately match the variable resolutions needed for different object scalings by choosing the closest matching (or pair of matching) images. You probably meant that, but it's not what the article is saying ...

very nice article. explains without confusing. gives enough information to inform, without being too technical, and gives the reader many avenues to explore if they want to know more.

Great article there, thanks for it ;) .
I already knew about AA & AF but the article helped me understand more without bombarding me with jargon

Originally Posted by drb
I think you have perhaps simplified too much ...

It's misleading to say mipmaps are lower-quality distance textures. They are multi-resolution texture image sets (pyramids) used to approximately match the variable resolutions needed for different object scalings by choosing the closest matching (or pair of matching) images. You probably meant that, but it's not what the article is saying ...

I wonder if this borders on nitpicky or just a good point. The idea wasn't to bludgeon the reader with insane tech details, and the article does convey its point without getting too over-the-top.

Great job on the article, though. Many people do not have this level of understanding, and it's a nice and clear read. Articles like this are hard to come by, because you often get mired in the tech details on other sites.

Originally Posted by Wil Harris
As a 3D scene stretches into the distance, less detail is required as the objects become further away. Rendering textures that are far away at a lesser detail than those that are closer to the player means that graphics cards can give better performance. These lower-quality distance textures are called mipmaps.

I was always under the impression that mipmaps were introduced back in the good-old-bad-old days of software rendering as a bit of a dodge to reduce unpleasant aliasing with minimal additional overhead. Certainly, storing multiple versions of each texture requires more memory, and selecting between them before entering into the texture-mapping loop requires more time. Taking bigger steps accross a texture (due to distance from the viewer) doesn't need more processing time, just bigger step increments. I think that Mr. Harris might have gotten the wrong end of the stick on this issue.

Otherwise, a lovely introduction to the basics of AA and AF!

Nezuji :)

Erm, another point (From page 1 no less).

There ARE TFT monitors that are more than 1280x1024 in a 17" area.

Look at Dell's Top Quality notebook screens: UXGA (1600x1200) in a 15" TFT screen.

The only thing I can't understand is why they haven't released any "Desktop" TFT's at these res' yet? Notebook TFT's have been that high for a long while now (At least a year)