When will we completely get rid of jagged edges in games?

[Tim Murray]

er, they show more signs of JPEG compression than edge enhancement...

 

[MuFu]

Yes.

 

[Jack_Tripper]

As for aliasing in the real world, that can only really occur in your peripheral vision (foveal sampling is anti-aliased optically and AFAIK, there's a huge margin for natural deviation). There's moiré, which you can see quite often in movies, but that isn't really a sampling issue as such.

Moire can be cured though via antialiasing last time I checked. Irrelevant to that, a weird example is if a window/door is screened with one of those ultra fine meshes for mosquitos and you look outside. Of course does it depend on what you're looking at, it's pattern, distance and a whole lot of other factors, but something like a fine fence outside or anything with very close parallel lines is about enough to expose a moire pattern.

That example hasn't of course anything to do with Pixar's movie in question.

But surely a shortcoming in the ability of our visual system to resolve fine detail would make us *less* sensitive to edge aliasing in a motion picture?

I'm not talking about any shortcomings or special abilities at all. Coincidentially I did notice something the first time I saw the movie and I thought that I was just a bit too tired or something else. I fooled around with the movie at home in quite a few scenarios and yes it was there. Extremely hard to notice. Most people concentrate on the movie itself and don't notice such fine details anyway; in my case it was a pure coincidence. Apart from that aren't they using a stochastic 64x sample FSAA algorithm? Why wouldn't a weird frequency and/or an extremely weird angle NOT in fact expose any form of aliasing after all? We're not talking about 10000x samples per pixel here, are we?

That is a true statment!

(snuck in a trip to Pixar earlier this year...I almost creamed my pants just standing in the middle of all that hardware and talent)



Jack

 

[Geo]

A couple interesting "stay tuned" from sireric here: http://www.rage3d.com/board/showthread.php?t=33796563

I found this particularly interesting:

It's generally agreed that the current API for MSAA is somewhat broken, but that can be fixed.

Fixing the api requires MS involvement and new rev of DX, right?

 

[tEd]

can be fixed maybe will with dx next or wgf whatever

 

[Hyp-X]

I found this particularly interesting:

It's generally agreed that the current API for MSAA is somewhat broken, but that can be fixed.

Fixing the api requires MS involvement and new rev of DX, right?

Well it depends on what he means by that.

I believe the API for MSAA sucks - the quality levels introduced in DX9 are completly useless as it is.
Especially as (AFAIK) the DDI doesn't allow the driver to specify quality levels for DX9 compliant MSAA modes...

I don't know if that's what he means though.
Maybe they'd have a new mode but it could be only put in the legacy (nonmaskable) category only - where noone would use it?

 

[Ailuros]

but please at a higher resolution

Windows output was at 1600*1200*32; it's not my fault that winDVD captures in that resolution and yes I did of course compress the jpegs even more to save bandwidth.

 

[Ailuros]

I dunno, I think I'd settle for that resolution on my PSP3. 8)

Digi... those frames show evidence of EE. It looks like they might have even been processed twice in such a manner (!).

See above; albeit winDVD has an inbuild edge sharpening filter, it wasn't enabled, neither any other additional filter. Only PowerDVD can capture shots also in the used windows display resolution; time for an obnoxiously high resolution capture I guess.

***edit: dang! it's of no use; winDVD has at least the decency to be stuck in 1024-whatever but captures with 96dpi horiz/vertical. PowerDVD5.0 might allow to capture in the display resolution but it uses a pathetic 2dpi per axis. Now imagine how horrible the video source at 720*576 with 2dpi looks like when captured in 2048 (not even worth the hussle). In any case here's a video source capture from PowerDVD:

http://users.otenet.gr/~ailuros/nemo3.jpg

 

[Babel-17]

Jeez, guys...... I hate to say this.... but age will take care of those jaggies.....

Ya can't see jaggies if'n ya can see.........

Great; then my eyes still work as they're supposed to, since I could even see some minor aliasing in rare spots in Pixar's Finding Nemo as an example.

***edit: even the human eye in real time isn't absolutely perfect, when it comes to antialiasing.

Are you speaking of a DVD or the film itself?

From Pixar we'd expect perfection but from my recent readings I've gotten the impression that if, when a DVD is encoded, the "flags" aren't done just right the process of deinterlacing can briefly fail.

If you are referring to texture aliasing such as blockiness then that can often be a matter of the bitrate not being sufficient to effectively fill out what is being rendered. The topic goes way over my head but I've read where there are even encoding options to bias the bitrate vis a vis foreground vs. background. A good example of that I think are many XviD mp4 encodes. Faces look sharp and detailed but a solid textures in the background will look a mess.

Lol, Hi Ailuros! In all honesty I started reaching to reply before I saw the post was from you.

At least you'll know just how much salt to take with my answer.

 

[davepermen]

but please at a higher resolution

Windows output was at 1600*1200*32; it's not my fault that winDVD captures in that resolution and yes I did of course compress the jpegs even more to save bandwidth.

windows output, yes, but dvd input, no i wonder how crisp and sharp nemo would look, if directly rendered with 1600x1200 as target res..

 

[Moloch]

I dunno, I think I'd settle for that resolution on my PSP3. 8)

Digi... those frames show evidence of EE. It looks like they might have even been processed twice in such a manner (!).

See above; albeit winDVD has an inbuild edge sharpening filter, it wasn't enabled, neither any other additional filter. Only PowerDVD can capture shots also in the used windows display resolution; time for an obnoxiously high resolution capture I guess.

***edit: dang! it's of no use; winDVD has at least the decency to be stuck in 1024-whatever but captures with 96dpi horiz/vertical. PowerDVD5.0 might allow to capture in the display resolution but it uses a pathetic 2dpi per axis. Now imagine how horrible the video source at 720*576 with 2dpi looks like when captured in 2048 (not even worth the hussle). In any case here's a video source capture from PowerDVD:

http://users.otenet.gr/~ailuros/nemo3.jpg

What about PDVD 6?

 

[Ailuros]

What about PDVD 6?

Haven't tried it yet. Am I missing anything important exept possible screenshot improvements?

 

[Babel-17]

Hi Ailuros, PowerDVD 6 offers vastly improved support for HD TS files and a major improvement in its deinterlacing.

Seeking and fastforwarding/rewinding with HD transport streams (that are out of spec) needs work.

Nero's file splitter seems better than whatever PowerDVD 6 uses (Windows default?). It's what I use in ZoomPlayer combined with PowerDVD 6's renderer.

 

[hkultala]

***edit: even the human eye in real time isn't absolutely perfect, when it comes to antialiasing.

did u install latest drivers for your eyes?
i hear they give 10-15% boost on sunny day. 8)

Last time I checked, with latest drivers for eyes LOD is set ridiculously low and there's quite few clipping plane hacks. Still it gives really good IRLMark score though...

clipping plane hacks?
what do you mean by them?

actually even without latest drivers the performance at sunny day is quite good, and the new drivers are just needed when there is not so much light. (without the new drivers there is some blur filter, especially on pixels with high z-value )

also there is this nice grayscale effect which appears when the amount of light gets to very low.

 

[Sobek]

I would like to think that quantum computing and entanglement would play a large part in removing jagged lines...those of you who know much about quantum theory would understand why..otherwise i'm just too tired to post my reasonings

 

[KimB]

I would like to think that quantum computing and entanglement would play a large part in removing jagged lines...those of you who know much about quantum theory would understand why..otherwise i'm just too tired to post my reasonings

Aside from the difficulties of setting a quantum computer up to process the entire area of the pixel, summing phases isn't the same thing as summing probabilities, which is what you'd want for antialiasing. So no, I don't see how quantum computing would really help here.

 

[MuFu]

What about a Quantum 3D Obsidian 2 X24?

 

[Martillo1]

Quantum aliasem hoc video.

 

[digitalwanderer]

Well I found out how to eliminate jaggies entirely RIGHT NOW!

I just take off my glasses and they're all gone....

 

[PowerK]

Well I found out how to eliminate jaggies entirely RIGHT NOW!

I just take off my glasses and they're all gone....

But by doing that, LOD decreases.