Screen-Space Ambient Occlusion( SSAO )

[nullsquared]

No, the depth is grabbed just on position.

I've personally just gotten to writing up some good displacement shaders, but wouldn't something like the following work out nicely?
- grab texture-space matrix as usual (tangent/binormal/normal)
- grab texture-space depth from parallax map
- rotate float3(0, 0, -depth) by TBN matrix (now the depth is in object-space)
- displace interpolated fragment position by this new vector
- rotate result by world and view matrices, and take that for the view-space depth
I think it might just work, unless I'm missing something.

[Omnicrash]

It seems to me that also grabbing the depth of the parallax or bump would be an effect that is barely noticeable, and cost additional rendering time. It doesn't seem like it would be worth the effort.

Remember one of the most important rule of making games in terms of graphics: Faster & more beautiful is more important than more realism.
Which is why people use SSAO instead of 'real' ambient occlusion.

[nullsquared]

It seems to me that also grabbing the depth of the parallax or bump would be an effect that is barely noticeable, and cost additional rendering time. It doesn't seem like it would be worth the effort.

Hm.  It really wouldn't be much slower.  First of all, it wouldn't make sense with parallax or bump mapping, only with some type of displacement mapping (such as POM, or relief mapping, or cone step mapping, steep parallax, and the likes).  Second, doing this would involve very few extra pixel shader instructions:
- one texture fetch
- two matrix multiplies
- one mul and one add

[GD]

well if he's already using parallax mapping in scene, with deferred rendering it isn't much of a difference.
anyhow I was just wondering about parallax, don't get all into it

[AriusEso]

This will be released soon.

[click for big]







Bottom left shows output mode.

[Happy]

most astonishing work in truevision history ever.... 

[Enzi]

After getting a little help from AriusEso I sat down and made another approach of my last SSAO impelemtation which performed, to be honest, pretty bad.
Reading white papers and mixing different improvement techniques from AoC, Crysis, Inigo and Arius I finally made it. Without these guys it wouldn't be possible.

So, big thanks.

[Zaknafein]

That looks awesome! I wonder how visually close it is to a real, ray-casted solution...

[WEst]

It looks really great!!

Except maybe the hard silhouette between the world and the void, I admit I haven't dealt with SSAO by myself yet.
Is it possible to add some depth threshold ?

[Raine]

Can't wait for a release to put my dirty hands on.

/evil laughter

[Enzi]

It looks really great!!

Except maybe the hard silhouette between the world and the void, I admit I haven't dealt with SSAO by myself yet.
Is it possible to add some depth threshold ?

The outliner is actually a bug in the merging shader. I don't know yet why it occurs but it's not rendered in the ssao texture. It isn't even rendered in the world shader. I'll surely fix it, but right now it's not one of the priorities. I'm still working on the variance shadow mapping (which are working in my xna project) and implementing casceded shadow maps to get a better look in large terrains.

HDR and my own interpretation of IBL is also implemented. (Still in a very rough way but 3d is all about faking )

Maybe when it's all done I can show it in the showcase thread. Right now I'm not really satisfied expect for the fake ambient occlusion.

[Mietze]

and implementing casceded shadow maps to get a better look in large terrains.

You may want to take a look at this then:

http://www.truevision3d.com/forums/indevelopment/parallel_split_shadow_mapping-t17694.0.html

 

[AriusEso]

Can't wait for a release to put my dirty hands on.

/evil laughter

It's coming, it'll be in my sample pack somewhere down the line. It'll also use occlusion maps, which work statically to give the bump map some kind of fake occlusion.

Images of occlusion mapped bump mapping( All done in XNA, but it'll work in TV3D also ):

http://azazeldev.org/cleaner.png
http://azazeldev.org/aha.png

Anyway, this mapping will be combined with the SSAO in the sample. Sorry it's taking so long to get a decent workable sample of this out. But I'm split about 4 ways at the moment( TV3D, XNA, OGL, DX ) so my time is a little thin for each.

[Toaster]

That looks really nice keep up the great work! 

[Raine]

Quite cool chap, quite cool *sips tea and nods*

[AriusEso]

Quite cool chap, quite cool *sips tea and nods*

*hands Raine a scone*

[Mietze]

http://azazeldev.org/aha.png

I just love the "Peanut - Wikipedia" window

[AriusEso]

I just love the "Peanut - Wikipedia" window

lmao, I was trying to proove to someone on IRC that they are peas and not nuts.

[Mietze]

lmao, I was trying to proove to someone on IRC that they are peas and not nuts.

This is.. eh... to avoid offtopic s**t:

Code:

Console.WriteLine(Peanut.ToGerman())

Outputs "Erdnuss", where "Erd" ^= ground and "Nuss" ^= "Nut".. so.. hm... If its really a pea, it would output "Erderbse" ^= ground pea, which sounds like "ground pee"... so peanuts are definately nuts, no?

Finally, we should add a [SOLVED] over here... rofl

[AriusEso]

This is.. eh... to avoid offtopic s**t:

Code:

Console.WriteLine(Peanut.ToGerman())

Outputs "Erdnuss", where "Erd" ^= ground and "Nuss" ^= "Nut".. so.. hm... If its really a pea, it would output "Erderbse" ^= ground pea, which sounds like "ground pee"... so peanuts are definately nuts, no?

Finally, we should add a [SOLVED] over here... rofl

Aha, no sir, they be peas I tells you. . They even come in a pod.

The plant's name derives from a combination of the morphemes pea and nut, causing some confusion as to the nature of the fruit. Although a nut in the culinary sense, in the botanical sense the fruit of the peanut plant is a woody, indehiscent legume and not a nut. The word pea describes the edible seeds of many other legumes in the Fabaceae family, and in that sense, a peanut is a kind of pea.

http://en.wikipedia.org/wiki/Peanut