This topic is 3854 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

im about to implement some shadows in my terrain engine... its a pretty large landscape with tress, grass and rocks... and im a bit unsure which shadowing technique to use... my only light source is the sun... i thinking about ortho cascaded shadowmaps... but i dont know if that the best and simplest solution...

##### Share on other sites
Yeah PSSM/CSM works really nicely. Very good example of PSSM on this page: http://appsrv.cse.cuhk.edu.hk/~fzhang/pssm_project/

##### Share on other sites
For just terrain (shelf) shadows google "Ambient aperture lighting". Fast, easy and good-looking, What else can you ask for? There's also an example in nVidia's FX composer 2 (beta).

##### Share on other sites
I have had good success with logarithmic shadows in my own project..very easy to implement and works in all cases..(with some artifacts but not much);

##### Share on other sites
Quote:
 Original post by Matt AufderheideI have had good success with logarithmic shadows in my own project..very easy to implement and works in all cases..(with some artifacts but not much);

EDIT:: googled didnt find any papers on it.. that were free

##### Share on other sites
I second looking into "Ambient Aperture Lighting" - there's a copy of the paper on ATI's developer site I believe. It's simple, cheap and looks quite good.

Crysis is apparently using variance shadow maps for their terrain shadows which seems like a good idea as the depth complexity of terrain is fairly low. VSMs can also be combined with other techniques such as PSSM to good effect.

##### Share on other sites
Quote:
 Original post by AndyTXI second looking into "Ambient Aperture Lighting" - there's a copy of the paper on ATI's developer site I believe. It's simple, cheap and looks quite good.Crysis is apparently using variance shadow maps for their terrain shadows which seems like a good idea as the depth complexity of terrain is fairly low. VSMs can also be combined with other techniques such as PSSM to good effect.

thx.. ive looked into Ambient Aperture Lighting and that seems quite good... also VSM seems interesting... but that goes for terrain...

for grass i thought PSSM or CSM.. or this logarithmic one... however which one would be simplest to implement?

##### Share on other sites
Quote:
 Original post by Dragon_Strikefor grass i thought PSSM or CSM.. or this logarithmic one... however which one would be simplest to implement?

Do you need to see the individual shadows for each blade of grass? That's gonna be pretty ugly no matter which method you use as grass is *way* too high frequency. I'd suggest looking into a probabilistic method for computing occlusion from grass (I believe several recent grass papers have done something similar).

##### Share on other sites
Quote:
Original post by AndyTX
Quote:
 Original post by Dragon_Strikefor grass i thought PSSM or CSM.. or this logarithmic one... however which one would be simplest to implement?

Do you need to see the individual shadows for each blade of grass? That's gonna be pretty ugly no matter which method you use as grass is *way* too high frequency. I'd suggest looking into a probabilistic method for computing occlusion from grass (I believe several recent grass papers have done something similar).

that would be an option... however since im using several different plants its not rly what im looking for... wouldnt it be possible to do alot of blurring on the shadows to hide the artifacts? sharp shadows arent required...

##### Share on other sites
Quote:
 Original post by Dragon_Strikethat would be an option... however since im using several different plants its not rly what im looking for... wouldnt it be possible to do alot of blurring on the shadows to hide the artifacts? sharp shadows arent required...

You can't blur standard shadow maps and expect to get proper (or even reasonable) results.

You *can* blur variance shadow maps, but you'll tend to just lose the shadows altogether if they are very high frequency (i.e. grass). That said, combining VSM with some sort of ambient occlusion approximation might get you what you need.

VSMs also have the advantages of hardware filtering, multisampling (if supported) and alpha blending (nice for foliage). Combined with PSSM they might give very good results, although I haven't tried them with a scene like you describe.

##### Share on other sites
Since Im currently also in the process of researching grass shadows, Id like to know what would you propose. Any ideas/links are useful.
Quote:
 Original post by AndyTXDo you need to see the individual shadows for each blade of grass? That's gonna be pretty ugly no matter which method you use as grass is *way* too high frequency.
Why should that be ugly ? Recently Ive watched Pixars Bugs life, and it looked prety good to me, although the shadows were hard.

Im thinking of grass blades self-shadowing itself so that you can clearly see the shadow of the individual grass blades moving all over the grass patch (fully-polygonal, about 150 tris per 15 grass blades). I dont necessarilly need dynamic grass-blades shadows over the terrain - thats actually pretty cheap to fake, just self-shadowing. Ill go for about 300k tris of grass per frame, so the faster the shadow method, the better. Static self-shadows of grass blades are fast and easy (and theyd probably still look good enough to an untrained eye), but what about dynamic ones ?

##### Share on other sites
Quote:
Since Im currently also in the process of researching grass shadows, Id like to know what would you propose. Any ideas/links are useful.
Quote:
 Original post by AndyTXDo you need to see the individual shadows for each blade of grass? That's gonna be pretty ugly no matter which method you use as grass is *way* too high frequency.
Why should that be ugly ? Recently Ive watched Pixars Bugs life, and it looked prety good to me, although the shadows were hard.

Im thinking of grass blades self-shadowing itself so that you can clearly see the shadow of the individual grass blades moving all over the grass patch (fully-polygonal, about 150 tris per 15 grass blades). I dont necessarilly need dynamic grass-blades shadows over the terrain - thats actually pretty cheap to fake, just self-shadowing. Ill go for about 300k tris of grass per frame, so the faster the shadow method, the better. Static self-shadows of grass blades are fast and easy (and theyd probably still look good enough to an untrained eye), but what about dynamic ones ?

well were gettinga bit of topic.. but i think i started a topic on this were it was discussed in quite detail...

http://www.gamedev.net/community/forums/topic.asp?topic_id=423966&whichpage=1�

however im still goin for shadowmaps for now atleast... and about individual grass blades.. its not required but it would be nice

##### Share on other sites
Quote:
 Original post by Dragon_Strikewell were gettinga bit of topic.. but i think i started a topic on this were it was discussed in quite detail...
Sorry, didnt mean to hijack your thread, it seemed to me relevant,because of previous posts. Ive seen that thread previously, but thanks.

Quote:
 Original post by Dragon_Strikehowever im still goin for shadowmaps for now atleast... and about individual grass blades.. its not required but it would be nice
Well, how would you escape self-shadowing of grass blades with shadow mapping ? Since its an image-space operation, youd get them automatically. Or am I missing something ?

##### Share on other sites
Quote:
Quote:
 Original post by Dragon_Strikewell were gettinga bit of topic.. but i think i started a topic on this were it was discussed in quite detail...
Sorry, didnt mean to hijack your thread, it seemed to me relevant,because of previous posts. Ive seen that thread previously, but thanks.

Quote:
 Original post by Dragon_Strikehowever im still goin for shadowmaps for now atleast... and about individual grass blades.. its not required but it would be nice
Well, how would you escape self-shadowing of grass blades with shadow mapping ? Since its an image-space operation, youd get them automatically. Or am I missing something ?

##### Share on other sites
I havent used shadow mapping with grass yet, so I cant say if its bad or if it isnt extreme performance overkill, but the only way you could achieve no grass self-shadowing seems to be in omitting grass rendering during shadow-map pass. Which, kinda defeats whole purpose.
Since I dont know anything about your camera, how much the player sees, from which height, and to what distance, its hard to say, if no grass shadows would make a big difference.

##### Share on other sites
You can also render a shadow map of object IDs rather than depths (or some hybrid). That will "avoid" self-shadows. I'm personally not a huge fan of that approach, but it could work in your situation.

##### Share on other sites

This topic is 3854 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Create an account

Register a new account

• ### Forum Statistics

• Total Topics
628642
• Total Posts
2983993

• 9
• 10
• 21
• 20
• 13