detail enhancement vs brute force.

[super gnu]

ive noticed, comparing renders done with brute force and those done with detail enhancement that i dont get the detailed shadows around small objects with detail enhancement that i do with brute force? can someone shed some light on why? see two examples.. the dado and picture rails in these two thumbnails are what im focusing on.

with imap i couldnt get clean edges there no matter what i did with the settings so i tried both detail enhancement and brute force.. both cleaned the problem, but the brute force has much nicer shadowing above and below the rails.. (done with the "universal method") the detail enhanced one looks a bit... flat. (the brute force one is grainy since that little thumbnail took 10mins to render.. i didnt fancy waiting an hour for a clean one.)

id have thought they would be almost identical.. after all, the detail enhancement basically uses brute force in the areas that need it.. or..?




also as another obsetrvation.. i have to drop the max rate of the imap when i use detail enhancement, or i -still get the blotchy edges, which is strange because with a radius of 20 pixels on the detail enhancement id have thought the imap wouldnt affect these areas.

[super gnu]

p.s. apologies for tiny thumbnails..

[Morne]

when using the detail enhancement, also try to tick "check sample visibility" - below that and a bit to the right

[super gnu]

hmm i had check sample visibility on and off during various tests to get a clean imap.. didnt seem to help in this case.

i tried enabling now with det.enh. on. but it didnt really make any difference.

[glorybound]

My understanding is BF samples every pixel and the other blend. I think Detail Enhancement relies on other V-Ray settings so a combination of several other things might make you Detail Enhancement look as good as you BF. I guess the question is which has better render times.

[super gnu]

well theres no doubt which has better rendertimes 1 min vs 10.. however as i understand it, detail enhancement -uses- BF in areas around details, using a similar system to AO to decide where to apply the effect. since the only settings we have for this are the radius (ive set it up to 100 pixels - larger than the whole image, with no real improvement) and the subdiv multiplier, which set at 0.5, and considering my imap setting of 200 hsph subdivs means 100 BF subdivs, im at a loss how to get it to "enhance my details"

also as mentioned, if i have my imap set to -3, 0 i -still get blotches along those problem edges despite having det. enh. set to 20 pixels, which if i understand correctly, should mean the imap isnt even being used in those areas, and they should be using pure BF.

ive had to set my imap to -4-2, which, used alone gives a clean but very flat lighting solution. adding detail enhancement helps a bit bring the detail back, but just seems lacking when compared to pure BF.. and i dont understand why.

[super gnu]

is it possible that the detail enhancement is being interpolated with the imap? i cant see how, because its calculated at rendertime, after the imap has finished calculating, but could explain the loss of little corner shadowing im getting.

[super gnu]

vlado you have any thoughts on this maybe? basically i love the idea of detail enhancement, but never end up using it as whatever i do i cant seem to get the same detailing in the corners etc that i do with brute force. in fact it never looks much different to the imap only image.

[3DMK]

Firstly I would try world units for DE [Detail Enhancement] I have always done it that way & always got good results so it may have something to do with you using screen space??

Just set a world scale distance.....I find about 1-2 cm perfect....& away you go.
You must also remember that the DE will use the Imap subdivision value for it's BF subdivision value so if your Imap subdiv was at 150 then your DE will be rendering BF at 150 Subdiv....NOT A GOOD IDEA.
The trick is to render your Imap at high Subdiv & then save & reload it & turn down the Subdiv to something like 32 otherwise render time blowout!!!

Also depending on your scene [indoors especially] you will need rather large subdivision setting for an Imap of -3 & 0 [maybe even up to 200 to 300 subdivisions] depending on how hard the renderer searches for lights.

& one last thing.....the quality of your secondary GI bounces [LightCache] also affects the quality of your final GI, especially with Imap as your primary bounce.

Hope this helps

BY the way VLADO...can we have a falloff control similar to the dirt map for blending the area between BF & Imap when using Detail Enhancement.
It often look ugly as you get a real defined line between the BF & Imap & there is currently no way to control this

[bardo]

Firstly I would try world units for DE [Detail Enhancement] I have always done it that way & always got good results so it may have something to do with you using screen space??

You must also remember that the DE will use the Imap subdivision value for it's BF subdivision value so if your Imap subdiv was at 150 then your DE will be rendering BF at 150 Subdiv....NOT A GOOD IDEA.
The trick is to render your Imap at high Subdiv & then save & reload it & turn down the Subdiv to something like 32 otherwise render time blowout!!!

But you can control this with the subdiv multiplier... so if you want 150 for imap and 32 subdiv for the DE, you can set subdiv multiplier for DE at 0.21.

[super gnu]

thanks for advice guys.. im currently using 200 hsph subdivs for imap, as mentioned, and the gi looks clean, apart from around those thin picture and dado rails. ive tried bf and lc secondaries and the lc is set very high.. its the same lc settings ive tried using bf as primaries and ive gotten a good result. i was also using -3,0 but as mentioned, this gave me the blotchies around those details. i -had- to drop the imap to -4 -2 to lose the blotchies, even when using det enh. which is kinda one of my points.. if det enh is set to a radius of 20 pixels, it should be overriding the imap in those areas, so why can i still see blotches there?

as bardo mentioned we have a subdiv multiplier, so whatever you set your imap to you can get the right subdiv level for the det enh. on a note about getting a defined line where the det enh. is, id suggest that is because you are using a very small (1-2cm) radius for the det. enh.. it only has a very short distance to blend over. i generally find that to avoid this visible boundary you have to use a much larger radius (the falloff basically) i usually lose the visible boundary with a distance closer to 10-20cm in world space in a normal sized interior.

my main issue however is that compared to BF (whn using the same sec. bounce lightcache settings) there seems to be much less detailed shadowing around objects with detail enhancement, whatever radius and subdivs i use. ive tried world space and screen space (up to radiuses that cover the whole scene), and ive tried imap settings from -4 -3 up to -3 0 and up to 500 hsph subdivs. as my multiplier is set to 0.5 thats a det. enh subdiv of 250!

[3DMK]

But you can control this with the subdiv multiplier... so if you want 150 for imap and 32 subdiv for the DE, you can set subdiv multiplier for DE at 0.21.

SORRY GUYS....had a brain fart

I did mix the explanation of DE up with the Use Irradiance map option on Vray material where when checked it will force object with that material to use BF instead of Imap [Vlado clarified once that the BF does work off the IMap subdivisions in the Use Irradiance map option]

That may be handy for what you want as you can have all 3 different GI engines working to the way you want it.