Title: Idea for inversions - no more fuzzy? Post by: DarkBeam on March 20, 2011, 12:00:06 PM :D
I have tried this modification for my sphere inversion; .Version = 2 .DEscale = -1 .SIpower = -1 Modified ONLY de-options I suppose that this recalculate the distance vector, so fixes the distance estimation bug, raising it by a power of -1. Descale... I don't know what it is exactly, but it fixes even better the distance estimation :D Is it possible to make the same with _reciprocalX and every other inversion? And how do I calculate and modify that blessed vector? ;D Now my sphere inversion works incomparably better! Before, shapes looked very odd. Near the center they looked kinda fuzzy, far from it they looked more and more rounded. Wrong! :angry: Attached images with vs without. :o Title: Re: Idea for inversions - no more fuzzy? Post by: DarkBeam on March 20, 2011, 12:38:21 PM Anyway the trick seems to work well with reciprocalX too... Now I rendered a Menger sponge with fixed reciprocalX and I see a bunch of small details, that before were hidden. Also a "wall" appears at x=0 but it's not a trouble (this is due to overflows I think, so he thinks that an overflow zone is a "zero" zone, can be fixed easily I think) :hmh:
(http://img40.imagefra.me/i53k/kriptokapi/l21o_ab9_u6bci.jpg) (http://i.imagefra.me/f02i40b7) Obviously, the more far you go from the origin the less works the de, but it's normal :) Only have set destop to 1.5 and not 1, to avoid noise in some spots Title: Re: Idea for inversions - no more fuzzy? Post by: MarkJayBee on March 20, 2011, 03:50:04 PM Hurrah! Great work Luca! :D My two current fave formulas, and you seem to
be on the track to fixing the one major pain with them - noise!! Looking forward to using them when available. :dink: Title: Re: Idea for inversions - no more fuzzy? Post by: DarkBeam on March 20, 2011, 04:12:48 PM Hurrah! Great work Luca! :D My two current fave formulas, and you seem to be on the track to fixing the one major pain with them - noise!! Looking forward to using them when available. :dink: The possible issue is that they are moved to "formula" folder and no longer in transforms... Now let's wait for Jessie :canadian: Title: Re: Idea for inversions - no more fuzzy? Post by: DarkBeam on March 20, 2011, 04:51:16 PM Okay: I added the new fixed sphere inversion. Obviously, I haven't changed the code but only the options so when the old one didn't worked at all this also doesn't work at all ;D ... The difference is only less noise, and the tab (shown in 3D formula tab and no longer in tranforms).
I don't fix all other ones because if I do so the 3D formula tab becomes plenty of transforms! :sad1: Also, added "Celtic mode". See the file for detailed info on usage :angel1: Title: Re: Idea for inversions - no more fuzzy? Post by: Jesse on March 20, 2011, 08:16:50 PM As far i can tell the reason is the same as why i put my speherinvc formula into the 3d section: without the .DEoption parameter it is set to 0, meaning it is calculated with Buddhis 4 point method. In this case it is better because the analytical calculation of the menger or other anaylical formulas are disturbed by the reciprocal thing. The Descale is a factor that will be multiplied with the calculated DE, dunno why -1 is working at all - it should not work! In a hybrid it will be averaged with other formulas scalings, but i would change it to 1 or leave it at all. Same for the Si (smoothed iteration) value. I have to make a switch to disable the analytical DE calculation at all, so we don't have to put all that addons to the 3d tab... an empty dummy 3d formula in the first position could do this job as well, maybe this could be an option until then? |