Title: New Rome (Mandelbox) Post by: David Makin on April 10, 2010, 04:38:29 AM Hi all, just trying a new ambient occlusion method for the Mandelbox.
"New Rome" (http://fc02.deviantart.net/fs70/f/2010/099/5/2/New_Rome_by_MakinMagic.jpg) (http://makinmagic.deviantart.com/art/New-Rome-160188432) If no image above then click the link. Title: Re: New Rome (Mandelbox) Post by: Power 8 on April 10, 2010, 05:00:31 AM Ave Caesar Davidus ! Mandelbox imperator! bravo! very roman indeed!
Title: Re: New Rome (Mandelbox) Post by: knighty on April 10, 2010, 06:32:28 PM Nice picture! :)
Could we see just the AO effect? Title: Re: New Rome (Mandelbox) Post by: Sockratease on April 10, 2010, 06:42:46 PM Very Nice!!
But... Was it built in a day? Title: Re: New Rome (Mandelbox) Post by: Saquedon on April 11, 2010, 12:58:17 PM Excellent! :D
Title: Re: New Rome (Mandelbox) Post by: David Makin on April 11, 2010, 09:50:46 PM Nice picture! :) Could we see just the AO effect? The method "Rectagular Ambient Occlusion" was suggested to me by Jos Leys. I've implimented the basic method such that the amount of occlusion at a given point is based on the maximum distance of the poiint from the centre of the cuboid along each axis (x/y/z). I also added an optional directional element. Here the basic method (location based only) on it's own for the "New Rome" fractal: (http://www.fractalforums.com/gallery/2/141_11_04_10_9_41_42.jpeg) And here's the enhanced method using the directionality - as applied in the original "New Rome" render: (http://www.fractalforums.com/gallery/2/141_11_04_10_9_45_43.jpeg) Note that the jet-black is not errors, it's holes through the Mandelbox to the background :) Of course the original "New Rome" is made up of the occluded ambient plus diffuse and specular from two light sources. Title: Re: New Rome (Mandelbox) Post by: Nahee_Enterprises on April 12, 2010, 04:58:15 AM Hi all, just trying a new ambient occlusion method for the Mandelbox. This is really cool !!! But something on the "busy" side. Kind of reminds me of the movie "Dark City", when the Strangers would modify the buildings while the people slept. (Looks like a few of yours are still in the process.) ;D Title: Re: New Rome (Mandelbox) Post by: David Makin on April 12, 2010, 11:13:14 AM Hi all, just trying a new ambient occlusion method for the Mandelbox. This is really cool !!! But something on the "busy" side. Kind of reminds me of the movie "Dark City", when the Strangers would modify the buildings while the people slept. (Looks like a few of yours are still in the process.) ;D It's only busy at this size - ideally it would be a 40" print or larger ;) Title: Re: New Rome (Mandelbox) Post by: Nahee_Enterprises on April 13, 2010, 12:53:51 AM It's only busy at this size - ideally it would be a 40" print or larger ;) Let me know when you have the "full size" image rendered, I would love to see it. :) Title: Re: New Rome (Mandelbox) Post by: David Makin on April 13, 2010, 03:44:17 AM It's only busy at this size - ideally it would be a 40" print or larger ;) Let me know when you have the "full size" image rendered, I would love to see it. :) I don't normally make the full-size versions available ;) So here's a taster, 640*480 detail from the original unresized 3840*2880 rrender: (http://www.fractalforums.com/gallery/2/141_13_04_10_3_39_45.jpeg) Of course a 40" print would be from a 12000*9000 render minimum, but I wouldn't increase the detail level any further. Title: Re: New Rome (Mandelbox) Post by: knighty on April 16, 2010, 05:37:05 PM Nice picture! :) Could we see just the AO effect? The method "Rectagular Ambient Occlusion" was suggested to me by Jos Leys. I've implimented the basic method such that the amount of occlusion at a given point is based on the maximum distance of the poiint from the centre of the cuboid along each axis (x/y/z). I also added an optional directional element. Here the basic method (location based only) on it's own for the "New Rome" fractal: /gallery/2/141_11_04_10_9_41_42.jpeg And here's the enhanced method using the directionality - as applied in the original "New Rome" render: /gallery/2/141_11_04_10_9_45_43.jpeg Note that the jet-black is not errors, it's holes through the Mandelbox to the background :) Of course the original "New Rome" is made up of the occluded ambient plus diffuse and specular from two light sources. Thanks! :) In case you are interrested, there is an AO technique based on Distance fields that was discribed by IQ in this presentation: http://iquilezles.org/www/material/nvscene2008/nvscene2008.htm (http://iquilezles.org/www/material/nvscene2008/nvscene2008.htm) Title: Re: New Rome (Mandelbox) Post by: David Makin on April 17, 2010, 02:24:12 AM Thanks for the link.
I confess I haven't looked but I suspect AO based on distance fields requires some sort of screen buffer ? I need a method that works solely on the information for each pixel in turn so very large renders up to say 12,000 by 9000 are possible in UF (without needing a new 64-bit UF and a 64-bit system). Title: Re: New Rome (Mandelbox) Post by: knighty on April 17, 2010, 04:16:27 PM No, it doesn't require any buffer.
From the document (p.48): "The idea: let p be the point to shade. Sample the distance field at a few (5) points around p and compare the result to the actual distance to p. That gives surface proximity information that can easily be interpreted as an (ambient) occlusion factor." He gives the foll0owing formula: AO=1-k*sum(i=1;i<=5;2^(-i)*(i*delta-DE(p+n*i*delta))) p is the point to shade. n is the unit normal vector at p. delta is the distance increment for sampling point. k is a factor for controlling the strength of the AO. Of course it's a fake AO (like SSAO or any fast AO) but it doesn't require a Z-buffer, it's quite fast and doesn't depend on the viewpoint. Title: Re: New Rome (Mandelbox) Post by: knighty on April 17, 2010, 05:53:55 PM Some results:
I used five samples and: k=5.5; delta=0.1 For the mandelbox I had to scale up the DE by a factor of two because the estimated distance is too small in this case. (http://www.fractalforums.com/gallery/2/932_17_04_10_5_45_17_0.jpeg) (http://www.fractalforums.com/gallery/2/932_17_04_10_5_45_17_1.jpeg) (http://www.fractalforums.com/gallery/2/932_17_04_10_5_45_17_2.jpeg) Title: Re: New Rome (Mandelbox) Post by: David Makin on April 17, 2010, 11:30:29 PM Now that's cool - it's also similar to an idea I was going to try - namely sampling the smooth iteration *density* at a point or points back along the ray towards the viewpoint and using that to give an AO value, though obviously IQ's method uses the normal as the direction and the DE value/s rather than the smooth iteration density :)
Thinking about it, of course going in the direction of the normal is definitely more correct ! Title: Re: New Rome (Mandelbox) Post by: Tglad on April 18, 2010, 02:04:13 AM Why are these areas so dark?
You would expect them to get a lot of surrounding light wouldn't you? Title: Re: New Rome (Mandelbox) Post by: Nahee_Enterprises on April 18, 2010, 02:46:06 AM I don't normally make the full-size versions available ;) So here's a taster, 640*480 detail from the original unresized 3840*2880 rrender: /gallery/2/141_13_04_10_3_39_45.jpeg Of course a 40" print would be from a 12000*9000 render minimum, but I wouldn't increase the detail level any further. Thanks for the closer peek. Now I know where all of those LEGOs went to. :D Title: Re: New Rome (Mandelbox) Post by: David Makin on April 18, 2010, 02:49:38 PM The rectangular and spherical AO methods produce results very close to IQ's distance field method:
"Ractangular AO" (http://www.fractalforums.com/gallery/2/141_18_04_10_2_46_52.jpeg) "Spherical AO" (http://www.fractalforums.com/gallery/2/141_18_04_10_2_47_24.jpeg) Title: Re: New Rome (Mandelbox) Post by: David Makin on April 18, 2010, 03:06:46 PM Sorry - I should have posted the parameters too :)
Ignore any warnings/errors - these will render with the latest released version of wip3D5: http://www.fractalgallery.co.uk/MMFwip3D.zip (http://www.fractalgallery.co.uk/MMFwip3D.zip) As always if downloading the above make sure the zip file is 95095 bytes, if it's smaller then you need to clear the older version from your internet cache. "Rectangular AO" Code: RectangularAO {"Spherical AO" Code: SphericalAO {Title: Re: New Rome (Mandelbox) Post by: knighty on April 18, 2010, 07:03:50 PM Why are these areas so dark? Yes!You would expect them to get a lot of surrounding light wouldn't you? There are two reasons: - Delta was too big. - Sampling is done along the normal vector. The formula I used gives generally good results with non fractal distance fields but don't seem to be good for fractals. I'm trying to find better formulas. :hurt: Here is the result i got while modifying the original formula: (http://www.fractalforums.com/gallery/2/932_18_04_10_5_52_48_0.jpeg) Humm... not perfect :-\ The rectangular and spherical AO methods produce results very close to IQ's distance field method: It is difficult to see the difference at first. In the picture above you can see the darkening at the corners which can't be obtained using rectangular AO.Title: Re: New Rome (Mandelbox) Post by: bib on April 18, 2010, 10:03:29 PM http://www.fractalgallery.co.uk/MMFwip3D.zip (http://www.fractalgallery.co.uk/MMFwip3D.zip) I had not been fractaling very intensively these days and I discover with pleasure your latest update. I am very enthusiastic about the work done on lighting. Title: Re: New Rome (Mandelbox) Post by: David Makin on April 18, 2010, 10:54:41 PM http://www.fractalgallery.co.uk/MMFwip3D.zip (http://www.fractalgallery.co.uk/MMFwip3D.zip) I had not been fractaling very intensively these days and I discover with pleasure your latest update. I am very enthusiastic about the work done on lighting. I think you'll find the best improvement is the rendering of the Mandelboxes/Juliaboxes when you use the analytical DE (with "Autobox" enabled). |