Most of my media experience is with emitting media. In an attempt to 
better understand scattering media I spent some time looking at our 
abyss.pov sample scene. I knew it to have some significant media shadow 
artifacts due the media settings. Namely multiple intervals of 3 and 
default sampling 3 and recursion 3. Turned out to be a way to look at 
assumed_gamma settings and the +ag anti-alias control with respect to 
highlights too.

Attached are two images. One is a study comprising 9 images:

1) The original 3.6 abyss.pov as rendered by 3.6.1.
2) (1) but modifying the media container so the light is outside.
3) (1) but changing the assumed_gamma to 1.0. You can get (3) to (1) 
easily in external software like gimp by applying a 1/2.2 gamma 
correction - except for tiny highlights and AA differences.
4) Is our current 3.7+ scene very like (3) except the camera view changed.
5) Some media corrections and camera view restored base upon (4).
6) (5) but modifying the media container so the light is outside.
7) (5) but gamma adjusting colors for assumed_gamma 1.0.
8) (7) but modifying the media container so the light is outside.
9) For artistic reasons using assumed_gamma srgb +ag4.4 for highlights / AA.

The other image is just comparing (1) to (9) at a larger resolution. I 
was thinking about submitting (9) on github as a replacement for our 
current abyss.pov sample scene, but I'm unsure? Probably posting my 
annotations here is enough for folks to find them on a search.

I expect Gilles knew the path he took to his artistic result. While my 
result (9) might better fit scattering media to the scene geometry in a 
way avoiding artifacts - it changes the look substantially.

Anyway, adding the very top of bottom of my current updated and 
annotated abyss.pov. I'd appreciate any review and especially 
corrections to my now somewhat improved understanding of scattering 
media behavior and +ag feature use.

Bill P.


Corrections not shown:
- Camera view restored to Gilles original. Added aspect warning check.
- Position of 'Persistence' text correct as top previously clipped.
- Rounded isosurface max gradients up as with some AA / media settings 
the current scene tripped max gradient warnings.

// Persistence Of Vision raytracer sample file.
//============================================
// The field, new improved version  October. 2001
// Copyright Gilles Tran 2001
// http://www.oyonale.com
// Simplified version of scene updated and annotated
// for v3.8. Feb 10, 2018.
//--------------------------------------------
// Render with a 2.67 ratio such as 320*120, 640*240,
// 1024*384, 1280*480
//--------------------------------------------
// -w320 -h120
// -w640 -h240 +a0.1
// -w768 -h288 +a0.1
// -w1024 -h384 +a0.1
//
//
// Note. For high quality result suggest adjusted media settings in
// comments at bottom of this file. Also aggressive adaptive
// anti-aliasing with dithering to avoid banding in the output
// image:
//     +a0.1 +am2 +r4 +thfs
//
// Note. Uncomment AreaOK=true below to turn on the area light. This
// will blur the shadow under the submarine but the rendering time
// will be extremely slow. The area light also compensates to some
// degree for any media under-sampling being itself a kind of
// additional sampling.
//
// Note. An 'assumed_gamma srgb' is being used for an artistic final
// result. With media especially, scene development is less
// confusing if done at the usual assumed_gamma of 1.0. At 1,
// densities and adjustments are more aligned with the visual
// result. For assumed_gamma other than 1.0 the visual result is
// distorted relative to the actual media densities and colors.
//
// Note. The srgb assumed_gamma essentially adds a second
// pow(color,2.2) adjustment from the internal color values on top
// of the usual one for the output file and display. POV-Ray's
// internal anti-aliasing works with a default 2.5 gamma in line
// with the usual srgb output. To handle colors - and highlights
// especially - in a manner aligned to the srgb color working space
// use +ag4.4 or similar. It's costly run time wise with deep
// adaptive anti-aliasing.

#version 3.8;

global_settings{ assumed_gamma srgb }

... (bulk of scene omitted some reformatting for posting) ...

//--------------------------------------------
// submarine and media
//--------------------------------------------
union{
   object{Submarine scale 3/4 translate z*-10 translate y*10}

//Original media container was sphere{0,1...
//In the update for the v3.8 release now using a difference so the
//light interacting with the media is outside the container.
//Perhaps intentional, perhaps not, but since v3.5 it has been the
//case with shape contained media the light being outside turns on
//media_attenuation and when inside media_attenuation is off no
//matter the setting on the light itself. When attenuation is off
//this image appears washed out.

   difference {
     // x,z scale by 2 eliminates shadow artifacts at rear of
     // media container.
     sphere{0,1 scale <2,1,2>}
     sphere{<0,1,0>,1/1.5}
     texture{pigment{Clear}}
     scale 410 hollow
     interior{
      media{
           //   Note. Tweaked media over original to not depend on
           //   weak sampling and artifacts for result.
              scattering {5,0.00133 eccentricity 0.90 extinction 1.0}
           //   Note. Original took mostly green out of the
           //   scattered light result by absorption.
           // absorption <255-23,255-171,255-239>*0.0005/255
           // intervals 1
           // method 3
           //   Note. Samples should be chosen with the media range
           //   and the size of the shapes casting shadows in mind.
           //   The old scene samples 9 equivalent worked for the sub
           //   itself, though with significant shadow artifacts, due
           //   the camera to inside media container distance being
           //   about 2000 and the sub length being maybe 700 as
           //   rotated. 2000/700 ~= 3.
           //
           //   Camera to far extent range for updated container ~=
           //   6225. Min rectangular sub extent about 130. 6225 / 130
           //   ~= 48.225. Picked 67 to allow for sub shape.
           // samples 67
           //   Note. Use the settings below for high quality renders.
           //   Need to be aggressive with adaptive sampling to
           //   minimize banding in result. A higher samples setting
           //   necessary to better resolve shadows given dimensions.
           //
           //   Min rectangular mine body extent about 13. 6225 / 13
           //   ~= 480. Picked 667 to allow for round mine body.
           // samples 667
           // aa_level 9
           // aa_threshold 0.05
      }
     }
   }

   scale 4.0
   rotate y*rotSub  translate posSub

Post a reply to this message

Attachments:
Download 'abyssbeforeafter.jpg' (172 KB) Download 'abyssmediastudy.jpg' (222 KB)

Preview of image 'abyssbeforeafter.jpg'

Preview of image 'abyssmediastudy.jpg'

That's a fascinating series of tests and code-fix changes you've come up with.
When I used to run the 'abyss.pov' file in v3.6.2 days, I too noticed that it
didn't look 'correct', in some hard-to-grasp way. (Not that the scene itself is
the problem; it's a beauty, regardless of any undelying problems or
media-interaction strangness.) At the time, I attempted to finesse some of the
scene values, but was never happy with my own results.

Over and above your technical/artistic fixes, I'm wondering which render version
Gilles himself would prefer-- the original, or your nicely murky and darker
final render. Of course, he made the scene based on the behavior of POV-Ray at
the time-- warts and all. ;-) But I wonder if an older render represents the
'look' he wanted to achieve-- OR, if that look was based on limitations that he
couldn't work around.

In any case, thanks for taking to time to delve into that complex scene--
especially regarding light/media interaction-- and for coming up with the
proposed fixes. That scene is a real showcase for POV-Ray.

Post a reply to this message

In your "abyss media study" image, I'm curious about 1) vs. 3).

When Gilles originally wrote the scene in v3.6.x (and posted an image, I
assume), was he rendering it with assumed_gamma 1.0 then, or 2.2? (Those years
were unclear times as far as gamma was concerned!)  Your 1) image *appears* to
show a gamma 2.2, yet the scene file itself (in my original 3.7.0 download)
shows it as 1.0. If he did run it at 1.0, it would match you image 3) rather
than 1)-- unless there was something really wonky about v3.6.x ! I'm wondering
if his scene file may have gotten updated from 2.2 to 1.0 at some time in the
past?

Post a reply to this message

On 02/21/2018 06:47 AM, Kenneth wrote:
> In your "abyss media study" image, I'm curious about 1) vs. 3).
> 
> When Gilles originally wrote the scene in v3.6.x (and posted an image, I
> assume), was he rendering it with assumed_gamma 1.0 then, or 2.2? (Those years
> were unclear times as far as gamma was concerned!)  Your 1) image *appears* to
> show a gamma 2.2, yet the scene file itself (in my original 3.7.0 download)
> shows it as 1.0. If he did run it at 1.0, it would match you image 3) rather
> than 1)-- unless there was something really wonky about v3.6.x ! I'm wondering
> if his scene file may have gotten updated from 2.2 to 1.0 at some time in the
> past?
> 
> 

I have copies of POV-Ray source and distributions going back in time. I 
have run-able POV-Ray versions back to v3.5 something (c ?).

What I did for the three v3.6.1 renders was use a version of the scene I 
pulled from my copy of v3.6.1 rendering those with a corresponding 
v3.6.1 compile. That version of the scene specified no gamma at all 
meaning it ran at an assumed_gamma of 2.2.

v3.6.1 supports assumed_gamma 1.0, but the abyss scene (Gilles called 
his 2001 version "The Field") is older - with a history going back into 
the 1990s. Gilles created the abyss.pov version of "The Field" 
especially for distribution with POV-Ray and I first see it in my older 
code bases at v3.5.

See too: http://www.oyonale.com/3D.php?lang=en&section=2001

(Oddly, a couple of the links off the abyss/The Field image page there 
are not working for me today and they did a couple weeks back...)

Anyway, the scene was indeed modified by the time it was shipped with 
v3.7.0 moving to 'assumed_gamma 1.0' and a changed camera definition - 
and by side effect, I'd guess, a changed camera view - but nothing else.

Bill P.

Post a reply to this message

William F Pokorny <ano### [at] anonymousorg> wrote:
>
> What I did for the three v3.6.1 renders was use a version of the scene I
> pulled from my copy of v3.6.1 rendering those with a corresponding
> v3.6.1 compile. That version of the scene specified no gamma at all
> meaning it ran at an assumed_gamma of 2.2.
>

No assumed_gamma at all, wow. (That was before my time-- ha-- as I started using
POV-Ray at v3.6.1, IIRC.)

So what Gilles himself saw when he created the scene was like you image 1). Got
it.

Funny thing: If *I* had created this scene (not likely!), I think I would have
chosen the more greenish color 'tint' of image 5)-- coupled with your nice fixes
in 9) of course. I just happen to like the mysterious more-green-than-blue look,
personal preference.

Post a reply to this message

Am 21.02.2018 um 14:36 schrieb Kenneth:

> So what Gilles himself saw when he created the scene was like you image 1). Got
> it.

Actually, I guess nobody except Gilles knows for sure what he himself
saw. Unless he happened to be using a calibrated display.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:

> Actually, I guess nobody except Gilles knows for sure what he himself
> saw. Unless he happened to be using a calibrated display.


Others preferred a blueish tint too - like in Pixar's "Finding Nemo"
(https://clementsnahs.files.wordpress.com/2015/08/mines4.jpg).

Hopefully Gilles T. got paid for it.


Norbert

Post a reply to this message

"Norbert Kern" <nor### [at] t-onlinede> wrote:
>
> Others preferred a blueish tint too - like in Pixar's "Finding Nemo"
> (https://clementsnahs.files.wordpress.com/2015/08/mines4.jpg).
>

Ha! So they do.

My humble opinion about *that*: A blue tint is fun and upbeat and not scary--
good for kiddies. A murky green tint is mysterious and foreboding-- like scary
submarine warfare.

;-P

Post a reply to this message

Am 22.02.2018 um 14:53 schrieb Kenneth:
> "Norbert Kern" <nor### [at] t-onlinede> wrote:
>>
>> Others preferred a blueish tint too - like in Pixar's "Finding Nemo"
>> (https://clementsnahs.files.wordpress.com/2015/08/mines4.jpg).
>>
> 
> Ha! So they do.
> 
> My humble opinion about *that*: A blue tint is fun and upbeat and not scary--
> good for kiddies. A murky green tint is mysterious and foreboding-- like scary
> submarine warfare.

https://youtu.be/mpsFaLH5BRc

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:


>
> https://youtu.be/mpsFaLH5BRc

One of my favorite films, especially the extended "director's cut".

Post a reply to this message