in news:3bd49d4f@news.povray.org Nathan Kopp wrote:

> Actually, I was planning to fix the granite pattern so that it
> produces similar results for all three noise generators.  Any
> comments? 

 For me it would be enough to change the default noise_generator to 1, 
instead of the current 3. This would maintain backward scene 
compatibility and give builders of new scenes the possibility to play 
with the noise_generator in combination with granite/wrinkles.
 Additionally Christophe could ad an extra "//noise_generator 3" to the 
scene templates to make it clear to the user that he can change the 
behavour.

Ingo

-- 
Photography: http://members.home.nl/ingoogni/
Pov-Ray    : http://members.home.nl/seed7/

Post a reply to this message

ingo wrote:
> 
>  For me it would be enough to change the default noise_generator to 1,
> instead of the current 3. This would maintain backward scene
> compatibility and give builders of new scenes the possibility to play
> with the noise_generator in combination with granite/wrinkles.

I don't think making 1 default would be good.  Most beginners will not
care about the noise generator and therefore not use the new one which
will lead to the old one staying the de-facto standard.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Christoph Hormann wrote:
> 
> ingo wrote:
> >
> >  For me it would be enough to change the default noise_generator to 1,
> > instead of the current 3. This would maintain backward scene
> > compatibility and give builders of new scenes the possibility to play
> > with the noise_generator in combination with granite/wrinkles.
> 
> I don't think making 1 default would be good.  Most beginners will not
> care about the noise generator and therefore not use the new one which
> will lead to the old one staying the de-facto standard.

The point being however is that method 1 HAS been the de-facto standard
for 10 years now. The new default breaks a lot of peoples existing scene
files including the behavior of several long standing textures in the
standard includes. You are right that a lot of beginners are not going
to care which noise generator is used. They have a lot of other things
to learn before they get that far. When they do they will start reading
the docs on the more advanced options available and will experiment with
the other methods available.

It seems to me it is a hand full of advanced users that are worried
which default will be used (more for their own selfish reasons) than
for concern about new users :)

-- 
Ken Tyler

Post a reply to this message

Wasn't it Nathan Kopp who wrote:
>
>"ingo" <ing### [at] homenl> wrote...
>> Histograms of the different noise_generators, one with the granite
>> pigment, the other with pigment{function{f_noise3d}}. Images rendered
>> to gamma=1.
>>  Maybe usefull in the docs, section "6.7.12.4".
>
>Actually, I was planning to fix the granite pattern so that it produces
>similar results for all three noise generators.  Any comments?  Also, would
>the wrinkles pattern need to be "fixed" too?

If you're going to "fix" wrinkles, then I reckon that you should do
"dents" as well.

There seem to be four groups of patterns to consider:-

Granite: Considerably different in a way that visibly alters existing
textures.

Dents and Wrinkles: Considerably different, but since they're mainly
used for normals the effect is far less noticeable.

Bozo and Bumps: Considerably different, but the point of the new noise
generators was to fix the existing plateaux problem with these patterns.

Everything else: Not affected.

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message

"Ken" <tyl### [at] pacbellnet> wrote in message
news:3BD56A04.E20EC0F9@pacbell.net...
>
>
> It seems to me it is a hand full of advanced users that are worried
> which default will be used (more for their own selfish reasons) than
> for concern about new users :)

I'd prefer it being the new "fixed" way by default.  This might be a good
reason, advanced versus new users, since people familiar with the program
can easily change it where applicable.  Only thing that hampers this idea is
when new people use old scene files, that would be the primary concern here.
But like I said, noise_generator 3 would be my choice.  Or beyond that,
anything else which is the better overall if it is ever changed before
POV-Ray v3.5 is final.

Bob H.

Post a reply to this message

"Mike Williams" <mik### [at] nospamplease> wrote...
>
> If you're going to "fix" wrinkles, then I reckon that you should do
> "dents" as well.

A little clarity to the nose_generator issues:
Inside POV there are two types of noise: DNoise and Noise.  DNoise returns a
vector, Noise returns a scalar.  Noise is the one that is affected by the
noise_generator.  DNoise is not affected by the noise generator.

Noise (as opposed to DNoise) is used in the following places:
* Isosurface functions:
  - f_hetero_mf
  - f_ridge
  - f_ridged_mf
  - f_noise3d
  - f_noise_generator
* dents
  - When used as a NORMAL, a combination of Noise and
     DNoise is used
  - When used as a PATTERN, the output of dents is Noise^3 (cubed)
* wrinkles
  - When used as a NORMAL, Noise is NOT used.
  - When used as a PATTERN, the Noise function is called
     10 times, and the results are combined by addition.
* crackle solid
* granite
  - Noise is called 6 times in a loop, and the results
     are combined by addition.
* bozo, spotted, bumps patterns
  - bumps only when used as a PATTERN (when used
     as a NORMAL, it uses DNoise)
  - bozo, spotted, and bumps patterns are all direct
     calls to Noise.
* Turbulence (but not DTurbulence)

Note that only the pattern (pigment) versions bumps and wrinkles are
affected.  The normal versions of these functions are not affected

Some patterns appear to have been "tweaked" to use the strange distribution
pattern of the old Noise function.  Such patterns could thus use a
"re-tweaking" so that they work well under all three noise generators.
Those patterns are:
* wrinkles as a pigment
* granite
* possibly dents

Note: Turbulence has already been "re-tweaked" (though there is a bug in the
re-tweaking that needed to be fixed).

The other functions probably shouldn't be "re-tweaked" because it appears
there is no tweaking in the current implementation and they are expecting
plateau-free noise (which is produced by noise generators 2 and 3).

Re-tweaking is easy.  It is done by taking the output from the new Noise
function, multiplying by 2 and subtracting 0.5.  Then clipping (to the
interval [0,1]) is done if necessary.  I applied this method to granite to
produce the attached results.

Attached is an image of the re-tweaked granite rendered using the three
noise generators.  Note that noise_generator 1 has not changed and produces
the old 3.1-style output perfectly.  It is interesting to note the clear
patterns that emerge from noise_generator 3.  This seems to confirm some
reports that noise_generator 3 doesn't seem to be very random.

-Nathan

Post a reply to this message

Attachments:
Download 'noise_generators.jpg' (62 KB)

Preview of image 'noise_generators.jpg'

Wasn't it Nathan Kopp who wrote:

> It is interesting to note the clear patterns that emerge from 
> noise_generator 3.  This seems to confirm some reports that 
> noise_generator 3 doesn't seem to be very random.

I'd noticed some subtler patterning in some of my own tests, but your
image shows it much more clearly.

It had been noted on another thread that noise_generator 3 produces non-
random results at integer positions. I guess that the grid pattern
visible in this image may be related to integer positions (possibly
scaled down by the code inside the granite pattern).

It looks to me like the stuff between the grid lines is nicely random,
but that the values returned from points on the grid are discontinuous
from the stuff between the lines. I thought that perlin noise was better
than that. Is it possible that there's a bug in the noise_generator 3
code when called for integer locations?

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message

The noise you used is scaled too large to give correct histograms. To reduce
statistical errors a large area of noise has to be evaluated. Therefore a
scaled down noise should be rendered at high resolution.

Roy

Post a reply to this message

Nathan Kopp wrote:
> 
> [...]
> 
> The other functions probably shouldn't be "re-tweaked" because it appears
> there is no tweaking in the current implementation and they are expecting
> plateau-free noise (which is produced by noise generators 2 and 3).
> 
> Re-tweaking is easy.  It is done by taking the output from the new Noise
> function, multiplying by 2 and subtracting 0.5.  Then clipping (to the
> interval [0,1]) is done if necessary.  I applied this method to granite to
> produce the attached results.
> 
> Attached is an image of the re-tweaked granite rendered using the three
> noise generators.  Note that noise_generator 1 has not changed and produces
> the old 3.1-style output perfectly.  It is interesting to note the clear
> patterns that emerge from noise_generator 3.  This seems to confirm some
> reports that noise_generator 3 doesn't seem to be very random.
> 

I made a another test image with various patterns involved:

http://www.schunter.etc.tu-bs.de/~chris/files/noise_generator_A.png

The structures in granite with noise_generator 3 are really quite obvious,
In the other patterns they seem to be at least less clearly visible.  

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

Christoph Hormann wrote:
> 
> The structures in granite with noise_generator 3 are really quite obvious,
> In the other patterns they seem to be at least less clearly visible.
> 

In fact they are visible, here some detailed samples:

http://www.schunter.etc.tu-bs.de/~chris/files/wrinkles1.png
http://www.schunter.etc.tu-bs.de/~chris/files/wrinkles3.png
http://www.schunter.etc.tu-bs.de/~chris/files/dents1.png
http://www.schunter.etc.tu-bs.de/~chris/files/dents3.png
http://www.schunter.etc.tu-bs.de/~chris/files/granite1.png
http://www.schunter.etc.tu-bs.de/~chris/files/granite3.png

You can also recognize some star like structure around the origin in the
old noise generator.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message