Hello!
I have created an image with a transparent material getting very thin in some
places, and I can see color changes there (marked red in attached). How do I
avoid that?

The solid object is defined as:   pigment{color rgb <0.39, 0.41, 0.43>}

The transparent one as:   pigment {color rgbf <0.25,0.5, 1, 0.5>}

I also use:
global_settings{
    assumed_gamma 1
    radiosity{
        pretrace_start 1
        pretrace_end 1
        count 400
        recursion_limit 1
        nearest_count 1
        error_bound 0.05
    }
}
#default {finish { ambient 0 diffuse 1 }}

Post a reply to this message

Attachments:
Download '' (187 KB)

Preview of image ''

hi,

"spigor" <nomail@nomail> wrote:
> Hello!
> I have created an image with a transparent material getting very thin in some
> places, and I can see color changes there (marked red in attached). How do I
> avoid that?

have you tried off-setting either object by a very small amount (an "epsilon") ?


regards, jr.

Post a reply to this message

On 28.10.2024 12:59, spigor wrote:
> Hello!
> I have created an image with a transparent material getting very thin in some
> places, and I can see color changes there (marked red in attached). How do I
> avoid that?

Looks like POVRay cannot decide which surface exactly the ray hits. As 
already suggested by jr, it's worth a try forcing POVRay to make a 
choice by slightly moving objects so that the surfaces do not touch but 
have clear intersection with non-zero thickness.

-- 
Ilyich the Toad
https://dnyarri.github.io/

Post a reply to this message

Le 2024-10-28 à 05:59, spigor a écrit :
> Hello!
> I have created an image with a transparent material getting very thin in some
> places, and I can see color changes there (marked red in attached). How do I
> avoid that?
> 
> The solid object is defined as:   pigment{color rgb <0.39, 0.41, 0.43>}
> 
> The transparent one as:   pigment {color rgbf <0.25,0.5, 1, 0.5>}
> 

What they described to you is what is called coincident surfaces.
That happen when two or more surfaces are touching.
Try moving the grey object down by a small distance.

Add something like
translate<0, -0.00001, 0>
to the grey object, or
translate<0, 0.00001, 0>
to the transparent object.

Post a reply to this message

"spigor" <nomail@nomail> wrote:
> Hello!
> I have created an image with a transparent material getting very thin in some
> places, and I can see color changes there (marked red in attached). How do I
> avoid that?

In my experience, transparent surfaces must be separated by at least 0.001 units
to avoid such artifacts.

Post a reply to this message

"Cousin Ricky" <rickysttATyahooDOTcom> wrote:
> "spigor" <nomail@nomail> wrote:
> > Hello!
> > I have created an image with a transparent material getting very thin in some
> > places, and I can see color changes there (marked red in attached). How do I
> > avoid that?
>
> In my experience, transparent surfaces must be separated by at least 0.001 units
> to avoid such artifacts.

Hi
Thank you All for your input, it's much appreciated.
The artifacts disappeared with 0.05 units separation.

spigor

Post a reply to this message

"spigor" <nomail@nomail> wrote:
> "Cousin Ricky" <rickysttATyahooDOTcom> wrote:
> > "spigor" <nomail@nomail> wrote:
> > > Hello!
> > > I have created an image with a transparent material getting very thin in some
> > > places, and I can see color changes there (marked red in attached). How do I
> > > avoid that?
> >
> > In my experience, transparent surfaces must be separated by at least 0.001 units
> > to avoid such artifacts.
>
> Hi
> Thank you All for your input, it's much appreciated.
> The artifacts disappeared with 0.05 units separation.
>
> spigor

radiosity{
        :
        error_bound 0.05
    }
It looks like the separation must be at least as big as the error_bound.
Reducing the error_bound reduces the size of artifacts but increases the
rendering time.

Post a reply to this message

Le 2024-10-28 à 05:59, spigor a écrit :

> I also use:
> global_settings{
>      assumed_gamma 1
>      radiosity{
>          pretrace_start 1
>          pretrace_end 1
>          count 400
>          recursion_limit 1
>          nearest_count 1
>          error_bound 0.05
>      }
> }

That radiosity block make me twitch.
pretrace_start and pretrace_end 1 are only to be used for the last pass 
of multipass rendering when all of the radiosity data have been saved 
from a previous pass.
Even the default of 0.08 and 0.01 are to large. Using pretrace_end 
0.0025 or less tend to give better result, and is often faster overall 
as more, and better, data are collected during pretrace.

count is OK. Using the two value version will often improve the quality 
at nearly zero cost.
count 400, 1111

error_bound is acceptable for a scene using only radiosity illumination. 
If you use conventional light, then the default of 0.5 is sufficient.

nearest_count 1 will often cause artifacts. Better to use the adaptive 
version : nearest_count 20, 3


You should NOT use diffuse 1. It can cause some nasty side effect. 
Better keep the default value.
So, change that #default to just :
#default{finish{ambient 0}}

And use emission 1 in the finish block to create some light emitting object.

Post a reply to this message

On 2024-10-30 04:20 (-4), spigor wrote:
> 
> radiosity{
>         :
>         error_bound 0.05
>     }
> It looks like the separation must be at least as big as the error_bound.
> Reducing the error_bound reduces the size of artifacts but increases the
> rendering time.

This has not been my experience.  I very rarely use error_bound lower
than 0.5, and the transparency artifact is independent of whether or not
I use radiosity.

Post a reply to this message

On 2024-10-30 08:41 (-4), Alain Martel wrote:
> 
> error_bound is acceptable for a scene using only radiosity illumination.
> If you use conventional light, then the default of 0.5 is sufficient.

The default error_bound is 1.8, although 0.5 is what I normally use.

> You should NOT use diffuse 1. It can cause some nasty side effect.
> Better keep the default value.
> So, change that #default to just :
> #default{finish{ambient 0}}

What sort of side effects?  I sometimes use diffuse 1, for example, with
spectral rendering, but when I do, I use darker pigment colors.  I have
never seen problems unless the diffuse times a color channel value was
greater than 1.0.

Post a reply to this message