On 10/12/2016 01:40 PM, omniverse wrote:
>
> I wouldn't have guessed right how you were doing that, thanks for explaining.
>
> Occurred to me that I wasn't using the blob 'sturm' keyword and gave that a try,
> not using AA or focal blur this time. Made matters worse, and I mean sturm did
> not just removing the AA.
> Closer look in images shown here.
>

Believe it or not, that sturm makes the fringe effect worse where blobs 
are used as a media container is a known issue. I avoided sturm knowing 
this.

I also cheated a little on my overall blob container making the 
containing blob slightly larger than the set-union of all the blob 
components making up the blob-color medias. This can often be done with 
the threshold alone by making the containing blob threshold a tiny bit 
smaller, such as the threshold minus 1e-4 or 1e-5. For my image I used 
turbulence and so needed to make the radius on my containing cylinders a 
little larger still by simply increasing the radius. Increasing the 
containing blob component radius upward is naturally another way to up 
the size of the containing blob.

Making the container a little larger doesn't eliminate the bad fringe 
media intervals, but rather it tends to move them outward to where there 
is a density of 0 - often helping with bright speckles at least.

As for the media fringe speckles themselves, I have the thought perhaps 
we could test media intervals for sanity. Is the mid-point of the range 
inside the container; Are points just a little beyond the two end points 
both testing as outside the container; ahead of sampling and calculation.

I've never examined the media code - so this ray interval sanity test 
thought is perhaps crazy. Such testing would certainly degrade 
performance for media itself, but, we are today adding run time outside 
by other means to address the fringe effects.

Mixed in this too is my softly proposed (I've not yet made it a pull 
request) blob code tuning from early this year. A branch with my current 
tuning value change is at:

https://github.com/wfpokorny/povray/tree/tune/blob_accuracy

The potential pattern is brand new. I am learning about its behavior 
too. Lots I haven't tried about which I wonder. Negative blobs; 
Semi-transparent containers textured via the uniquely flexible 
negative/positive blob component methods. With isosurface based 
potentials we can of course manipulate the container and media together. 
What about blob based potential patterns as functions defining 
isosurfaces.

Lastly, could I please get a copy of your last scene to add to my blob 
artifact cases? It happens to be somewhat unique to those I have in my 
collection in having lots of both bright and dark speckles.

Bill P.

Post a reply to this message