I've been looking at blobs on and off since Sven's December post to 
povray.bugs. There is a numerical accuracy issue which is real and 
briefly I thought I too was seeing something wrong in how finish 
elements were being handled.  However, it turned out I'd missed adding a 
texture on one of the negative blob components in an overall blob.

To my surprise POV-Ray includes textures attached to negative blobs 
along with those attached to positive blobs when it interpolates to a 
final texture. Further, if a negative component is missing a texture, 
POV-Ray will use the overall blob texture, if any, and if none is found, 
the running default texture. In the attached SvenFinishQ.jpg image the 
negative components missing a specific textures and thus affected are 
labeled as regions B and A, respectively.

So. Negative blob textures matter... This is a powerful feature. Namely, 
it is possible to create invisible ink, so to speak, on negative blobs 
which only become visible when interacting with portions of the blob, 
which are positive and greater than the blob threshold. Further, nothing 
stops us from making our positive blob components with clear textures!

In the BlobImages.jpg image we have four blobs intersected by one clear 
and 3 black boxes to keep  only the +Z portions of the blobs. These box 
intersections were done to avoid noise/image-echo from the interior 
backside of the blob surface. Further, in all four cases we have a 
vertical cylinder of a very small radius scaled left and right to get a 
reasonably flat shape after which we attach an image mapped texture.  In 
the upper left and lower right the image texture is attached to 
flattened cylinder(s) with positive strengths. In the upper right and 
lower left the image is attached to a slightly larger, but similarly 
flattened, negative strength cylinder. These negative components are 
enclosing the clear positive strength cylinder which turns the negative 
components visible due texture interpolation.

We see images in the upper right and lower left ONLY due the interaction 
of the negative, image, blob  component with the positive clear 
component. In the lower left image the negative strength has been 
cranked up enough so the left and right portions of the clear shape 
start to disappear. It would be easy to create an animation where the 
image shrinks from the left and right edges, until it finally goes poof, 
by ramping the strength more negative until the clear-catalyst shape 
disappears.

In the lower right we get rotational blur from the interpolation of 
positive, image textured, components variously rotated. If we wanted to 
create a rotating propeller like blur, we'd move the image textures onto 
negative blob components and introduce a clear component catalyst to 
turn the interpolated result visible. Not tried it yet, but I believe we 
can get propeller blur by rotating a non-flattened cylinder with a 
simple color texture attached.

Old blob, new tricks. :-)

Bill P.

Post a reply to this message

Attachments:
Download 'blobimages.jpg' (127 KB) Download 'svenfinishq.jpg' (99 KB)

Preview of image 'blobimages.jpg'

Preview of image 'svenfinishq.jpg'

Hi William,

I was reading your posting, but am not so sure if these new discoveries
on blobs and blob textures mean any improvement for my own blob problem.

Are you saying, I have to use one more texture for blobs, and in some
special way?

-----------

Another question: when I compiled an own PVEngine64.exe, and apply your
suggested change (2 becomes to 3), then the program crashes each time.
Did it function on your computer?

Post a reply to this message

On 01/14/2016 01:46 AM, Sven Littkowski wrote:
> Hi William,
>
> I was reading your posting, but am not so sure if these new discoveries
> on blobs and blob textures mean any improvement for my own blob problem.
>
> Are you saying, I have to use one more texture for blobs, and in some
> special way?

No.

I have since found the technique is tangled in the same numerical 
accuracy issue. Meaning, I often need my patched POV-Ray for such - 
clear-pos-texture-neg test scenes to work without artifacts.

>
> -----------
>
> Another question: when I compiled an own PVEngine64.exe, and apply your
> suggested change (2 becomes to 3), then the program crashes each time.
> Did it function on your computer?
>

I saw your posts, but I am unable to assist directly. I run on linux and 
patched the older stable branch invalid for a compile on windows. I too 
doubt your windows crash issue is related to the numeric change.

Are blob-less scenes of similar complexity and render size running fine 
with your compiled exe ? Does your crashing scene run fine with the 
pre-compiled binary to which Christoph pointed you?

I've been collecting & cleaning up problem blob scenes from years past 
with the hope the scenes might be of use at a later time to the core 
developers. Your problem blob, though simply textured, is in this 
collection.

The patched code(1) doesn't fix every blob issue I've re-discovered, but 
it does fix many - including FS288 as it turned out. Yet to see it break 
anything.

Aside: I forgot to add a credit for the photo used in my original 
posting. It's a press photo for the musical group "The Defibulators."

Bill P.

(1) - I have since moved to 1e-4 in my testing because this accuracy 
scaling (and AA as Alain noted at the time) was required for a scene 
Cousin Ricky posted back in 2008.

Post a reply to this message

Am 14.01.2016 um 15:57 schrieb William F Pokorny:

> I've been collecting & cleaning up problem blob scenes from years past
> with the hope the scenes might be of use at a later time to the core
> developers. Your problem blob, though simply textured, is in this
> collection.
> 
> The patched code(1) doesn't fix every blob issue I've re-discovered, but
> it does fix many - including FS288 as it turned out. Yet to see it break
> anything.
...
> 
> (1) - I have since moved to 1e-4 in my testing because this accuracy
> scaling (and AA as Alain noted at the time) was required for a scene
> Cousin Ricky posted back in 2008.

All this is valuabe input; thanks a lot for your continuing
investigation into this topic!

Post a reply to this message