Subdividing boundary edges

[beeemtee]

Hi,

in Houdini both the Subdivide SOP and the "Polygons As Subdivision" render option seems to "protect" boundary edges, so their position doesn't get smoothed. A rectangular grid for example keeps its shape and doesn't become an oval. Is it possible to change this behavior and not "protect" boundary edges?

thanks in advance

bmt

[lavadht]

The SOP's manipulate > smooth-node smoothens an input mesh and it's boundary edges. It doesn't add subdivisions however.

One way to do both, is to create a Foreach node, set to work as "each number", in it the following structure could be added;

(the method I explain here, with subdivisions and without an oversmoothened result, will probably only work on a basic rectangular shape like a grid - but it may still be of help)

SOP

>FOREACH

_> each1 (input)

_> smooth1 (operation 1)

_> to chop (operation 2)

CHOPNET (operation 2 explained)

> geometry1 (get each1's channels)

> geometry2 (get smooth1's channels)

> fetch1 (fetch Y from geo1)

> fetch2 (fetch X and Z from geo2)

> merge (fetch1 + fetch2 = XYZ)

_> channel (read CHOP data, apply to smoothened but undivided geometry)

_> subdivide

Then, the ammounts of subdivisions, and degree of smoothing, will be dependend on the itteration number of the foreach node, and possibly the relative ammounts of smooth vs subdivision nodes (or, preferably just the settings of these nodes) within that foreach node. Using a math-node (in the chopnet) to average the Y positions of the 'each1' and the 'smooth1' node, instead of only taking the Y position of the initial input, will decrease the ammount of smoothing; which I noticed get's quite extreme when only smooth and subdivision nodes are used.

Edited December 6, 2010 by lavadht

[beeemtee]

Hi Laurens,

and thanks for your answer. If I get it right you suggest to use an alternating sequence

of a Subdivide and a Smooth operator. While this would surely produce an oval from a grid,

in my opinion it's more like a hack than a real subdivision method. I mean, whith

this approach it's probably very hard to match the exact result of

a subdivision algorithm like PRMan's or Maya's.

To allow me to check my preliminary intuition, would you post a hip file containing the net

you've described?

thanks

mate

[lavadht]

Hi Máté,

Looks like you're right. I thought it would be a relative simple setup and didn't bother to check its appearance relative to a true circle; It seems quite difficult to get an accurate match, and even if it's possible, the fiddling with parameters would certainly label it as a hack.

Still, the method could be used depending on the subject and the targeted accuracy of it.

About the attached file;

The node set as visible is the main test I did; it shows a square which deforms to an approximation of a circle - as well as a star.

To the right, there's a wiremodel of a circle, which acts as a reference for a 'perfect circle'.

To the left, there's a more complex input, which I added to further test whether it'd be worthwhile to tweak the parameters in order to get exactly that which you are after - and it appears it's not.

EDIT; I just had to use the subdivide OP on a non solid geometry myself - and I think I'm misunderstanding your problem; as boundary edges as I thought I knew them do subdivide and change their position/shape. My simple use is a box with one face deleted.

What's the context in which you're bumping into this problem? Care to share your working file or a part of it?

SmoothDivide.hip

Edited December 6, 2010 by lavadht