[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-bugs-dist
Subject:    [palapeli] [Bug 337972] IDEA: Non-uniform piece size based on image detail
From:       Matthew Woehlke <mw_triad () users ! sourceforge ! net>
Date:       2014-08-04 16:01:54
Message-ID: bug-337972-17878-F60b4ArSwk () http ! bugs ! kde ! org/
[Download RAW message or body]

https://bugs.kde.org/show_bug.cgi?id=337972

--- Comment #3 from Matthew Woehlke <mw_triad@users.sourceforge.net> ---
(In reply to Ian Wadham from comment #2)
> (In reply to Matthew Woehlke from comment #0)
> > I had a dream (not the M. L. King kind, but the waking-from-sleep kind) this
> > morning, in which there was a jigsaw puzzle. This puzzle was unusual for
> > having pieces of widely varying sizes, based on the local level of detail.
> 
> Hopefully not of the Stephen King kind... But what would be the objectives
> of such a puzzle from the user's point of view?

Does it *need* an objective other than being interesting and fun? :-)

> Seems to me it -might- have effects like:
> 
>  - Evenly colored objects such as sky become easier, through having fewer
> pieces,
>  - Fractal type objects such as trees become harder, through having lots of
> pieces covering similar-looking details,

Yes, I think both of these would happen.

>  - Non-fractal type objects such as buildings become easier, through having
> details segregated onto one piece, rather than randomly split.

Ideally this wouldn't happen; places of "constant" detail should still end up
split more or less randomly. It's more the above two points that are intended.

> You might also be interested in these links, which I found via Google images
> for "Voronoi grid":
> http://www.swstechnology.com/blog/voronoi-cells-and-modflow-usg
> http://www.cs.cmu.edu/~quake/triangle.html


Voroni is pretty well understood (though the first link is about as good a
reference as I've seen; thanks).

Also interesting:


http://www.cnblogs.com/pulas/archive/2013/04/02/2994922.html (non-uniform
density Poisson disk sampling)
http://bost.ocks.org/mike/algorithms/ (lots more on Poisson disk sampling;
FASCINATING read!)

> > There are two other desired refinements (actually, these would apply to the
> > existing irregular slicer also, if it doesn't use them already). First, for
> > each cell vertex, find the largest angle Ī±, and compute a weighted random
> > chance (Ī±=180 ° ā†’ pā‰ˆ0.9, Ī±=90 ° ā†’ pā‰ˆ0.3) of forming the edges of said angle to
> > be G2 continuous at that point. Second, for each edge, compute a weighted
> > random chance (A=1.6 ā†’ p=1, A=2.5 ā†’ pā‰ˆ0.1, with A = the ratio between the
> > edge length and the nominal plug (minimum?) width) of omitting a plug for
> > that edge.
> 
> I don't really understand the purpose of the first refinement, but would the
> second refinement help eliminate the odd nipples and cavities that appear on
> short edges in the Voronoi-based slicer?

Yes, that was largely what inspired the second. I'll add an SVG illustrating
both effects shortly.

-- 
You are receiving this mail because:
You are watching all bug changes.=
[prev in list] [next in list] [prev in thread] [next in thread]