(Ping ...?) Nothing out there? Really ...??
--
To post to this group haxe...@googlegroups.com
http://groups.google.com/group/haxelang?hl=en
---
You received this message because you are subscribed to the Google Groups "Haxe" group.
For more options, visit https://groups.google.com/d/optout.
I'm confused by what you're looking for. I just did a search for "image pyramid" because I'd never heard of it before and came up with this:
http://docs.opencv.org/doc/tutorials/imgproc/pyramids/pyramids.html
As far as I can tell it is simply a "stack" of resized versions of an image no? Hence the suggestion of the resized above?
What's the purpose of it? Are you trying to make tiles for a map?
As far as I can tell it is simply a "stack" of resized versions of an image no? Hence the suggestion of the resized above?
On Mar 25, 2015 10:31 AM, "Mike Robinson" <sundial....@gmail.com> wrote:
Nope, I don't readily see how "cubic interpolations" have anything whatsoever to do with the construction of map-tile pyramids . . . am I missing something?
On Tuesday, March 24, 2015 at 8:31:39 PM UTC-4, JLM wrote:Is this java relevant?
http://www.paulinternet.nl/?page=bicubic
Would need to study it some more to work out using it for an image but I think it could work.
Bicubic interpolation is cubic interpolation in two dimensions.
I'll only consider the case where we want to interpolate a two
dimensional grid." 4x4 matrix.
The concept is that using matricies you can generate points in a
sparce matrix ( fill in holes ) using predictive equations
(interpolation), basically your applying an equation to a matrix
so taking say a grid of 16 from your image and apply an equation
to predict other points then repeating for each color so the
equation essentially adds weighting to points and tries to curve
fit them to a cubic. in real coding terms matrices are often
pretty inefficient way to calculate something fast, better to
break down the actual operations and more or less just pass them
in the output, without bothering with creating generalized matrix
code structures and operators. The link was one referenced on a
page about up and down sampling but one that contains real code
and not just matrices, and seems to be doing this and is better
than say averaging nearest or some of the other common but simpler
approaches.
The code seems to be setup to upsample ( add intermediate points
to the data ) I think ( but not sure ) between points although
probably overlapping 4x4 grid would be needed and overlaps
averaged or similar to process the whole image experimentation is
probably needed to get good results but certainly this might be a
good start.
To down sample you can use just some of the generated between
points and throw all the other points away so for every 4 points
you would get a point you interpolated that was in the middle of
them.
Apparently if you wanted something even better, say using
fractals you can get data that looks real when you upscale because
it keeps the same amount of detail and never ends up blured, kind
of hard to explain unless you read a few books on chaos, but
neadless to say that is probably going to be fairly slow
processing if you try to use fractals unless your very smart in
use?
Bicubic is probably one of the drop down options on photoshop
export for web when resizing an image.
I figured the code might be enough if ported and understood to
work in a recursive pyramid or a structure that process all groups
of 16. Often it's tricky to work out how to put maths into code
and this seemed to contain some of the essence although I admit
even the author suggests that it's a bit more work to apply to an
image. I suspect the python code does similar, but python is less
close to haxe so maybe no so easy to port.
but what's a tile pyramid? There's tile textures (which could be packed together in a batch), there's tile-maps (like what tiled might produce). Never heard of a tile pyramid before.
--