RE: Fwd: another filter to consider?

Hi Erik,

All I know about these filters (other than my fun experiments with them below) is found at http://msdn.microsoft.com/en-us/library/ms533091(v=vs.85).aspx . As I say, they seem to have been available there for many years and resemble transforms more than filters.  I wanted to do something like that (in my attempts to build image morphing software in the browser); poked around; played with them and was somewhat pleased, but of course, bumped into their limitations and wanted more.

To do it with deDisplacement one would have to have a gradient which is sinusoidal rather than linear or radial, which explains a part of my interest in those things as well. You could use script to stitch a series of piecewise linear gradients together into a quilt to simulate sinusoidal ones (I have some examples) and then do it, but even then feDisplacement seems to operate on pixels rather than on vectors, and the results are often very grainy. Having access to only 256 shades of underlying colors for the displacement also seems to be a bit of a limitation at times.

Maybe instead of handling it with filters, it would be easier to just define a collection of rather common transforms (like various lens effects, parabolic, spherical, conic and sinusoidal deformations, as well as the standard geographic projections and splines? Maybe the same math engine that underlies all these transforms could also do the corresponding classes of gradients? 

I might be missing something obvious or leaping through a singularity though.

Cheers
David

-----Original Message-----
From: Erik Dahlstrom [mailto:ed@opera.com] 
Sent: Monday, April 18, 2011 4:27 AM
To: public-fx@w3.org; ddailey; Doug Schepers
Subject: Re: Fwd: another filter to consider?

Hi David,

I'd be happy to hear some more details about that filter. Is it a  
simplified syntax for displacementmap transforms? Because it looked to me  
like it could be done with feDisplacementMap and gradients.


On Sat, 16 Apr 2011 19:08:06 +0200, Doug Schepers <schepers@w3.org> wrote:

> Forwarding to FXTF list, since we are now working on filters as a joint  
> SVG-CSS project.
>
> -------- Original Message --------
> Subject:  another filter to consider?
> Resent-Date:  Sat, 16 Apr 2011 15:57:01 +0000
> Resent-From:  www-svg@w3.org
> Date:  Sat, 16 Apr 2011 11:53:55 -0400
> From:  David Dailey <ddailey@zoominternet.net>
> To:  <www-svg@w3.org>
>
>
>
> Internet Explorer/ HTML/CSS has had, for at least 10 years, something
> present, I think, in versions 5 through 9, a wave filter. See
>
> http://srufaculty.sru.edu/david.dailey/javascript/ani/waves.html .
>
> If you don’t have IE, basically, it allows sinusoidal distortion of an
> image (until today I always used jpgs but today I discovered that it
> also works in IE9 with <img src=”file.svg”>, though the wave distortion
> is applied to a pixel-based rendition of the SVG file). It’s not quite
> as flexible, as I recall, as the bivariate sinusoidal wave function in
> Adobe Photoshop (that allows arbitrarily many generators to be added to
> the transform), but it is a step in the right direction!
>
> Anyhow, it is more what, in SVG, we would call a “transform” than a
> “filter”, but this sort of transform as well as projections (conic,
> spherical, and so forth) would be nice additions as well. I think the
> underlying math is probably simpler than the non-affine transforms being
> discussed for SVG 2.0.
>
> Applications are, as always, artistic, scientific and geographic.
>
> Regards
>
> David
>


-- 
Erik Dahlstrom, Core Technology Developer, Opera Software
Co-Chair, W3C SVG Working Group
Personal blog: http://my.opera.com/macdev_ed

Received on Thursday, 21 April 2011 00:59:53 UTC