MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: 2D Convolution

  • To: mathgroup at
  • Subject: [mg76164] Re: 2D Convolution
  • From: Mathieu G <ellocomateo at>
  • Date: Thu, 17 May 2007 05:55:36 -0400 (EDT)
  • References: <f29389$m7p$> <f2bunr$jig$>

dh a =E9crit :
> Hi Mathieu,
> Note that the Gaussian and Hole covolution is symmetrical and can
> therefore done as a much simplier 1 dim problem. However, if you need
> 2dim, I would advice, due to computing time, to approximate the 2D
> convolution by a numerical approximation. Here is a small example:
> Hole[x_,y_]:=If[0.2<x^2+y^2,0,1];
> Gaussian[x_,y_]:=Exp[-10(x^2+y^2)];
> fun[x_/;NumericQ[x],y_]:=
>    NIntegrate[Gaussian[x-xx,y-yy] Hole[xx,yy],{xx,-2.,2.},{yy,-2.,2.}] =
> fun1=FunctionInterpolation[ fun[x,y]  ,{x,-1.,1.},{y,-1.,1.}]
> hope this helps, Daniel

Thank you for your reply!
Your comments are useful in my discoering of Mathematica.

I am not too sure why there is no NumericQ checking on y in the <fun>
function you propose? Can you tell me why please?

  • Prev by Date: Mathematica 6.0 and MatrixManipulations
  • Next by Date: how save newly created palette?
  • Previous by thread: Re: 2D Convolution
  • Next by thread: Re: 2D Convolution