Next: fsharpen Up: Reference Previous: forder   Contents   Index

#### fsepconvol

Name

fsepconvol Convolution with a separable 2D-kernel

Command Synopsis

fsepconvol [-c width] [-g std] [-b b] in out [xker yker ]

-c width : unit mass constant kernel of specified width

-g std : Gaussian kernel, standart deviation std (precision 1e-3)

-b b : boundary condition: 0=zero, 1=symmetry (default), 2=torus

in : input Fimage

out : output (convolved Fimage)

xker : use a Fsignal (and its shift) as kernel along x axis

yker : use a Fsignal (and its shift) as kernel along y axis

Function Summary

Fimage fsepconvol (in , out , xker , yker , width , std , b )

Fimage in , out ;

Fsignal xker , yker ;

int *width , *b ;

float *std ;

Description

This modules allows to quickly apply a 2D-convolution to a Fimage in the case of a separable kernel,

K(x, y) = xker(x) . yker(y).

The kernel K can be given with xker and yker (Fsignal), or chosen between a constant kernel or a Gaussian kernel, both with total mass 1. The boundary assumption is, depending on the value of the b parameter, that
• b = 0: the image is zero outside its domain
• b = 1: the image is symmetrized on each side (eg if u > 0, u(- i) = u(i - 1))
• b = 2: the image is periodic (eg u(i) = u(i + n))
See module fconvol for general convolution with non-separable 2D kernels.

NOTE: calling this module with in=out is possible

Version 1.2

Last Modification date : Thu Nov 29 20:23:56 2001

Author

Lionel Moisan

Next: fsharpen Up: Reference Previous: forder   Contents   Index
mw 2004-05-05