pfftconvolution

Performs convolution of image by kernel.

Synopsis

pfftconvolution [im_in1|-] [im_in2|-] [im_out|-]

Description

pfftconvolution performs a convolution the input image im_in1 with the kernel given in the input image im_in2. The size of the kernel image im_in2 must be lower or equal than the input image im_in1.

The convolution uses the frequency domain. (It differs from the pconvolution operator that uses the spatial domain.) It consists in a complex multiplication of the Fourier transform of the two input images:

imi=fft(im_in) * fft(im_in2)
im_out=iift(imi);

where * is the complex multiplication.

Inputs

im_in1: a 2D image.
im_in2: a 2D image (same type as im_in1 and a size lower or equal to im_in1).

Outputs

im_out: a float image of the same size as the input image im_in1.

Result

Returns SUCCESS or FAILURE.

Examples

Adds a motion blur on the tangram.pan The motion blur is generated by the way of an oblique line:

   pshapedesign 10 10 0 3 10 1 line.pan
   protation 0 45 line.pan line1.pan
   pfftconvolution tangram.pan line1.pan out.pan

Prototype C++

Errc FFTConvolution( const mg2duc &im_in1, const Img2duc &im_in2, Img2dsf &im_out );

Version française

Convolution d'une image par un noyau.

Author: Régis Clouard