Re: how to calculate gabor filter Convolution template

From: "Chen Sagiv" <chensagivron@xxxxxxxxx>
Date: Sat, 3 May 2008 09:52:03 +0000 (UTC)

Dear superwind wang,

An easy way could be to determine an array that is as large
as the image, where the filter takes only a small portion
of this array. Then, do the convolution in the frequency
domain.

Suppose that your image is img. Make its dimensions odd.

siz = size(img);
[x,y]=meshgrid(1:siz(1),1:siz(1));
x0 = (siz(1)+1)/2;
y0 = (siz(2)+1)/2;
filt = 1/(2*pi*σ^2)exp[-1/2(((x-x0)+(y-y0))/σ)^2]exp[j2*pi*f
((x-x0)*cosθ+(y-y0)sinθ)];

ConvRes = real(ifft2(fft2(img).*fft2(ifftshift(filt)));

I hope this works well for you.

Best,

Chen
"superwind wang" <wangbing7928@xxxxxxx> wrote in message
<fvhbm4$l6v$1@xxxxxxxxxxxxxxxxxx>...

Dear All:
I use this gabor filter
G(x,y)= 1/(2*pi*σ^2)exp[-1/2((x+y)/σ)^2]exp[j2*pi*f
(xcosθ+ysinθ)] for image edge detection.
now i know the value of σ,f and θ,how can i calculate
the gabor template, in other words,when i use the

function

conv2(A,B,'same') to get the filtered image,how can i
choose the values of variable x and y to generate the
matrix B.
Please help me ,i am in trouble.
Thank you!

References:
- how to calculate gabor filter Convolution template
  - From: superwind wang