Prev: cognitive radio
Next: when exact differential equation solvers go bad!
From: center15 on 7 Dec 2009 03:09
Hi,
I got a project which extracting vessels from retinal image. The filter must be in frequency domain. I am having a problem with my filter algorithm. On my research, usually gaussian filter method modified for matched filter. Then use imrotate 90 degree and use sigma 1 and 3. After all sigma1-sigma3 will give final result, but I cannot get the final result as correct. I thing my filter algorithm is wrong or I am missing some method after filter. Could you please help. Please dont send any link about matched filter, I allready checked. Thank you

function [sigm1] = test(im)
imf = fftshift(fft2(im));
[co,ro]=size(im);

H1 = Gfilter(co,ro, 1); %sigma 1
out1= imf.*H1;
sigm1=abs(ifft2(out1));

H3 = Gfilter(co,ro, 3); %sigma 3
out3= imf.*H3;
sigm3=abs(ifft2(out3));

imshow(sigm1-sigm3,[]);

function H = Gfilter(nx,ny, s)
x_orig =floor(nx/2) + 1;
y_orig = floor(ny/2) + 1;

H=zeros(nx,ny);
for x=1:nx
for y=1:ny
u = sqrt((x-x_orig)^2 + (y-y_orig)^2);
H(x,y)=exp(-u/2/s^2);
end
end
H=imrotate(H,90);
H=imresize(H,[nx,ny]);
 | 
Pages: 1
Prev: cognitive radio
Next: when exact differential equation solvers go bad!