bd2Fdesc.m

% Convert (binary) mask image to Fourier Descriptors (FDs) code of its contour

% Required function: find_close_indeces.m

% inputs 
% img: (binary) mask image
% ncoef: the number of FDs required (<= the total number of contour pixels)

% output 
% FD code (FD complex numbers and the total number of contour pixels)
%

function o = bd2Fdesc(img,ncoef)

img(img~=0)=1;
img = bwperim(img);
image_edged = double(img);

% border_fft=zeros(size(image_edged));

% find border
f=find(image_edged);
lenf=length(f);
[new_indeces]=find_close_indeces(image_edged);
ii=new_indeces(:,1);jj=new_indeces(:,2);
border_cmplx=ii+1i*jj;

border_fft =fftshift(fft(border_cmplx));

% deal with even and odd nums
if mod(lenf,2)
    k=(lenf-1)/2;
else
    k=lenf/2;
end
rc = k+1;

% if mod(lenf,2) % odd
%     lenf = lenf-1;
% end
% rc = fix(lenf/2)+1;

p1=(rc+1):(rc+1+ncoef-1);
p2=(rc-1):-1:(rc-1-ncoef+1);

o = zeros(2*ncoef+1,2);

n = 1;

% try

for ind =(ncoef):-1:1
    o(n,1)=real(border_fft(p2(ind)));
    o(n,2)=imag(border_fft(p2(ind)));
    n=n+1;
end

%add dc
o(n,1)=real(border_fft(rc));
o(n,2)=imag(border_fft(rc));
n=n+1;

for ind = 1:(ncoef)
    o(n,1)=real(border_fft(p1(ind)));
    o(n,2)=imag(border_fft(p1(ind)));
    n=n+1;
end

% catch 
%     disp (lenf)
%     disp (ind)
% end

% save the total number of contour pixels
o(n,1)=length(f);
o(n,2)=length(f);

end