diff --git a/code_forMetrics/antonioGaussian.m b/code_forMetrics/antonioGaussian.m
index b700b32..37f7688 100755
--- a/code_forMetrics/antonioGaussian.m
+++ b/code_forMetrics/antonioGaussian.m
@@ -1,4 +1,4 @@
-function [BF, gf]=gaussian(img, fc)
+function [BF, gf]=antonioGaussian(img, fc)
 % Gaussian low pass filter (with circular boundary conditions).
 % 
 %  [BF, gf]=gaussian(img, fc);
@@ -24,32 +24,42 @@
 % due to zero padding of the input.
 
 % Antonio Torralba, 1999
+% edited: David Berga, Jan 2017
 
 [sn, sm, c]=size(img);
 n=max([sn sm]);
 n=n+mod(n,2);
 n = 2^ceil(log2(n));
 
+img2 = zeropadimage(img,n); %image with zero padding
+[sn2, sm2, c2]=size(img2);
+n2=max([sn2 sm2]);
+%n2=n2+mod(n2,2);
+%n2 = 2^ceil(log2(n2));
+
 % frequencies:
 [fx,fy]=meshgrid(0:n-1,0:n-1);
 fx=fx-n/2; fy=fy-n/2;
 
 % convert cut of frequency into gaussian width:
-s=fc/sqrt(log(2));
+s=round(n2/n) * fc/sqrt(log(2)); %re-scaled to padded image dimensions
 
 % compute transfer function of gaussian filter:
 gf=exp(-(fx.^2+fy.^2)/(s^2));
-gf = fftshift(gf);
+
+gf2 = zeros(n2,n2,c); gf2(n2/2-n/2+1:n2/2+n/2,n2/2-n/2+1:n2/2+n/2,:) = gf;
+gf2 = fftshift(gf2);
+
 
 % convolve (in Fourier domain) each color band:
-BF = zeros(n,n,c);
+BF = zeros(n2,n2,c);
 for i = 1:c
-    BF(:,:,i)=real(ifft2(fft2(img(:,:,i),n,n).*gf));
-    %BF(:,:,i)=real(ifft2(fftshift(fftshift(fft2(img(:,:,i),n,n)).*gf)));
+    BF(:,:,i)=real(ifft2(fft2(img2(:,:,i),n2,n2).*gf2));
+    %BF(:,:,i)=real(ifft2(fftshift(fftshift(fft2(img2(:,:,i),n2,n2)).*gf2)));
 end
 
 % crop output to have same size than the input
-BF=BF(1:sn,1:sm,:);
+BF=BF(n+1:n+sn,n+1:n+sm,:);
 
 % if no input parameters are provided, then it shows a section of the
 % gaussian filter:
@@ -63,5 +73,19 @@
    xlabel('cycles per image')
    ylabel('amplitude transfer function')
 end
+end
+
 
+function [Ipad] = zeropadimage(I,p)
+
+%Find size of image
+[h, w] = size(I); 
+
+%Pad edges
+Ipad = zeros(2*p, 2*p);  
+
+%Middle
+Ipad(p+1:p+h, p+1:p+w) = I;
+
+end