find shift using cross correlation

Dear all;

i have a confusion when i am using xcorr function in matlab , suppose i have attached two signals , is this function give the amount of shift in between signals ? is there any conversion factor that shall i use ?

Answer 1

[Image Analyst]

I suggest you start with a simple example, like

signal1 = zeros(1, 10);
signal1(5) = 1;
signal2 = zeros(1, 10);
signal2(8) = 1;

Then look at the length of the output and where the max/peak/spike is located. You'll figure it out.

By the way, I don't think it's well known that the max of the correlation is not always when the two signals are optimally overlapped - it just tells you when the area under the curve of the multiplication of the two signals is greatest. I believe this is why normalized cross correlation was invented. This is done by the function normxcorr2 in the Image Processing Toolbox. (Demo available upon request).

[mustafa alnasser]

Dear Image analyst;

Thank you very much for help . in fact , i understand th concept but i was worried if i am understand in built function well or not.

the above two lines are boundary between two areas in am image ( after extraction). these image are taken from a video and i am trying to estimate how much the two objects are moving with time (in other words , estimate distance shift (in pixels)) then later on convert this to an actual distance in meter. the signals are normalized as shown below: s1= EE1(:,1); s2= EE2(:,1);

s1_sum=0;
s2_sum=0;
for i=1:L1(1,1)
    m1= s1(i)*s1(i);
    s1_sum= s1_sum+m1;
    m2= s2(i)*s2(i);
    s2_sum= s2_sum+m2;
end 
%// Normalize signals 

norm= sqrt(s1_sum*s2_sum);
s1 = (s1 - mean(s1)) /norm;
s2 = (s2 - mean(s2)) /norm;

But the issue that I am not getting satisfactory results when i compare it with actual ones

[Image Analyst]

Attach s1 and s2 in a mat file so we can run your code.

[mustafa alnasser]

Dear Image analyst; Thank you very much for your help.

Attached are mat files for original signals . the code shown below

s1= EE1(:,1);
s2= EE2(:,1);
L1=size(s1);
s1_sum=0;
s2_sum=0;
for i=1:L1(1,1)
    m1= s1(i)*s1(i);
    s1_sum= s1_sum+m1;
    m2= s2(i)*s2(i);
    s2_sum= s2_sum+m2;
end 
%// Normalize signals 

norm= sqrt(s1_sum*s2_sum);
s1 = (s1 - mean(s1)) /norm;
s2 = (s2 - mean(s2)) /norm;

%// Compute time lag between signals [c,lag] = xcorr(s2(:,1), s1(:,1), 'none' ); %// Cross correlation

% % lag = mod(find(c == max(c)), length(s2)) %// Find the position of the peak % [x,I] = max((c)); Distshift = lag(I)

   figure (3)
    subplot(311); plot(s1(:,1)); title('s1'); 
    subplot(312); plot(s2(:,1)); title('s2'); 
    subplot(313); plot(lag,c); 
    title('Cross-correlation between s1 and s2')

FOV=0.10; % Field of View (Meter) FW=1632; % Frame Width (Pixel) Actual_Distance= Distshift*FOV/FW; % Calcuate the actual distance in meter

[mustafa alnasser]

any feedback about the code?

[Angel Romero]

Hi there! I tried to solve your problem but I wasn´t able. Any solution so far?

[Angel Romero]

Were you able to fix it?. I'm looking forward to hearing from you. Thanks.