Image interpolation scheme of windows deformation (forward shift vs symmetric shift)

[wywisfah sgzksf]

Hello, everyone. I am learning piv from source github code. 

I notice that the windows deformation algorithm in multi-pass process uses forward shift interrogation (see code), which apply local coarse velocity to the second image:

X2 = interp2(X,Y,U,X1,Y1,'*linear') + repmat(X1,size(Y1, 1),1);

Y2 = interp2(X,Y,V,X1,Y1,'*linear') + repmat(Y1,1,size(X1, 2));

The following commented code about symmetric/centre interpolation of image A and B 

%symmetric interpolation of image A and B
%X2 = interp2(X,Y,U*0.5,X1,Y1,'*linear') + repmat(X1,size(Y1, 1),1);
%Y2 = interp2(X,Y,V*0.5,X1,Y1,'*linear') + repmat(Y1,1,size(X1, 2));
%X2_2 = interp2(X,Y,U*-0.5,X1,Y1,'*linear') + repmat(X1,size(Y1, 1),1);
%Y2_2 = interp2(X,Y,V*-0.5,X1,Y1,'*linear') + repmat(Y1,1,size(X1, 2));

However, the paper claimed that symmertric/centre shift interrogation scheme "has the advantage of being second-order time accurate, and therefore yields a considerable improvement when a large time delay occurs between exposures. "  (F Scarano 2002 Meas. Sci. Technol. 13 R1, page 6)

I want to know why does pivlab use the forward shift interrogation instead of the high-order symmetric shift interrogation scheme?

[William]

Hi,

in my tests (a pretty long time ago) there was not a large difference in the results (bias / RMS error), but the processing time was considerably slower. Please feel free to validate this!