Normalized cross correlation of matrices along specified dimensions

I need a function that computes the normalized cross correlation along every column/row of a matrix with the corresponding column/row of another matrix

I could do:

for i = 1 : size( A , 2 )

x(:,i) = normxcorr2( A(:,i) , B(:,i) )

end

but this runs very slowly for large matrices. I found the function fastConv.m on the file exchange which computes the convolution of two matrices along a particular dimension which could be used to find the cross correlation though not normalized.

http://www.mathworks.com/matlabcentral/fileexchange/authors/115845

I need to somehow combine fastConv with normxcorr2, but I cannot figure out how to adapt the code in normxcorr2 to get accurate normalization.

any help would be much appreciated.

thank you Sam