Well I guess you want to get single scalar value as distance of A2 for each Chanel in A1. If so then I try to do in steps. There are multiple ways but I would do like this.
First i will make A2 as single column vector
n_A2 = A2(:);
then I will make one matrix by A1, where each column is one image.
n_A1 = zeros(length(n_A2), size(A1,3)); % this creates matrix of rows as length of and columns are equal to number of images in A1
for i=1:size(A1,3)
t= A1(:,:,i);
n_A1(:,i) = t(:);
end
and Finally now compute the Euclidean distance n_A2 with all images in n_A1
A= repmat(n_A2,1,size(n_A1,2));
E_distance = sqrt(sum((A-n_A1).^2));
Now E_distance contains the output distance. There must be one value 0 as you are taking query image from the same pool of images.
I wish it works for you. I executed this on small example. Once you get A1 and A2 from Ds. then copy paste the following code below. It is the same code written above.
n_A2 = A2(:);
n_A1 = zeros(length(n_A2), size(A1,3));
for i=1:size(A1,3)
t= A1(:,:,i);
n_A1(:,i) = t(:);
end
A= repmat(n_A2,1,size(n_A1,2));
E_distance = sqrt(sum((A-n_A1).^2));
