wolffd@0: N = 3;
wolffd@0: x = rand(N,2); % each row is a feature vector 
wolffd@0: m = mean(x,1);
wolffd@0: xc = x-repmat(m, N, 1);
wolffd@0: 
wolffd@0: C = eye(N) - (1/N)*ones(N,N);
wolffd@0: xc2 = C*x;
wolffd@0: assert(approxeq(xc, xc2))