“…Based on the BMF idea in [30], since the reference wafer and the m-th wafer share a similar spatial pattern, we expect that their DCT coefficients {α t,r,k ; k = 1, 2, ..., K} and {α t,m,k ; k = 1, 2, ..., K} are similar. Such a prior knowledge can be mathematically encoded by the following prior distribution:…”