Unified filter modulation synthetic discriminant function

To improve correlation performance, we give further consideration to the filter- encoding constraint applied to the SDF construction. The idea is to synthesize the SDF from the linear combination of a set of training images that are already filter modulated, i.e. preprocessed, so that the constructed SDF is dominated by the higher, not the lower, frequency components of the individual training-set image. The filter-encoding constraint is then applied to the SDF. For convenience this is called the modified fSDF (MfSDF).