Techniques to rate the validity of multiple methods to process multi-dimensional data

A method to operate a computing machine to rank the validity of a plurality (n) of data processing algorithms A1-An on input data I, wherein the input data is processed by each of the data processing algorithms to produce associated output data P1-Pn, the method including:

• generating one or both of phase representations and phase congruence representations of the input data, including:
• (1) computing an n Dimensional Fourier Transform or discrete cosine transform of the input data, and (2) computing the phase of the transform computed at (1);
• generating one or both of phase representations and phase congruence representations of the associated output data for each of the plurality of data processing algorithms, including:
• (3) computing an n Dimensional Fourier Transform or discrete cosine transform of the associated output data for each of the plurality of data processing algorithms, and (4) computing the phases of the transforms computed at (3);
• generating a similarity metric of one or both of the generated phase representations and phase congruence representations of the input data with one or both of the generated phase representations and phase congruence representations of the associated output data for each of the plurality of data processing algorithms, including (5) computing the normalized cross correlation (NCC), mean absolute phase difference (MAPD), normalized mean square error (NMSE) or scale invariant normalized mean square error (S1NMSE) of the input data transform phase computed at (2) with the output data transform phase computed at (4) for each of the data processing algorithms;
• and
• ordering the generated similarity metrics, wherein the order of the similarity metrics is a ranking of the validity of the data processing algorithms and is free from use of a ground truth.