PDF version of this document

Effects of the Shuffle Transform

The experiment described in the previous section was repeated for shuffle neighbourhoods of 1x1 (Euclidean distance), 3x3, 5x5, and 7x7, to test the hypothesis that this would extend the range over which different degrees of mis-registration could be discriminated.

  

Fig. 6. Specificity and Generalisation for increasing mis-registration of different shuffle neighbourhood sizes.