PDF version of this entire document
In Chapter 6, I devised a measure of image distance for computing the overlap between distributions. With many images to be considered, I could also attempt systematic alterations.
Of particular interest is the notion of sensitivity as it enables one assessment method to be compared against another. Although sensitivities were shown to culminate at a particular value of the shuffle distance, other approaches that I had investigated could perhaps entail superior sensitivity, at the expense of computational power. It is worth exploring and finding if a more complex approach, e.g. one which considers an average or median in a neighbourhood of pixels/voxels, outperforms shuffle distance.
