Shape Theories. II. Compactness Selection Principles

Abstract

Shape(-and-scale) spaces - configuration spaces for generalized Kendall-type Shape(-and-Scale) Theories - are usually not manifolds but stratified manifolds. While in Kendall's own case - similarity shapes - the shape spaces are analytically nice - Hausdorff - for the Image Analysis and Computer Vision cases - affine and projective shapes - they are not: merely Kolmogorov. We now furthermore characterize these results in terms of whether one is staying within, or straying outside of, some compactness conditions which provide protection for nice analytic behaviour. We furthermore list which of the recent wealth of proposed shape theories lie within these topological-level selection principles for technical tractability. Most cases are not protected, by which the merely-Kolmogorov behaviour may be endemic and the range of technically tractable Shape(-and-Scale) Theories very limited. This is the second of two great bounds on Shape(-and-Scale) Theories, each of which moreover have major implications for the Comparative Theory of Background Independence as per Article III.