Reliability criteria in quantitative texture analysis with experimental and simulated orientation distributions

Abstract

Quantitative texture analysis reliability factors are examined from an experimental point of view, using real-sample and modelled orientation distribution refinements. The classicalRPfactors of texture analysis are shown to depend on the texture strength, and their representation of the various textures is not homogeneous. The surface-weightedRPwhomologues exhibit lower texture strength dependency and better homogeneities, but still do not allow for comparison of refinements operated on samples with different textures. New factors,Rw, weighted by the counting statistics, show the lowest dependency and best homogeneity. FromRw1(Rw0) curves a new criterion is established, which allows the detection of poorly refined orientation distributions. This study highlights a unique entropy-to-texture index relationship, which also gives a new criterion for testing refinement reliability, and proposes two different ranges in which to apply the texture index and the entropy factors.