Stochastic-Wavelet Preprocessing of Heterogeneous Data for Finite Element Analysis

This paper extends the texture descriptors used by the stochastic-wavelet supervised classification technique and applied to two sets of test suites. Overall, these additional parameters improved the classification accuracy. A comparative analysis, between the dimensionality of texture feature-space against four classification schemes, is carried out to study the effect of classification accuracy. The study shows that a combination of different types of texture features is better w.r.t. classification accuracy but is computationally expensive.