On the Reliability of Computing Wigner Texture Features

In this work we analyse the robustness of the computation of pseudo-Wigner texture features using both analytic and statistical methods. It is shown that if the input error is normally distributed and the SNR of the input is not very low (i.e. if SNR ≥ 3), then the output error is also normally distributed with known mean and variance. The error distribution in some typical simple signals is considered. The effects of quantisation are also investigated.

[1]  Henrik I. Christensen,et al.  EditorialPerformance characteristics of vision algorithms , 1997, Machine Vision and Applications.

[2]  Josef Kittler,et al.  Error-sensitivity assessment of vision algorithms , 1998 .

[3]  C. K. Yuen,et al.  Digital Filters , 1979, IEEE Transactions on Systems, Man, and Cybernetics.

[4]  Josef Kittler,et al.  Texture crack detection , 2005, Machine Vision and Applications.

[5]  Neil A. Thacker,et al.  Algorithmic modelling for performance evaluation , 1997, Machine Vision and Applications.

[6]  I. Miller Probability, Random Variables, and Stochastic Processes , 1966 .