||Steerable filters are evaluated as tools for determining fibre orientation
in paper. Filters due to Freeman and Adelson and to Jacob and Unser are
presented and compared. Both methods are compared with a simple gradient filter.
Several model fibre orientation probability distributions are compared. The effects of e.g.
filter order and filter kernel width are studied. The filters' noise properties
and their ability to reproduce known fibre orientations are also investigated.
Steerable filters, like the gradient filter, seem to reproduce the fibre orientation angle well. They are less successful in estimating the fibre orientation anisotropy, although the gradient filter yields even worse estimates.
The Freeman and Adelson filter tends to over-estimate the anisotropy, while it is under-estimated by the Jacob and Unser filter (and the gradient filter). Anisotropy estimates appear to grow (and sometimes, but not always, saturate) with filter order and filter kernel width. Steerable filters are observed to have better noise properties than the gradient filter. Steerable filers, however, are also much slower than the gradient filter.
Some modifications to Jacob and Unser's original work is also suggested.