Optimal filters for the detection of linear patterns in 2-D and higher dimensional images

摘要

It is shown that the rotation-invariant operators form a set of orthogonal basis functions which is optimal for a wide range of 2-D and 3-D feature-detection operations. The optimality criterion used is based on the Karhunen-Loève expansion of the local image data. This is a generalization of the work of Zucker and Hummel who showed that these filters are optimal filters for the detection of step edges in 2-D and 3-D images.