A Linear Discriminator of Width

José A. F. Leite, Edwin R. Hancock

Abstract

Our aim in this paper is to present a methodology for linearly combining multi-channel filter information, using principal components analysis. We take as our application vehicle the detection and classification according to width of line-features. Here we operate on the raw data with oriented multi-scale line detection kernels. The objective in this study is to develop a linear discriminator of line width which can be used to classify the detected features. By performing principal components analysis on the response of the multi-scale filter bank to a set of example line-profiles, we transform the filter-outputs. Under this transformation, the principal component of maximum variance is demonstrated to have linear dependance upon line-width. We demonstrate the utility of our line-classification process under conditions of controlled noise.
BibTeX
@inproceedings{leite1995linear,
  author = {José A. F. Leite and Edwin R. Hancock},
  title = {{A Linear Discriminator of Width}},
  booktitle = {ICIAP},
  year = {1995},
  doi = {10.1007/3-540-60298-4\_301},
}

Loading related content...

Figure