Question: Accuracy of orientation estimation algorithms
Vision List Digest:
Article 4,
Volume 14, Issue 31
From: Naoki Saito
Post-Followup: submission@VISLIST.com
Hello! I would like to know whether there is a paper or report on the
accuracy of the orientation information (of edges or lines). I know there
are many papers on the orientation estimation algorithms (e.g., Kass & Witkin,
Rao & Schunk, or steerable filters of Freeman and Adelson, Perona etc.),
but the thing I really want to know is:
(1) Is there any quantitative comparison on the estimated angles between the
method using two orthogonal derivative operator (e.g., first derivative of
Gaussian along x, y directions) and the methods using more convolution kernels
(such as 8 orientations of the elongated Gaussians used by Zucker et al.)?
(2) How about the effect of the quantization of edges or lines on digital
images? Any (analytic) straight line cannot be precisely represented on the
digital images: it becomes a staircase. So, viewing locally, it is not
really the straight line. Is there any study on how to estimate accurately
the orientation of such lines with minimal quantization effect?
I would greatly appreciate any help/advice/pointer!
Best regards and thanks in advance,
Naoki Saito Schlumberger-Doll Research
Email: saito@ridgefield.sdr.slb.com
Old Quarry Road
Voice: (203) 431-5209 Fax: (203) 438-3819 Ridgefield, CT 06877 USA
http://www.vislist.com