Image Features Based on Local Hough Transforms

Authors:
Andrzej Śluzek
Affiliations:
School of Computer Engineering, Nanyang Technological University, Singapore 639798 and Faculty of Physics, Astronomy and Informatics, Nicolaus Copernicus University, Toruń 87-100
Venue:
KES '09 Proceedings of the 13th International Conference on Knowledge-Based and Intelligent Information and Engineering Systems: Part II
Year:
2009

Citing 4
Cited 0

The Vector-Gradient Hough Transform

IEEE Transactions on Pattern Analysis and Machine Intelligence
Use of the Hough transformation to detect lines and curves in pictures

Communications of the ACM
LIGHT: Local Invariant Generalized Hough Transform

ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 03
Finding Picture Edges Through Collinearity of Feature Points

IEEE Transactions on Computers

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new method of building local image features is proposed. The features are represented by various shapes (patterns) that can be approximated using Hough transforms. However, the transforms are applied locally (to the current content of a scanning window) so that the shape's location is fixed at the current window's position. Thus, the parameter-space dimensionality can be reduced by two (compared to globally computed Hough transforms) and the transforms can be effectively applied to more complex shapes. More importantly, shapes can be decomposed (two decomposition schemes are proposed) so that the overall complexity of the shapes used as features can be very high. The proposed feature-building scheme is scale-invariant (if scale is a dimension of the parameter space) subject only to diameters of scanning windows.