Genetic programming for automatic construction of variant features in edge detection

Abstract

Basic features for edge detection, such as derivatives, can be further manipulated to improve detection performance. However, how to effectively combine different basic features remains an open issue and needs to be investigated. In this study, Genetic Programming (GP) is used to automatically and effectively construct rotation variant features based on basic features from derivatives, F-test, and histograms of images. To reduce computational cost in the training stage, the basic features only use the horizontal responses to construct new horizontal features. These new features are then combined with their own rotated versions in the vertical direction in the testing stage. The experimental results show that the rotation variant features constructed by GP combine advantages from the basic features, reduce drawbacks from basic features alone, and improve the detection performance. © Springer-Verlag Berlin Heidelberg 2013.