Traditional Lines and Circles detector

Abstract

The traditional method for extracting features from image is accomplished by Hough Transform. When an image is represented mathematically, it becomes easy to perform computation according to a certain algorithm and detect the required features from the image. The presence of such features is determined by a voting procedure in a parameter space, from which the selected features are obtained as local maxima by creating a certain threshold. This procedure for simple detection of lines was patented by Paul V.C. Hough in 1962. There has been various advancement in this field for detecting lines and circles, all of them are based on the parent Hough Transform way to vote for the corresponding features. The various ways of detecting features and the mathematics behind them is discussed in this report.

Detecting Features via Modified Harris Corners and Matching Them through SIFT

“…The textures will vary from image to image and hence we can't take it as good features. Circular arcs cannot be used to detect corners [2]. If the edges are connected from image to image, we can get a wireframe 3D model of the region as 3D surfaces.…”