Fast Line Detection by the Local Polar Coordinates Using a Window

摘要

The Hough transform is known as a technique for detecting image components (primitives) such as lines or circles. We previously proposed a "primitive detection method using local polar coordinates," which can quickly detect primitives by reducing the dimensionality of the vote array of this Hough transform. However, this method has the property that the primitive detection speed is affected by the position and quantity of noise that is contained in the image (this property is called "noise dependency"), and this property had become a problem when the method was actually used. Therefore, in this paper, we propose a solution for this noise dependency problem. The proposed method sets a window in the vicinity of a local pole origin and determines whether or not a local pole can be set according to an operation (multiple votes per pixel) within this window. From various types of experimental results, we show that the proposed method is highly effective even though it is simple.