Straight Line Extraction Using Iterative Total Least Squares Methods

In this paper we present a new algorithm for enhancing the accuracy of the parameter extraction of straight lines in a two-dimensional image. The algorithm achieves high accuracy in comparatively less computational time than most traditional methods and is invariant under rotation and translation. The Iterative Total Least Squares (ITLS) method starts from an initial estimate of the line parameters. When no a priori information about the image is available this estimate can be assigned randomly. Alternately, a lower accuracy method can be used to generate an initial estimate which will result in faster convergence. Then, a rectangular window is centered using the current line approximation, and a new line estimate is generated by making a total least squares fit through the pixels contained within the window. This is repeated until convergence is reached. Adaptively adjusting the window size yields the 4D ITLS process. In addition, a pairwise accelerated ITLS method has been developed which substantially increases the convergence rate. We conclude with some examples where the ITLS method has been used successfully