1 May 2010 Contour estimation in images using virtual signals
Author Affiliations +
Optical Engineering, 49(5), 057002 (2010). doi:10.1117/1.3421576
The line-fitting problem has been transposed to the signal-processing framework: Array-processing methods can be applied to virtual signals generated from the image, to estimate straight-line orientations. This paper deals with the estimation of straight and distorted lines in images by fast array-processing methods. Hough transform and snake methods retrieve straight lines and distorted contours, but present limitations. We adapt a fast high-resolution method, the propagator method, to the estimation of multiple distorted contours. For the first time, a method is proposed to cope with the intrinsically limited size of images, which reduces the accuracy of the high-resolution methods due to the low number of signal realizations. Moreover, an extension to images impaired by correlated noise is proposed. For this, an extension of the subspace-based methods to a method based on higher-order statistics is proposed. Distorted contours are assimilated to distorted wavefronts and retrieved with a novel optimization method. The performance of the proposed method is validated on several images.
Salah Bourennane, Caroline Fossati, Julien Marot, "Contour estimation in images using virtual signals," Optical Engineering 49(5), 057002 (1 May 2010). https://doi.org/10.1117/1.3421576

Back to Top