Segmentation of Quadrangles for Calibration Purposes

T. Hanning, P. Barth, and M. Kellner (Germany)


edge detection, sub-pixel accuracy, piecewise constant ap proximation, image acquisition, convolution


A common approach to image segmentation is the Mumford-Shah model, which formulates an approximation of a given image by piecewise continuous functions with certain constraints. For computational needs a simpliļ¬ca tion to piecewise constant functions is often made. In the case of extraction of features with sub-pixel accuracy - in particular for calibration purposes - we are not interested in a segmentation of the discretized image, but in a local ization of the edges of an object in reality. Therefore we must analyze the image acquisition process. In this article we model this process as a convolution of an input func tion followed by a discretization. We present an algorithm which detects the edges of simple primitives with sub-pixel precision independently from the convolution kernel.

Important Links:

Go Back