Steger's Line Detection Routines

A gravitational wave signal in interferometer data $h(t)$ should produce a ridge in the TFD $\rho(t,f)$. Therefore, to detect GWs, a ridge detection algorithm (or equivalently line detection algorithm if $\rho(t,f)$ is represented as a gray-scale map as in Fig. ) is required. Fortunately, there are a number of ridge detection algorithms in the digital image processing and computer vision literature from which to choose.

We use Steger's second-derivative hysteresis-threshold algorithm [35]. The essential idea of this scheme is simple. A ridge in a surface will have high curvature (second derivative of $\rho(t,f)$) in the direction perpendicular to the ridge. Furthermore, the first derivative will vanish at the top of the ridge, since it is a local maximum. Thus, ridges are identified as contiguous sets of point at which $\rho(t,f)$ has a high-curvature local maximum. We have included Stegers code [34] in the GRASP package with minor changes. The main aim of making the the changes was to enable Steger's routines to take a map whose pixels are of type float rather than unsigned char. For details about this algorithm please see [33,35]