3. Spatio-temporal trajectories
In this Deferasirox section, we provide the technical details behind our spatio-temporal DGTC classification method. A simplified workflow of the major elements of our system is shown in Fig. 1. We implemented our algorithms in Python and make use of several libraries including Scipy/Numpy and OpenCV (ver2.4), a well-known open source library for computer vision. We also developed a graphical interface in PyQT (QT4 library extensions for python) and produce real-time 3-dimensional plots of the spatio-temporal with MayaVi.
3.1. The MVFI spatio-temporal template
In Olivieri et al. (2012), we described the MVFI (Motion Vector Flow Instance) spatio-temporal template that encodes the velocity field of different human movements. These templates, corresponding to each image frame, are constructed by mapping the optical flow field, ff, of the foreground motion onto an evenly spaced grid, given by (xn,ym)(xn,ym). At each grid point (xi,yj)(xi,yj), the magnitude and direction of the optical flow vector vi,jvi,j are encoded; the direction by a bounding box, while the magnitude by pixel color (Fig. 2). For an input video consisting of NfNf frames, hypertonic procedure will produce a corresponding video sequence of template frames having the Nf-1Nf-1 frames. Fig. 2 illustrates these concepts and shows the MVFI encoding scheme for the dense optical flow of a video sequence.