Cager: From 3d performance capture to cage-based representation

Figure 1: CAGER Processing Pipeline. From left to right: given a raw 3D+time sequence and an initial cage, we first extract an optimal subset of positional constraints for the cage coordinate inversion. Then, the cage coordinates are inverted for each frame of the input sequence. A selective enforcement of regularization terms is defined to affect the cage vertices where the inversion is the most unstable. The resulting smoothly varying cage sequence faithfully reconstructs the input sequence when applied to a single mesh frame