Active blobs: region-based, deformable appearance models

Abstract

A region-based approach to nonrigid motion tracking is described. Shape is defined in terms of a deformable triangular mesh that captures object shape plus a color texture map that captures object appearance. Photometric variations are also modeled. Nonrigid shape registration and motion tracking are achieved by posing the problem as an energy-based, robust minimization procedure. The approach provides robustness to occlusions, wrinkles, shadows, and specular highlights. The formulation is tailored to take advantage of texture mapping hardware available in many workstations, PCs, and game consoles. This enables nonrigid tracking at speeds approaching video rate. Ó 2003 Published by Elsevier Science (USA).

Fast Non-Rigid Surface Detection, Registration and Realistic Augmentation

“…This has been successfully demonstrated in the context of nonrigid tracking (Baker et al, 2004;Cootes et al, 2001;DeCarlo and Metaxas, 1998;Sclaroff and Isidoro, 2003) but typically requires a good initialization because the criteria being minimized tend to have many local minima.…”

“…The two are of course complementary since trackers require initialization and, no matter how good they may be, will sometimes lose track, for example, because of severe occlusions. Non-rigid object tracking has also been convincingly demonstrated, for example in the case of animated faces (Baker et al, 2004;Cootes et al, 2001;DeCarlo and Metaxas, 1998) or even more generic and deformable objects (Bartoli and Zisserman, 2004;Sclaroff and Isidoro, 2003). However, the automated detection of such deformable objects still lags behind and existing methods (Belongie et al, 2002;Ferrari et al, 2004) are far less convincing for real-time applications.…”

Retexturing in the Presence of Complex Illumination and Occlusions

“…Image warping approaches have long incorporated polynomial [17] or spline [16] illumination models. Similarly, [5] allows for an affine illumination change.…”

Shape Particle Filtering for Image Segmentation

“…Statistical models of global object appearance are widely used for image segmentation [7,13,12], and form powerful tools especially in the case of missing or locally ambiguous boundary evidence.…”