Abstract

The construction of Facial Deformable Models (FDMs)is a very challenging computer vision problem, since theface is a highly deformable object and its appearance dras-tically changes under different poses, expressions, and il-luminations. Although several methods for generic FDMsconstruction, have been proposed for facial landmark local-ization in still images, they are insufficient for tasks such afacial behaviour analysis and facial motion capture whereperfect landmark localization is required. In this case,person-specific FDMs (PSMs) are mainly employed, requir-ing manual facial landmark annotation for each person andperson-specific training. In this paper, a novel method for the automatic construc-tion of PSMs is proposed. To this end, an orthonormal sub-space which is suitable for facial image reconstruction islearnt. Next, to correct the fittings of a generic model, image congealing (i.e., batch image aliment) is performed by employing only the learnt orthonormal subspace. Finally, the corrected fittings are used to construct the PSM. The image congealing problem is solved by formulating a suitable sparsity regularized rank minimization problem. The proposed method outperforms the state-of-the art methods that is compared to, in terms of both landmark localizationaccuracy and computational time.