A Simple Coupled Statistical Model for 3D Face Shape Recovery

Abstract

We focus on the problem of developing coupled statistical models that can be used to recover surface height from brightness images of faces. Our approach consists on using a simple model that assumes that the height eigenmodes are identical to the intensity eigenmodes. We recover the height function directly from the best-fit intensity parameters. As a result the computations involve only a straightforward matrix-vector multiplication. Experiments show that this method generate accurate height surfaces from out-of training intensity images.

Learning 3D shape from a single facial image via non-linear manifold embedding and alignment

“…It can be seen that our algorithm can obtain accurate reconstructions in spite of illumination and pose variations. The reconstructed error in each pose is shown in Figure 7, which shows that our algorithm is fairly insensitive to pose variations and achieves the same level of accuracy as the methods [11,4,16] in all poses. The recovery accuracy for the frontal facial images in our method is slightly better than that of those methods, but our method can handle illumination and pose changes.…”

“…(a) (b) (c) (d) For comparison we also show the best reconstruction errors of CCA [16], CSM [4] and Tensor+CCA [11]. These methods can only deal with the frontal images of face without illumination variation.…”

“…Lei et al [11] present an approach (Tensor+CCA) similar to [16], while the mapping is learned from the NIR tensor space to the 3D shapes. Castelan and Hancock [4] apply coupled statistical models (CSM) to recover surfaces from brightness images of faces. However, these statistical learning approaches can handle the shape recovery only from a frontal face image.…”

“…Another class of methods use machine learning techniques to reconstruct the 3D shape(e.g., [4,11,16]). These Learning-based methods take advantage of the availably of prior training data, i.e., face images with the corresponding shapes, from which the relationship of shapes and facial images can be inferred.…”

Recovering 3D facial shape via coupled 2D/3D space learning

“…Recently researchers begin to realize this problem. To make use of the coupling information, Casteln et al [13] proposed a coupled statistical model for facial shape recovery. The main idea of this coupled model is coefficients sharing.…”