3D face recovery from intensities of general and unknown lighting using Partial Least Squares

3D-reconstruction of human jaw from a single image : integration between statistical shape from shading and shape from shading.

“…In this case, the shape is solved for using a series of matrix operations guaranteeing faster shape recovery when compared to its iterative counterpart. This was proven to yield comparable results in terms of reconstruction accuracy [39]. Moreover as we saw recently that we modeled the HP image model by using PCA as , in the same concept we model a sparse 2D model Landmarks for annotation landmarks by using PCA as 2 = + .We use the partial least squares regression(PLS)…”

“…There has been a substantial amount of work regarding statistical shape recovery for human face modeling and biomedical structures with distinct shapes -e.g., modeling the corpus callosum, the kidney and spinal cord; it is an active research area under shape and appearance modeling (e.g., [39,40]). Atick et al [5] proposed the first SSFS method where principal component analysis (PCA) was used to parameterize the set of all possible facial surfaces.…”

“…This has also been validated for near illumination conditions. Since then, SH was incorporated in SFS framework to tackle the problem of illumination [4,12,38,39]. Zhang and Samaras [43] primary goal was to recover the SH basis images from a single test image.…”

“…Castelan and Horebeek [12] and Rara [39] cast their models in a regression framework using the Partial Least Squares (PLS) method which uses a few matrix operations for shape reconstruction to provide a computationally efficient alternative to the iterative methods used in [13][38]. and summed-up to construct the harmonic projection (HP) images which encodes the illumination conditions of the input image.…”

Face recognition in the wild.

“…As an example Figure 52 shows the first nine Spherical Harmonics (SPH) constructed on USF database ( [48][57]).…”