Illumination and Person-Insensitive Head Pose Estimation Using Distance Metric Learning

Abstract

Abstract. Head pose estimation is an important task for many face analysis applications, such as face recognition systems and human computer interactions. In this paper we aim to address the pose estimation problem under some challenging conditions, e.g., from a single image, large pose variation, and un-even illumination conditions. The approach we developed combines non-linear dimension reduction techniques with a learned distance metric transformation. The learned distance metric provides better intra-class clustering, therefore preserving a smooth lowdimensional manifold in the presence of large variation in the input images due to illumination changes. Experiments show that our method improves the performance, achieving accuracy within 2-3 degrees for face images with varying poses and within 3-4 degrees error for face images with varying pose and illumination changes.

Regression from local features for viewpoint and pose estimation

“…Since we are using the labels to define the neighborhood, this is a supervised enforcement of the data manifold constraint. Enforcing the manifold constraints have been shown to highly improve regression results in many applications [3,18,25,9]. However all these applications used vectorized representations of the raw intensity.…”

“…In many of these problems, a regression function is learned from a vectorized representation of the input. For example, in head pose estimation, researchers typically learn regression from vectorized representation of the raw image intensity, e.g., [14,3,8,25,9].…”

Automatic head pose estimation with Synchronized sub manifold embedding and Random Regression Forests

“…In contrast Ben Abdelkader [16] shows a direct way to incorporate label distance related information into the objective functions of LLE and LE. Wang, et al, [17] performs two steps of dimensionality reduction, first an unsupervised step consisting of ISOMAP which is followed by a supervised step using linear Local Fisher Discriminant Analysis (LFDA) [18]. The linear mapping obtained by this method also allows the subsequent projection of out-ofsample examples.…”

Novel Backprojection Method for Monocular Head Pose Estimation

“…Previous head-pose estimation approaches can be grouped into feature-based, appearance-based, or model-based approaches [6][7][8][9][10].…”