Improved SIFT matching for pose robust facial expression recognition

This paper presents a method to enhance the recognition of facial expressions under varying poses, based on the scale invariant feature transform (SIFT). Affine transformation is used to recover the effect of the unknown depths of keypoints extracted from the 2D images, allowing the problem to be reduced to an absolute orientation with scale problem, which is solved using SVD. The proposed Π-SIFT method has been experimented with BU-3DFE database acquired with different perspective camera views and has been shown to outperform regular SIFT and D-SIFT.