3DMM Fitting for 3D Face Reconstruction

Abstract

Most facial recognition techniques are self-contained on a three-dimensional model to ensure the challenge posed by facial expressions that are variable depending on the situation. In this research we aim to provide the latest SDK package which is used to identify faces with attention to flipping because the face changes its expressions and using the technique of 3D-expandable Model 3DMM we can introduce facial changes, 3DMM also enables us to isolate identity variations from those resulting from changes in facial expressions. We face two problems that need to be addressed: accurate measurement of the parameters of the situation and computational efficiency. When the verification is performed with the adjustment, a new face view is created where the situation is corrected and the expression is disabled to define the expression we provide two methods for it. The first depends on the prior knowledge to illustrate the neutral expression image of the input image. While the second method is based on the idea of verification on the transfer of expression of the exposed face to the probe. Experiments using neutral and equivalent view with the FR SDK commercial standard are demonstrated on two-face databases, PIE and AR, thus, demonstrating a significant improvement in the experimental SDK's performance in terms of expression.