Three-Dimensional Modeling of Faces Utilizing Robust Characteristics

Abstract

The modeling of human faces is an integral part of a variety of computer applications, from animation and entertainment programs to security and facial recognition apps. This paper presents an alternative method that, given a photo as input, extracts a set of landmarks, anatomical points extracted from a human's face, modifying a triangular polygonal mesh using principles of computer graphics so that it resembles the face whose characteristics have been extracted without requiring additional input from the user. After the creation of the initial mesh, this method, implemented in Python for testing purposes, utilizes principles of three-dimensional mesh modeling such as the use of δ-coordinates to allow the user to perform controlled modifications in the mesh by moving the landmarks. The meshes generated by this method maintain the format and colors of the original face, extracted from the photo offered as input, with tests showing the resistance of the mesh to deformation, as well as corroborating that there is little to no effect on the speed of how modifications are performed based on the degree of the vertex or by how much it is moved.