3D face modeling from single image based on discrete shape space

Abstract

In this article, we propose a novel 3D face modeling method which constructs a new 3D face model from a low-dimensional feature space consisted of a large set of blend shapes based on the discrete shape space theory. The details of original face features are completely retained during the modeling process and a large number of new natural faces are constructed by several face samples. The optimization process of our method is independently decoupled for different facial attributes (identity, expression, and head pose), which improves the application flexibility and reduces the probability of it falling into a local optimal situation. The new facial data with new attributes are constructed based on the geodesic path search in discrete shape space with sufficient freedom and accuracy. In experiments and applications based on public databases (Helen, LFW, and CUFS), the modeling results show our method can provide high-quality 3D face model, with enough freedom for face expression editing and natural facial expression animation from a small facial sample set. K E Y W O R D S 3D face modeling, discrete shape space, facial landmarks, geodesic path 1 INTRODUCTION 3D face modeling is an important issue for many applications such as game production, movie making, and cartoon animation creation. The face data can be edited using the reconstruction face model, which is a powerful tool for imaginative creative activities and business applications (DAZ3D, Maya, and Meitu). It is generally difficult for building a high-quality 3D face model from a single image, because the faces in images have different head poses, face scales and the geometric features are incomplete, which is termed "ill-posed." To remove the influence of ill-posed features, the prior knowledge of the 3D face data should be used to guide the modeling process. Simultaneously, to synthesize the 3D face object from an image, the modeling method is required to match the geometric features (facial landmarks or contours) between the 3D face data and 2D facial image. In general, a standard 3D face object (from Facewarehouse or Bospho-rusDB) has more than 10,000 vertices, it cause that the high computational complexity of directly 3D face modeling in original 3D face objects, which it is unattainable. Therefore, we propose a 3D face modeling method which represents 3D face by facial landmarks can quickly and correctly construct 3D face model based on the discrete space in this article.

Face Recognition in Single Sample Per Person Fusing Multi-Scale Features Extraction and Virtual Sample Generation Methods

“…There are several methods to solve the problem of SSPP face recognition: virtual sample generation, patch-based method, an image segmentation method, generic learning method, and so on [7][8][9][10]. Among these methods, virtual sample generation is a commonly used method to generate more samples.…”