Facial Pose Image Synthesis with a Single Image Using Tensor Subspace Approximation Method

摘要

In this paper we proposed a new method called tensor subspace approximation (TSA) to do the facial pose image synthesis with a single image. We make use of the morphable model to all the image data previously in order to separate the image information into two parts: shape and texture. This method applies high-dimensional PCA method which deal with the objective data in the form of tensor to estimate the identify vectors of each samples among the varied images. MaVIC Database (KAO-Ritsumeikan Multi-angle View, Illumination and Cosmetic Facial Database) is used in our experiments and results are improved the accuracy of the method.