Non-rigid registration using AAM and factorization method for augmented reality applications

This paper presents a novel non-rigid registration method for augmented reality applications using AAM and factorization method. The method can be divided into two stages: offline construction of 3D shape basis and online estimation of the 3D pose parameters together with the 3D shape coefficients. In offline stage, we get the training data with the use of the AAM algorithm, then we use factorization method to obtain the object 3D shape basis set from image sequences. In online stage, given an initial 3D shape coefficients and the rotation matrix, we can get the non-rigid object coordinates under the world coordinate systems by projecting the 3D points which are the linear combination of the 3D shape basis to the 2D image. Then we use Levenberg-Marquardt algorithm to optimize the 3D shape coefficients and the 3D pose parameters simultaneously. Some experimental results demonstrate this method is effective and useful for augmented reality applications.