Abstract: This paper presents a method of facial expression transfer using expressions from same person. Its target is to improve the quality of facial expressions in photos. The inputs of this method include two face images, neural face database and template blendshape face model. Firstly, according to the contour of face area, a near frontal face image is chosen from these two input face images. Secondly, based on the neutral face database and template blendshape face model, a specific blendshape model for this person is generated. Thirdly, using this blendshape model, the 3D face model of the input face is generated and refined using RBF mesh deformation. Finally, with these two generated 3D face models, the two input images are distorted and facial expression transfer image is generated. The experimental results show that the proposed method can generate natural expressions transferred images and improve the quality of facial expressions in original photos.