This paper presents a sharpness-based method for hole-filling that can repair a 3D model such that its shape conforms to that of the original model. The method involves two processes: interpolation-based hole-filling, which produces an initial repaired model; and post-processing, which adjusts the shape of the initial repaired model to conform to that of the original model. In the interpolation-based hole-filling process, a surface interpolation algorithm based on the radial basis function creates a smooth implicit surface that fills the hole. Then, a regularized marching tetrahedral algorithm is used to triangulate the implicit surface. Finally a stitching and regulating strategy is applied to the surface patch and its neighboring boundary polygon meshes to produce an initial repaired mesh model, which is a regular mesh model suitable for post-processing. During post-processing, a sharpness dependent filtering algorithm is applied to the initial repaired model. This is an iterative procedure whereby each iteration step adjusts the face normal associated with each meshed polygon to recover the sharp features hidden in the repaired model. The experiment results demonstrate that the method is effective in repairing incomplete 3D mesh models.

Image mosaicing is the act of combining two or more images and is used in many applications in computer vision, image processing, and computer graphics. It aims to combine images such that no obstructive boundaries exist around overlapped regions and to create a mosaic image that exhibits as little distortion as possible from the original images. In the proposed technique, the to-be-combined images are first projected into wavelet subspaces. The images projected into the same wavelet space are then blended. Our blending function is derived from an energy minimization model which balances the smoothness around the overlapped region and the fidelity of the blended image to the original images. Experiment results and subjective comparison with other methods are given.

A new approach to the construction of a person's three-dimensional face model from the person's two orthogonal views is presented. This approach is a no hair-style interference method, which constructs the person's face model and hair model independently. The person's fine facial structures, i.e., eyes, nose, and lips, as well as the person's outline in two orthogonal views are obtained. A feature-based generic three-dimensional bald head model is then used to adjust its predefined features to fit the person's fine structures and outline morphologically in two respective views. The independently obtained three-dimensional hair model is put on the morphed bald head model to be the person's three-dimensional geometrical face model. The person's realistic face model is finally obtained by using the texture mapping of the person's image in two orthogonal views onto the person's three-dimensional face model. This paper shows the detail process of the construction of a person's face model from the person's two orthogonal views.

Image mosaic combines two or more images. It has found many applications in computer vision, image processing, and computer graphics. A common goal of the problem is to join two or more images such that there is an invisible boundary around the seam line and the mosaic image is as little distorted from the original images as possible. We propose a new image mosaic method by wavelet multiresolution analysis and variational calculus. We first project the images into wavelet spaces. The projected images at each wavelet space are then blended. Variational calculus techniques are applied to balance the quality between the smoothness around the seam line and the fidelity of the combined image relative to the original images in image blending. A mosaic image is finally obtained by summing the blended images at the wavelet spaces. Experimental results based on our method are demonstrated.

A new approach to the generation of a feature‐point‐driven facial animation is presented. In the proposed approach, a hypothetical face is used to control the animation of a face model. The hypothetical face is constructed by connecting some predefined facial feature points to create a net so that each facet of the net is represented by a Coon’s surface. Deformation of the face model is controlled by changing the shape of the hypothetical face, which is performed by changing the locations of feature points and their tangents. Experimental results show that this hypothetical‐face‐based method can generate facial expressions which are visually almost identical to those of a real face.

A new approach for the generation of feature point-driven facial animation is presented. This approach is based on the construction of a hypothetical face formed by connecting face feature points into a net and representing each facet of the net by a mathematically expressed surface to control the deformation of a real face model. The deformation is controlled by changing both the locations of the feature points and their tangents. The experimental results show that the hypothetical-surface-based method can generate almost identical facial expressions to a real face.