Computer Animation and Virtual Worlds

This paper proposes an automatic approach to extract keyframes from a motion capture sequence. The method treats the input sequence as motion curves; reduces the dimension of the motion capture data using Principal Component Analysis; and obtains the most salient parts of these curves using a new multiscale metric, called 'motion saliency'.

We propose a novel surface deformation method based on a feature sensitive (FS) metric. Using a FS Laplacian operator and the tetrahedron constraints, our new method preserves features of triangular meshes very well during deformation.

This paper presents a skeletal rigid skinning approach. First, a skeleton extraction technique is introduced that produces refined skeletons appropriate for skinning. Then, to avoid shortcomings of previous skinning approaches, an efficient and robust rigid skinning technique is developed that applies blending patches around joints. To achieve real time animation, all steps of the rigid skinning algorithm have been ported on the GPU. Finally, an evaluation of the proposed techniques is presented against four criteria: performance, quality, scope, and robustness.

Nowadays, the marker-based mobile augmented reality should rely on time-consuming iterative scheme in pose estimation procedure for accuracy and stability due to the absence of redundant point correspondences. We present an efficient LUT-based pose estimation solution which achieves high stability in the presence of noise better than the most robust and accurate iterative solutions available in the field, with the same level of accuracy and a much lower computational complexity.

Combining path planning and motion graphs introduces several problems. We will identify two problems and propose possible solutions. The first problem is selecting an appropriate distance metric for locomotion synthesis. We have evaluated three common distance metrics and propose a set of guidelines. The second problem is that there is no single point on the body that can follow a generated path without causing unnatural artifacts. We will show that our solution, which uses path abstractions, generates significantly better animations.