Improved 3D face feature-point nearest neighbor clustering using orthogonal face map

This research is part of facial animation development using multi feature-point cluster as a base. Clustering process applied by modifying k-nearest neighbor method into a feature-points nearest neighbor using an orthogonal face map. The use of 3D models of human face and a cartoon character face is meant to observe the reliability of the synthesis of the clustering approach in handling the automation of grouping the face area based on the position of the motion features as its center point. This process will support the adaptive establishment of the area affected by the movement of the weight of each facial feature-points, corresponding to a 3D face model imposed. Weights obtained from each vertex of the clusters formed is used in the deformation process skinning, when a movement in the face is detected during the animation process. We assume that for acquiring a good clustering result modification has to be made in the process of clustering. The result evaluated and compared using regular clustering methods. It shows that using this modification could improved the result visually and objectively.