Motion level of detail

Abstract

As computer graphics technology improves, methods for animating scenes with a large number of articulate figures are attracting more attention in movies, games and VR applications. However, an enormous number of characters can hardly be simulated in real-time. Several researchers have tried to improve the performance of animation by using an imagebased technique such as ‘impostor’. This method improved the speed of the animation; however, the accuracy, memory and interactivity problems related to motion remain to be resolved. In this regard, a ‘motion simplification’ for an articulate figure is proposed, which not only improves the speed of the animation but also conserves the features of the original motion. Research on crowd animation can be classified into two major categories. An approach that focus on offering realism of the crowd motion and another technique that improves the speed of the animation. During the past decade, a wealth of research on realism such as crowd behavior and collision has been presented. Recently, for the sake of GPU evolution, research on improving speed has also received much attention. According to the result of our experiment, we determine that the number of joints have a much greater effect than anticipated on the animation speed. In order to resolve this problem, a new level-of-detail (LoD) technique named ‘motion LoD’ is proposed. Motion LoD is a joint-reduction technique operated in the preprocessing stage. The whole process consists of motion clustering and simplification. At the motion clustering process, motion sequence is analyzed and clustered by using two approaches joint posture clustering (JPC) and mean shift motion clustering. These clustering processes extracts the key postures of a motion and automatically generates the priority of joint reduction. The clustered postures are directly utilized as an initial posture of the proposed motion simplification. At the stage of motion simplification, ...