Nonuniform Segment-Based Compression of Motion Capture Data
This paper presents a lossy compression method for motion capture data. Each degree of freedom of a motion clip is smoothed by an anisotropic diffusion process and then divided into segments at feature discontinuities. Feature discontinuities are identifi
- PDF / 532,765 Bytes
- 10 Pages / 430 x 660 pts Page_size
- 8 Downloads / 204 Views
Abstract. This paper presents a lossy compression method for motion capture data. Each degree of freedom of a motion clip is smoothed by an anisotropic diffusion process and then divided into segments at feature discontinuities. Feature discontinuities are identified by the zero crossings of the second derivative in the smoothed data. Finally, each segment of each degree of freedom is approximated by a cubic B´ezier curve. The anisotropic diffusion process retains perceptually important high-frequency parts of the data, including the exact location of discontinuities, while smoothing low-frequency parts of the data. We propose a hierarchical coding method to further compress the cubic control points. We compare our method with wavelet compression methods, which have the best compression rates to date. Experiments show that our method, relative to this work, can achieve about a 65% higher compression rate at the same approximation level.
1 Introduction Motion capture data is now widely used in games, simulations, and animation. A motion is represented in motion capture data as an ordered sequence of frames. Each frame specifies all positions, orientations, and joint angles for a pose at a certain point in time. Each pose parameter is called a degree of freedom (DOF). How to efficiently store such data in usable form with limited memory resources is still a challenging problem. Compression is possible for motion capture data because of two properties: temporal coherence and DOF correlations [1]. Temporal coherence is present in motion capture data because all DOFs are sampled simultaneously from the same continuous physical motion. Due to the physical limits inherent in its origin, motion data can be approximated by continuous functions, such as splines, wavelets, or linear dynamical systems, which can capture temporal coherence. Correlation between different DOFs is caused by the fact that all DOFs are related to the same physical structure. Also of importance is the fact that some DOFs are more important perceptually than others. Therefore, it is possible to remove redundancy between DOFs using dimensionality reduction techniques and also to omit some uninfluential DOFs without significantly affecting the perceptual accuracy. There are two kinds of compression: lossless and lossy. Lossless compression has zero error but cannot achieve high compression rates. Lossy compression techniques can achieve higher compression rates but always involve a tradeoff between compression rate and error. The goal of a lossy compression method such as the one we present here is to achieve the best ratio between compression rate and error. Compressing in G. Bebis et al. (Eds.): ISVC 2007, Part I, LNCS 4841, pp. 56–65, 2007. c Springer-Verlag Berlin Heidelberg 2007
Nonuniform Segment-Based Compression of Motion Capture Data
57
both temporal and DOF spaces can achieve better compression rate than temporal compression only [2]. However, compressing individual DOFs makes reuse of the motion data and update of the database easier.
Data Loading...
