J. Zhang, J. Xu, and P. Fisher (USA)
octree, animation, geometry compression
The original octree-based animation compression method partitions the vertices into different octree nodes based upon their locations relative to the bounding box of the current frame to encode. Because the positions of vertices change from frame to frame, two neighboring vertices partitioned to the same octree node in a frame may be partitioned into two different octree nodes in another frame, and later be partitioned to the same octree node again. This can cause a larger distortion in the compressed data resulting in a lower compression ratio. We propose a static partitioning octree-based animation compression method in which a vertex’s belongingness to an octree node in each depth of octree is predetermined and fixed across all frames. Given the same threshold that controls the quality of compression, results show that the static partitioning method consistently outperforms the original method ranging between 1.2 to 1 and 2.2 to 1 in five animation sequences. Experimental results show that the selection of the base frame for static partitioning has significant impact on the compression ratio. The significant improvements achieved in this study show that there is significant room for further improvements to 3D animation compression.
Important Links:
Go Back
