NLME: a nonlinear motion estimation-based compression method for animated mesh sequence

This paper proposes an efficient compression algorithm for animated three-dimensional meshes by introducing nonlinear transformations to model the motion field of deforming patches. First, a segmentation process is applied to separate the 3D model into different patches which have similar motion patterns through the sequence. Next, the motion of the resulting patches is accurately described by a nonlinear motion estimation model. The main idea is to exploit the temporal coherence of the geometry component by using a nonlinear predictor in order to get better approximation of vertex locations. Nonlinear motion transforms are computed at previous frame to match the subsequent ones. Moreover, an adaptive bit allocation algorithm is employed to determine the near-optimal number of bits for quantizing the prediction errors. The number of quantization bits for each segmented patch is determined by analyzing the geometry complexity of the patch and the statistical properties of the prediction errors. Finally, an extensive experimental study has been conducted to evaluate the coding efficiency of the proposed compression scheme. Simulation results demonstrate that the proposed method is very efficient in terms of rate-distortion performance, particularly for the animated models with non-rigid deformations, and outperforms the state-of-the-art methods.