Fast Image Segmentation and Animation Generation Algorithm Based on Depth Image Sequence

In this paper a continuous automatic segmentation method for color sequence slice images of Chinese virtual human is proposed. Its main idea is based on region growing algorithm, and animation is a new algorithm for segmenting moving target images from complex background by using sequence images. Firstly, the animation uses Markov random field (MRF) model and continuoushneprocess to establish a more accurate animation objective function, and uses parallel computing method to calculate the velocity field. Then, based on this, the parallel segmentation of the target animation is realized. Depth image fitting or depth image segmentation is to divide point cloud data with the same geometric features into the same area and perform surface fitting. There are two main methods for deep motion image segmentation: one is based on edge segmentation, and the other is based on region growing. Due to the characteristics of the depth image acquisition method, the point cloud data is often discontinuous and contains more noise. Using the edge based segmentation algorithm, the boundary point segmentation can be effectively realized only when the processed point cloud data has animation continuity and less noise points. In this paper, a method of animation parameter optimization calculation based on depth image sequence is proposed. According to the depth image sequence under different attitudes, the optimization function of animation parameters and attitude parameters is designed, and the deformation results under different attitudes are comprehensively considered. By alternately optimizing the animation parameters and attitude parameters, the animation parameters are output until the objective function converges.