立体视频对象分割及其三维重建算法研究*

为更加有效分析立体视频对象，本文提出了一种基于离散冗余小波变换的立体视频对象分割算法，首先采用离散冗余小波变换提取特征点结合DT网格技术的视差估计方法，获得了可靠的视差场，再利用视差信息对立体视频中静止对象进行分割。对于立体视频序列中的运动对象，采用离散冗余小波提取运动区域的方法进行分割。实验结果表明，本算法对有重叠的多视频对象具有较好的分割效果，可同时分割静止物体和运动物体，具有较好的精确性和鲁棒性。对于分割出的立体视频对象，结合深度信息对其进行三维重建，得到较好的三维效果。

Research for stereo video object segmentation and 3D reconstruction

For more effective analysis of stereo video object, this paper put forward a discrete redundant wavelet transforms based stereo video object segmentation method. First, the method obtained the disparity map by discrete redundant wavelet transforms and the disparity map was used to do video object segmentation. For the moving objects in the stereo video sequence, the discrete redundant wavelet transforms was used to extract the motion region. Experimental results show that the method can not only segment the overlapping objects, but also can segment the stationary objects and moving objects at the same time with better accuracy and robustness. According to depth map, the visible scene surface is represented with a para-metrically deformable, spatially adaptive, wireframe model.