摘要
视频对象平面 (VideoObjectPlane ,VOP)的生成是基于对象的编码技术以及基于图像内容的操作的基础。提出了以对象运动特征 ,在 2D +T三维空间进行图像分割的VOP生成技术 ,特别是在摄像机运动的动态场景中 ,以全局运动估计技术补偿摄像机的运动 ,同时以边缘图像与差分图像进行匹配 ,克服对象高速运动时背景的遮拦与显露造成的影响 ,准确地生成VOP图像。
Object-oriented coding technology and content-based functionalities are based on VOP generation. This paper proposes a novel VOP generation technology based on the image segmentation with motion information in 2D + T three-dimension space. Firstly, this technology compensates cameras' motion with global motion estimation technology in moving scenes. At last, it compares edge image with different image by Hausdorff distance in order to overcome background's occlusion and presentation. The VOP's obtained by this algorithm is more accurate than by other technologies. The results of processing real scene images demonstrate the performance of the algorithm.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2001年第2期74-77,共4页
Systems Engineering and Electronics
