This paper proposes a video-based rendering algorithm using hybrid light-field and geometry rendering.It synthesizes the first frame of a new view with a light-field rendering algorithm, and computes the foreground region by using the Gaussian Mixture Model and geometry computing.Updating the foreground region in the last synthesized frame results in a frame of the new view.This scheme effectively avoids re-rendering the background region which occupies a large area but changes slowly, and hence improves the rendering performance.To eliminate the accumulated error, a new cyclic pattern of “the first frame and the subsequent frames” is employed.In the meantime, the distribution of the foreground points of the scene is gained in a cyclic pattern, and is used to divide the scene of the next cycle.Experimental results show that our approach can synthesize high-quality images efficiently.%提出一种基于光场与几何混合绘制的视频绘制算法.算法以光场绘制算法计算起始帧,以新视点的上一帧为基础更新前景;结合高斯混合背景建模和场景几何计算方法来计算新视点的当前帧的前景区域,避免了重复绘制占据图像大部分而变化缓慢的背景区域,提高了视频绘制的效率.为了消除累积误差,采用"起始帧+后续帧"的循环模式,同时在循环中统计前景点在场景中的分布,自适应地划分下一个循环的场景层次.实验结果表明,文中算法效率高,所生成的图像质量好.
展开▼
