基于光场渲染的动态3D目标重构技术

3D dynamic object reconstruction technique based on light field rendering

提出了一种基于光场渲染的动态3D目标重构的方法.目前基于图像建模的方法对于复杂场景难以建模,而基于图像渲染的方法因数据量大不利于动态场景的实时渲染.因此采用了模型和图像相结合的方法,从多视点视频图像中重建动态3D模型,采用光场映射算法对重建的3D模型进行光场采样,然后对样本分解压缩.实验结果表明,在保证真实感的同时,减少了数据量,并可实现动态3D目标的重构.

This paper proposes a method of 3D dynamic object reconstruction based on light field rendering. At the present, for IBM（image based modeling） it is difficult to construct models for complex scene, while for IBR（image based rendering） real-time rendering of dynamic scene is infeasible due to a large amount of data. Thus in this paper, we introduce the method of modeling associated with images. We reconstruct 3D dynamic model from multi- view video images, utilize LFM（light field mapping） algorithm to sample light field of 3D model, and then decompose and compress the samples. The experimental results show that the proposed method can reduce amount of data while guaranteeing the realities, and realize reconstruction of 3D dynamic object.