摘要
在增强现实领域,光影一致性的保持十分重要,之前的研究者们主要集中在光照估计然后模拟生成阴影。大多需要知道物体的投射物体的具体形状,很少考虑真实场景阴影和虚拟物体之间的光影相互影响。提出一种原创的面向增强现实的阴影检测算法。首先利用Meanshift算法进行图像预分割,检测阴影边缘。然后有效结合2D和3D信息,基于局部一致性进行误检测去除,最后得到理想的阴影边缘,从而可以逆向生成阴影区域,更好地模拟阴影交互。实验结果证明算法的有效性。
In Augmented Reality （All）, keeping the illumination consistency is very important. Most previous AR techniques about shadow process- ing mainly focus on illumination estimation and then generate the shadow of the virtual objects. Mostly, the shadow interaction between the virtual objects and the real scene has been ignored. Proposes a novel shadow detection algorithm for AR. Firstly, every frame will be segmented with mean-shift method, and then the shadow edges will be detected. The method based on local consistency refines the shad- ow edges combining the 2D information and 3D information effectively. Then, the virtual shadow can be generated from the shadow edges reverse. Experiments demonstrate the stability and effectiveness of our method.
