Abstract: Based on the theory of visual hulls，this paper presents a method for point based modeling and rendering from real objects’ images.For point based modeling，a series of images is taken for the object from a set of viewpoints.Then，this paper uniformly point-samples the visual hull formed by these images.During the sampling，a regular-interval index table is used to organize the silhouette of each image，which increases the modeling efficiency.For point based rendering，a series of discrete reflectance properties of the points is obtained from the images under various directional lighting conditions.Then，realistic rendering is achieved with an improved algorithm to interpolate the reflectance properties.The experimental results show that the method has great efficiency，and provides high realistic rendering.

Key words: point based models, visual hull, layered depth image, environment mapping

摘要： 从视觉凸壳的理论出发，提出了一种基于图像的真实物体的点云建模和绘制方法。在建模方面，首先从不同视点采集目标物体的图像，然后对采样图像所形成的视觉凸壳进行均匀的点采样。同时，利用等间隔索引表来组织每幅采样图像的轮廓边，从而提高了建模效率。在绘制方面，首先分析目标物体在各种不同光照条件下的图像，并得到点云模型的离散反射属性，然后通过优化的插值方法实现真实感的绘制效果。实验表明，该方法的建模速度快，绘制结果具有很强的真实感。

关键词: 点云模型, 视觉凸壳, 层次深度图, 环境映射