显著性特征保持的点云模型缩放

Salient Features Preserved Resizing of Point-Based Geometry

为了对现有三维模型的尺寸进行自适应调整以满足具体应用的需求,提出一种基于能量线抽取的显著性特征保持的点云模型缩放算法.首先为输入点云模型指定上、下边界线,并采用距离演化方法在模型上计算出从上边界线到下边界线的距离场;然后依据曲面的局部显著性特征,采用动态规划算法计算出模型的能量场,根据能量场计算一条从上边界到下边界的最小能量曲线;最后将最小能量曲线上的点从模型中删除,并局部拼接完成模型的一次缩放.重复上述步骤,直至模型的尺寸符合要求.实验结果表明,该算法在达到缩放目的的同时既能较好地保护模型的显著区域,又能有效地解决现有方法在模型缩放过程中出现某些部分分辨率过高,导致产生视觉不自然甚至失真现象.

To adjust the size of existing three-dimensional model adaptively for meeting the requirements of various applications, this paper presents a salient features preserved resizing technique for point cloud model by deleting or copying the energy curves. Firstly, we specify the upper and lower boundary points on the point set surface, and construct the distance field for the model based on distance evolution. Then with the local salient features of surface, we evaluate the energy field for the underlying model using dynamic programming technique, and compute the minimum energy curve from upper boundary to lower boundary according to the energy field. Finally, we remove the points which lie on the minimum energy curve, and splice both sides of the curve to finish once model resizing operation. Repeat the above steps until the size of the resulted model meets our requirements. Experimental results show that the proposed algorithm can produce appealing resized results, and it can both protect the significant regions and avoid the problems of existing methods, for example, the resolution becomes too high in some resized parts so that the model looks unnatural and even distorted.