基于骨架的单幅图像三维建模

Skeleton-Based Single Image 3D Reconstruction

为了快捷直观地对单幅图像进行三维重建,论文提出一种基于骨架的单幅图像三维建模方法。首先提取图像中目标物体的轮廓,将其分为主体和侧肢组件;对主体组件提取骨架,结合轮廓的曲率和几何特征,寻找最优"轮廓-骨架"点对,将骨架进行空间提升,再依照点对进行自动生长并网格化,侧肢组件进行参数化模板建模;最后合并组件、纹理映射得到三维模型。实验结果表明,该方法能够很好地保持原二维轮廓的曲率和几何特征,提高了建模效率。

In order to reconstruct 3D model with single image quickly and intuitively,a skeleton-based single image 3D reconstruction method is proposed. First,the contour of the object in the image is extracted and divided into the main body and the side limb components,then,the skeleton of the main component is calculated. by combining the curvature and geometric feature of the contour, the optimal contour-skeleton point pairs are determined. The skeleton is automatically elevated into the depth space and all the contour-skeleton point pairs grow with the same rule into a 3D mesh. The side components are reconstructed by parameterization template method. The final 3D model is acquired after component merging and texture mapping. Experimental results show that the proposed method can maintain well the curvature and geometric features of the 2D contour of the object,and improve the reconstruction efficiency.