中文摘要：

为了方便快捷地重建出输入图像中的物体, 基于单幅图像, 提出一种利用二维手绘线画图重建生成三维对称自由形体的方法. 在以单幅图像为背景的画布上, 用户首先根据图像中的物体位置为待重建模型指定一个对称面;然后为确定了对称面的物体绘制出其各个子部分的手绘线条, 每个子部分通常由4 条手绘曲线表示, 即1 对对称曲线和1 对非对称的一般曲线, 对称曲线的深度坐标信息可以利用物体对称面信息进行计算, 一般曲线的深度坐标信息则可以由同一子部分的三维对称构造曲线计算得到; 利用恢复了深度信息的三维构造曲线, 自由形体的每个部分可以分别由1 对对称构造曲线和1 对非对称构造曲线利用旋转混合方法生成其参数曲面, 并离散化为三角网格表示; 根据输入图像中物体的纹理信息进一步合成自由形体每一部分的表面纹理, 最终得到三维物体. 实验结果表明, 该方法得到的三维自由形体具有较强的真实感.

英文摘要：

To create real-word objects from a single input image, with the help of 2D sketches, a novel 3D reconstruction approach for symmetric free-form objects is presented in this paper. Upon the sketching plane defined by the input image, users first assign a symmetry plane for the reconstructed model and draw the sketch curves for each object component interactively. Each shape component can then be created from four sketch curves, that is, a pair of symmetric curves and a pair of non-symmetric general curves. The 3D depth information of the symmetric curves can be determined with respect to the symmetry plane, and the 3D co-ordinate information of the general curves can also be calculated due to the symmetric curves of the same shape component. According to the recovered 3D constructive curves, each object component can be gener-ated by an improved cross-sectional surface blending scheme and be represented as a triangular mesh. The texture of free-form objects can finally be synthesized from the input image. Experimental results show that our proposed method can generate the real-world objects from a single image effectively.