Traditional portrait bas-relief modeling methods are limited by the shape of the input 3D model. In this paper, we introduce a novel approach to generate portrait bas-relief with various poses and expressions. In the alignment stage, we segment the input model into several meaningful regions so that the facial features of each region can be separately aligned. To guaranty the quality of expression simulation, we consider both the landmarks of fitted 3DMM model and input portrait image to drive the bi-Laplacian deformation process. After alignment, we apply normal enhancement in the deformed model, and the bas-relief is achieved by solving a sparse linear system in the normal field. Furthermore, the input model is transformed into a compressed height field to be the constraint in bas-relief modeling. The final bas-relief preserves the subtle details and depth structure. Experimental results demonstrate that our method is effective in generating the attractive appearance of portrait bas-reliefs.

传统的肖像浮雕建模方法受输入3D模型的形状限制。在本文中，我们介绍了一种新颖的方法来生成具有各种姿势和表情的肖像浅浮雕。在对齐阶段，我们将输入模型划分为几个有意义的区域，以便可以分别对齐每个区域的面部特征。为了保证表情模拟的质量，我们同时考虑了拟合的3DMM模型的界标和输入肖像图像来驱动双Laplacian变形过程。对齐后，我们在变形模型中应用法线增强，并且通过解决法线场中的稀疏线性系统来实现浅浮雕。此外，将输入模型转换为压缩高度场，以作为浅浮雕建模中的约束。最终的浅浮雕保留了细微的细节和深度结构。实验结果表明，我们的方法可以有效地产生肖像浮雕的诱人外观。