An example-based approach to 3D man-made object reconstruction from line drawings

Abstract

3D reconstruction from a single 2D line drawing is an important but challenging problem in computer vision. Existed methods usually fail when line drawings contain large degree of noise named sketch errors. In this paper, we present an example-based approach to reconstructing 3D object, either planar or curved, from a single-view line drawing with sketch errors. Our method is to first decompose the input line drawing into primitive components and cluster them into local groups, then turn each group into 3D shapes via a novel example-based algorithm, and lastly integrate those recovered 3D shapes from all groups to build a final complete 3D model. Comprehensive experiments on a wide range of line drawings depicting man-made objects show that the proposed approach outperform previous work, especially for line drawings containing large degree of sketch errors.