Complex 3D General Object Reconstruction from Line Drawings

Linjie Yang, Jianzhuang Liu, Xiaoou TangICCV 2013 · OpenReview

Abstract

An important topic in computer vision is 3D object reconstruction from line drawings. Previous algorithms either deal with simple general objects or are limited to only manifolds (a subset of solids). In this paper, we propose a novel approach to 3D reconstruction of complex general objects, including manifolds, non-manifold solids, and nonsolids. Through developing some 3D object properties, we use the degree of freedom of objects to decompose a complex line drawing into multiple simpler line drawings that represent meaningful building blocks of a complex object. After 3D objects are reconstructed from the decomposed line drawings, they are merged to form a complex object from their touching faces, edges, and vertices. Our experiments show a number of reconstruction examples from both complex line drawings and images with line drawings superimposed. Comparisons are also given to indicate that our algorithm can deal with much more complex line drawings of general objects than previous algorithms.

Introduction

Interpreting 3D structure from 2D line drawings is a fundamental challenge in computer vision. While humans perform this task effortlessly, computational models often struggle with the high-dimensional search space and the presence of local minima. Traditional methods are frequently restricted to manifold objects—solids where every edge is shared by exactly two faces. However, real-world objects often feature non-manifold geometries, such as internal faces or dangling edges, which complicate the reconstruction process.

Methodology: DoF-Based Decomposition

The core innovation of this research is a "divide-and-conquer" strategy based on the Degree of Freedom (DoF) of 3D objects. The authors define a "primitive solid" as a structure with a DoF of 4, meaning its entire geometry is determined once the depth of four non-coplanar vertices is known.

Decomposition of complex line drawings into simpler primitives.

The reconstruction pipeline follows these steps:

  1. Face Extending: Starting from adjacent faces, the algorithm "grows" sub-objects by adding faces that maintain a DoF of 4. This isolates the fundamental building blocks of a complex sketch.
  2. General Decomposition: To handle overlapping primitives, the problem is modeled as a Maximum Weight Clique search. This identifies the optimal set of non-conflicting sub-objects that explain the maximum number of faces in the drawing.
  3. Reconstruction and Merging: Each simplified primitive is reconstructed independently using geometric regularities like planarity and parallelism. These components are then merged at their shared boundaries to form the final complex object.

Experimental Results

The proposed method was evaluated on synthetic complex drawings, real-world images with superimposed lines, and noisy sketches. A key success was the reconstruction of a complex castle model, a task that typically causes global optimization-based methods to fail due to the complexity of the vertex depth search.

3D reconstruction of a complex castle model.

The algorithm demonstrated significant robustness to "shaky" input lines, as geometric constraints effectively corrected vertex positions during the 3D optimization phase. Furthermore, the system is computationally efficient, processing complex models in approximately one minute on standard hardware.

Conclusion and Future Work

By decomposing complex geometries into low-dimensional primitives, this approach significantly expands the range of objects that can be reliably reconstructed from line drawings. While currently limited to orthogonal projections and planar faces, future work aims to incorporate perspective projection and curved surface support to further enhance its applicability to real-world photography and diverse design sketches.

Summary updated

Abstract

An important topic in computer vision is 3D object reconstruction from line drawings. Previous algorithms either deal with simple general objects or are limited to only manifolds (a subset of solids). In this paper, we propose a novel approach to 3D reconstruction of complex general objects, including manifolds, non-manifold solids, and nonsolids. Through developing some 3D object properties, we use the degree of freedom of objects to decompose a complex line drawing into multiple simpler line drawings that represent meaningful building blocks of a complex object. After 3D objects are reconstructed from the decomposed line drawings, they are merged to form a complex object from their touching faces, edges, and vertices. Our experiments show a number of reconstruction examples from both complex line drawings and images with line drawings superimposed. Comparisons are also given to indicate that our algorithm can deal with much more complex line drawings of general objects than previous algorithms.
BibTeX
@inproceedings{yang2013complex,
  author = {Linjie Yang and Jianzhuang Liu and Xiaoou Tang},
  title = {{Complex 3D General Object Reconstruction from Line Drawings}},
  booktitle = {IEEE/CVF International Conference on Computer Vision},
  year = {2013},
  url = {https://openreview.net/forum?id=BJ4xr-MdZH},
}
Figure