Visualization and GraphicsInteractionDept ICSFaculty of ScienceUU

//webspace.science.uu.nl/~telea001/Site/TopBar

Interactive 3D Shape Rotation Using 2D Skeletons

Rotating 3D shapes is difficult. Imagine a typical 3D viewer in which you have a complex 3D shape. How to ``specify`` the rotation of this shape ``easily''?

Typical mechanisms include virtual trackball, which is easy to use but hard to control, or axis-aligned rotation, which is easy to use and control, but limited in flexibility.

Shape-based rotation

We propose to support users in rotating shapes along their natural symmetry axes. In detail

  • local symmetry axes are computed from a 2D projection (view) of any 3D shape, using image skeletons
  • these 2D axes are back-projected to 3D using a simple approximation
  • the shape is rotated along the emerging 3D axes

Pipeline

Our skeleton-based rotation tool is surprisingly simple to implement. The pipeline is below:

  • select any viewpoint, e.g. using trackball (a-c)
  • compute the silhouette of the 3D shape in that view (d)
  • compute the 2D skeleton of the silhouette (d)
  • fit a 2D line to skeleton points close to the mouse pointer (e)
  • backproject the 2D line to 3D (e)
  • rotate the shape around the 3D line based on mouse motion (f)

Features

This method is surprisingly simple, fast, and generic, since

  • 2D image skeletons can be computed in real-time (see CUDASkel)
  • it works for anything whose 2D silhouette can be computed (polygon meshes, polygon soups, splats)
  • it only requires a click and drag in the window (so, suitable also for touch interfaces)
  • all computations can be done very efficiently on the GPU (NVidia CUDA)
  • it integrates with any 3D renderer

Videos

A video illustrating the interactive 3D axis rotation is available below

Skeleton based 3D rotations

Implementation

The method is implemented in C++ using OpenGL1.1 and NVidia CUDA. Sample code is available here.

Publications

Interactive Axis-Based 3D Rotation Specification Using Image Skeletons X. Zhai, X. Chen, L. Yu, A. Telea. Proc. GRAPP, 2020