Physically-Based Modelling for Graphics and Vision

Our research in physically-based modelling for graphics and vision

Physically-based modelling uses physical laws and numerical analysis for computer animation and computer graphics.

In recent years, physically-based modelling has become one of the most important elements of digital applications in the creative industries, largely because of the realism and automation that it offers.

The demand for physically-based simulation is increasing at a rapid rate in the creative industries, particularly in computer games and movie production.

The latter uses some of the world’s most powerful computers: for example, in the movie Avatar, over 60% of the frames were generated using computer simulation; over 40,000 processors were used for simulation and rendering.

By combining the research in computational mechanics, computational fluid dynamics and computer graphics, we develop better algorithms and new tools for graphics and vision applications, in particular computer animation.

We have close research collaborations with leading computer scientists from Tsinghua University, China, Cardiff University, UK, and North Carolina University, USA. 

Research areas

  • Realistic fluid animation
  • Fisheye video correction

Realistic fluid animation

We focus on the development of novel computational technologies and highly efficient codes, to produce visually-realistic animations of various practical fluids. The goal is to provide practical tools for animators, designers and directors to create visually impressive 3D animations more easily and more quickly. Our research covers gaseous and liquid fluids, multiphase and multi-component flow, and fluid-structure interaction etc. 

This research is in collaboration with Prof. Shimin Hu at Tsinghua University, China, and Prof. MC Lin at North Carolina University, USA. 

Fisheye video correction

Comparing with standard camera lenses, fisheye lenses have the advantage of covering a wide field of view (180 degree). However, this is compromised by significant barrel distortion, making fisheye video difficult to interpret. We have developed the first practical method to recover, without loss of visual information, natural-looking video sequences from fisheye videos. The invention has promising potentials in surveillance monitoring, movie/TV production, sport coverage etc. 

This research is in collaboration with Prof. SM Hu at Tsinghua University, China, and Prof. RR Martin at Cardiff University, UK.