Our research in visual computing spans a wide range of topics in the fields of computer vision, computer graphics, geometric computing and both image and video processing. A significant theme in our work considers the input, description and editing of solids, surfaces and curves. These are represented analytically, as CAD models and as meshes. Other aspects of our work include the analysis, use and generation of static data such as images, surface meshes and 3D depth scans, as well as time-varying data such as video and 4D scans of moving objects.
Recognising geometric regularities in a reconstructed CAD model can improve its usefulness.
Our expertise in this area has been applied to problems from a range of disciplines including engineering, earth sciences, psychology, biology, medicine and even quantum control.
Reverse Engineering and Computer Aided Design
A major theme of our research concerns reverse engineering of CAD models - regenerating CAD models from scan data. As well as devising a range of general purpose algorithms for this task in areas such as registration, segmentation and curve and surface fitting, we have developed higher level algorithms for analyzing symmetry and design intent of such models, as well as special purpose algorithms for processing bas reliefs. Other recent CAD work undertaken has investigated sketch input of solid models, split line design for moulding and improving boundary curves used in machining.
Analysis of High Dimensional Image Feature Spaces
We have been applying eigenmethods, principle component analysis (PCA) and non-linear dimension reduction variations to many applications. Initially we developed some core algorithms to incrementally build and learn eigenmodels. This has led to further research into computational issues and manifold representation of high dimensional image feature spaces. This work has been applied to articulated human motion analysis, human faces and facial dynamics, biometrics, biology and audiovisual tasks. Recently we have been developing groupwise methods to automatically build 2D, 3D and dynamic PCA models.
Segmenting a mesh model - dividing it into a disjoint set of naturally meaningful pieces - provides a basis for many applications.
Triangle meshes have many applications in computer graphics as well as CAD. We have developed approaches for many geometric operations on mesh models, including noise filtering, segmentation, morphing, texture transfer, parameterization, watermarking and remeshing. Other research of a geometric nature comprises algorithms for construction and analysis of curves, surfaces and solids, fusion of geometric data from different sensor types, object recognition, multiscale representations, shape retrieval, low-discrepancy sampling, object vectorization and the theory of geometric shape measures.
Researching in this Field
- Professor S. Hu
- Dr Y. Lai
- Dr F.C. Langbein
- Professor A.D. Marshall
- Professor R.R. Martin
- Professor P.L. Rosin
- Mr L.R. Semmens
- Dr K. Sidorov
- Dr X. Sun
See what this group is discussing in the School's VLunch Seminar programme.
Particular areas of strength and expertise in the School include:
- Audiovisual face processing
- CAD algorithms
- Characterisations and analysis of shape
- Geometric computing
- Human motion analysis
- Identification of physical and biological systems
- Image and 3D shape retrieval
- Mesh processing
- Point-based modelling
- Quantum engineering
- Reverse engineering of solid shape
- Solid, curve and surface modelling
- Video processing
- Applications of geometric and image processing algorithms with many interdisciplinary partners
- Avoiding antisocial behaviour by modelling crowd behaviour
- CAD algorithms developed jointly with leading UK CAD supplier
- Reverse engineering algorithms used in world leading commercial systems
Recent successful PhD students have submitted the following theses:
- S. Caton - On-demand distributed image processing over an adaptive campus-grid
- L. Truong-Benedikt - Using 3D facial motion for biometric identification
- O. Samko - Low dimension hierarchical subspace modelling of high dimensional data
- J. Quinn - Low-discrepancy point sampling of 2D manifolds for visual computing
Current Grants & Research Projects
|Holder||Project Title||Source||Value (£Ks)|
|Dr PL Rosin & Professor T Wess (OPTOM)||High definition X-ray microtomography and advanced visualisation techniques for information recovery from unopened historical documents||EPSRC||250.91|
|Professor S. Bordas (Engineering), Professor R.R. Martin and Dr F.C. Langbien||Integrating numerical simulation and geometric design technology||European Commission (FP7)||1032.2|
|Dr F. Langbein||Optimal control technologies in quantum information processing||European Commission (FP7)||15.6|
|Professor R.R. Martin, Professor P.L. Rosin and Dr X. Sun||Realistic shape from shading||EPSRC||398.7|
|Professor RR Martin, Professor NJ Avis, Dr PL Rosin, Dr AD Marshall & Dr FC Langbein with Bangor, Aberystwyth and Swansea||RIVIC: One Wales Research Institute for Visual Computing||HEFCW||1244.3|
|Dr X. Sun, Professor R.R. Martin and Professor P.L. Rosin||Robust and sensitive methods for non-rigid and partial 3D model retrieval||EPSRC||387.4|
|Professor R.R. Martin||Structural analysis and interactive composition of visual media||EPSRC||111.76|
|Professor RR Martin and Dr Y Lai||Using mosaicing for finite element meshing||Airbus||45|