Výzkum (Text pouze v angličtině)

Predictive Rendering

One of the main research areas within CGG is predictive rendering, the sub-discipline of computer graphics that attempts to provide reliable predictions of object appearance. In an industrial setting, this sort of capability is mainly needed for reliable virtual prototyping applications. Apart from a number of other technological keystones, requires accurate models of reflectance, as well as robust and unbiased simulations of light transport. Both these topics are active research areas within CGG, as well as the general research area of which physical effects have the capability of influencing object appearance sufficiently to warrant their simulation in a predictive renderer. ART, our predictive rendering research system, can be downloaded via its project page.

Robust and Efficient Light Transport Simulation

We focus on improving the robustness and efficiency of realistic rendering with global illumination. The goal is to design algorithms that can render environments with complex geometry, lighting, and materials in acceptable computation time. In the past, we have focused on illumination caching (SIGGRAPH 2008 Course, book, EGSR 2009 paper) and many-light rendering (SIGGRAPH Asia 2009, SIGGRAPH Asia 2010, SIGGRAPH 2010, EG 2012, EG 2013 STAR).

More recently, we have shifted our attention to robust Monte Carlo-based algorithms. Our major results in this area include Bidirectional Photon Mapping, Vertex Connection and Merging, Joint Importance Sampling, On-line Learning for Importance Sampling, Unified points, beams, and paths.

We regularly present reviews and surveys on global illumination and light transport simulation: SIGGRAPH 2010 Course, SIGGRAPH 2012 Course, EG 2013 STAR, SIGGRAPH 2013 Course, SIGGRAPH 2014 Course, ACM TOG 2014 paper.

Segmentation and visualization of medical volume data

We specialize in processing of data generated by medical tomographic scanners (CT, MRI). Currently we work on CT enterography data processing and segmentation of patological structures in liver and kidneys. We use modern GPGPU technologies (CUDA and OpenCL) - implementation of fast nonlocal means denoise, interactive segmentation based on watershed transformation, etc.