Optics-inspired rendering techniques can yield fast and realistic real-time rendering algorithms.
Non-Invasive 3D Acquisition of Gas Flows
The Eikonal rendering forward model can be computationally inverted, resulting in a technology for the non-invasive 3D imaging of dynamic gas flows.
Computational processing allows for the decomposition of images with complex inter-reflection patterns into their constituting views. The resulting single-shot surround imaging techniques can more efficiently use todays large scale sensors to record additional information.
Mirror systems can be used to perform surround geometry and reflectance scanning without moving parts and using only a single camera and projector.
Reconfigurable Plenoptic Imaging
An optical add-on that turns a standard camera into a snapshot HDR, multi-spectral, polarization, or light field imager.
Computational Imaging - Interdisciplinary Research at the Intersection of Computer Science and Optics
My research is centered around the physical forward and backward modeling of imaging processes and their theoretical analysis. The first line of research yields novel rendering algorithms for complex optical phenomena. The second results in 3D acquisition or other optical measurement devices.
In fact, acquisition today often results in data that is not necessarily two- or even three-dimensional. Many imaging techniques yield coded data that can only be made comprehensible to humans by computational post-processing and associated visualization techniques.
This change in acquisition paradigm has been enabled by the digitization of the complete imaging pipeline: The primary observer and analyst of data is the computer rather than the human. The design of imaging and measurement systems should take this fact into account. Co-designing optical systems and the associated algorithms can yield substantial quality benefits and novel capabilities.
My research team is working on exploring these new possibilities. Examples of our work can be found below.
We introduce three-dimensional kaleidoscopic imaging for recording multi-view imagery. We show that this approach can generate a large number of high-quality views that are well distributed over the hemisphere surrounding the object. The technique is single-shot and therefore video-capable.
3D Fluid Velocity Estimation
Captures obtained by tomographic scanning are used as inputs to our method which estimates physically plausible dense velocity fields.
Tomographic Volume Editing
Volumetric phenomena are important for realistic rendering.Volume properties are, however, not easily edited. Tomographic techniques may aid in the design of volumetric data sets for rendering purposes.
Kaleido Camera Add-On
We propose a non-permanent optical add-on that enables plenoptic imaging with standard cameras which we refer to as KaleidoCamera. We design different configuration that enable light field, HDR, hyper-spectral, and polarization imaging.
3D Fluid Capture
We have demonstrated the first time-resolved Schlieren tomography system for capturing full 3D, non-stationary gas flows on a dense volumetric grid.
Plenoptic Multiplexing Theory
Multiplexing is a common technique for encoding high-dimensional image data into a single, two-dimensional image. We have developed a general theory of multiplexing the dimensions of the plenoptic function onto an image sensor.