Light-field photography extends conventional digital photography by multi-perspective information that enables 3D processing and viewing (including synthetic refocusing, 3D depth reconstruction, large depth-of-field photography, auto-stereoscopic viewing, etc.). With increasing resolution of imaging sensors, light-field photography is now becoming increasingly practical, and first light-field cameras are already commercially available (e.g., Lytro, Raytrix, and others).

Applying common digital image processing techniques (like panorama imaging) to light-fields, however, is in many cases not straight forward. The reason for this is, that the outcome must not only be spatially consistent, but also directionally consistent. Otherwise, refocussing and perspective changes will cause strong image artifacts.

Panorama imaging techniques already became an integral part of digital photography. We recently realized a first approach towards the construction of panorama light-fields (see the video).

The goal of this project is to develop new techniques for panorama light-field imaging that allow to create light-field panoramas as simple and robust as conventional image panoramas. The video on the right shows an already existing light-field panorama stitching technique that we want to improve and generalize.

The combination of advanced software algorithms and optics opens up new possibilities for display, imaging, and lighting. It makes possible responsive optical systems that adapt to particular situations automatically and dynamically. Visual computing is a relatively young research field that provides a foundation for many of these approaches. It represents a tight coupling between image synthesis, image analysis, and visual perception. While optics is all about image formation, visual computing deals with the general processing of images.

We have built a microscope (see image on the right) that captures high-resolution images and low-resolution light-fields of a specimen simultaneously. In this project, we want to investigate the possibility of combining both (the high-resolution image and the low-resolution light field) recordings to enhance imaging quality.

Light-field photography extends conventional digital photography by multi-perspective information that enables 3D processing and viewing (including synthetic refocusing, 3D depth reconstruction, large depth-of-field photography, auto-stereoscopic viewing, etc.). With increasing resolution of imaging sensors, light-field photography is now becoming increasingly practical, and first light-field cameras are already commercially available (e.g., Lytro, Raytrix, and others).

The Raytrix light-field camera is based on the plenoptic 2.0 principle. The main camera lens generates an intermediate image. A microlens array acts as camera array that focuses the intermediate image onto the image sensor. Micro lenses with different focal length allow for an extended depth-of-field and a higher effective resolution than other cameras.

In this project we want to experiment with a new Raytrix R29 camera. The goal is to find and document the optimal calibration procedure and settings given a desired depth range and to capture representative light fields. Furthermore tools for converting the Raytrix plenoptic 2.0 to an plenoptic 1.0 representation should be extended and optimized to work with this new camera.

Light-field photography extends conventional digital photography by multi-perspective information that enables 3D processing and viewing (including synthetic refocusing, 3D depth reconstruction, large depth-of-field photography, auto-stereoscopic viewing, etc.). With increasing resolution of imaging sensors, light-field photography is now becoming increasingly practical, and first light-field cameras are already commercially available (e.g., Lytro, Raytrix, and others).

Applying common digital image processing techniques (like panorama imaging, retargeting, feature extraction, etc.) to light-fields, however, is in many cases not straight forward. The reason for this is, that the outcome must not only be spatially consistent, but also directionally consistent. Otherwise, refocusing and perspective changes will cause strong image artifacts.

The goal of this project is to develop new and to convert existing light-field processing techniques and merge them into a new, flexible, extendable and easy to use toolbox for light-fields. The initial work of establishing a framework for light-field processing is the ideal starting point to get into the topic of light-fields. The gained knowledge can then be used to work on new processing techniques (e.g. depth reconstruction, segmentation, stitching, relighting, simple color conversions, etc.) at a difficulty level suitable for the chosen project type (practicum or thesis).

At our institute we are developing a novel imaging device that is based on a thin-film luminescent concentrator. This special transparent foil transports light that falls on the surface to its edges, where line scan cameras measure the amount of transported light. From the measured data we reconstruct an image that was focused on the surface of the luminescent concentrator foil. This can either be done by solving a system of linear equations or by using the filtered backprojection technique that is knows from CT scanners.

The goal of this project is to improve the image reconstruction procedures that are currently in use, and to further investigate the possibilities of filtered backprojection. For example, the solving of a linear equation system can be improved by finding appropriate preconditioners. The use of filtered backprojection is necessary to allow the reconstruction of images with a higher resolution. However, the filters known from CT image reconstruction cannot be directly used for the transpartent image sensor.

Nowadays, many different science fields, such as biology or medicine, are faced with large amounts of complex data. In order to solve current problems in these fields, it is essential to include multiple heterogeneous datasets in the sense-making process. Gaining new insights requires users to identify patterns and trends in the data.

In recent work we have developed a "navigation system for data analysis" that is able to pinpoint an analyst to possibly interesting patterns and correlations in the data. The system on-the-fly reduces the vast options space and interactively present a ranked list of potentially interesting data combinations that the user then can investigate.

The goal of this thesis is to even go a step further and use the ranked list of potentially interesting patterns for giving the user an automated tour through the highlights of his data. This process is comparable to a city trip of a tourist who books a personal tour guide to discover unknown territory.

Large scale initiatives, like the The Cancer Genome Atlas Project, collect vast amounts of heterogeneous raw data (matrices, lists, meta-data, etc.). In order to gain new insights, the data is processed on a regular basis in large and complex pipelines, e.g., the Firehose Pipeline – resulting in derived data such as clustering information, reports, images, ranked lists, etc.

The problem is that the pipeline as well as the input data are constantly changing over time. New raw data items are collected, processing parameters optimized and the processing workflow manipulated. However, the changes are not coordinated in a structured way, leading to unexpected changes in the output as well as errors in the pipeline.

The goal of this project is to develop a visualization technique that allows researchers to analyze such complex and large pipelines with all its parts on multiple levels over time. Seeing the evolution of the pipeline will help them to find out the causes of changes in the output and also to identify and correct errors in the pipeline.

Over the last few decades many fields of science have been confronted with tremendous amounts of data and continuously increasing annual growth rates. The sheer amount of data, however, is only one aspect of the problem. In order to solve complex analysis questions, it is necessary to cope with heterogeneous data sets from different sources, on distinct levels of scale and stored in various formats and types such as text, maps, graphs and images. Tailored visual analysis tools, such as the one demonstrated in the video, are well suited for solving a specific, predefined analysis task. However, it is highly unrealistic that it will be possible to create a super-application that covers all kinds of data and analysis tasks from different domains. Thus, users want to use a combination of existing applications to perform their analysis. However, the single tools are usually not integrated with each other.

The goal of this thesis is to develop a meta-visualization that is able to run an analysis within such a multi-application scenario. In order to realize a visual integration of the single tools, it will be essential to find ways for navigating, arranging, extracting, and adapting the individual application's interfaces.