Systems for processing, generating, and transmitting digital images (image systems) often have hard constraints on compute performance, latency, throughput, and costs. Typical examples are medical image processing, computer games, or video compression on camcorders. In order to fulfill these constraints often dedicated hardware accelerators are used, as well as graphics processors (GPUs) or digital signal processors (DSPs).
The resulting image systems are heterogeneous in two respects: First, within a system the computation is spread to several components, second, there is a large heterogeneous set of architectures on which different image applications are executed. Both types of heterogeneity lead to engrossing and important research problems, which are examined in the graduate school. Three major topics will be considered: dedicated hardware architectures for heterogeneous image systems, tools and methods for the programming of heterogeneous image systems, and applications and algorithms for heterogeneous image systems. Image systems are of high importance, both in research and in practice. Their planning, development and realization requires comprehensive and interdisciplinary knowledge in soft- and hardware, method and tool design, and algorithmic development.
Today, a variety of dedicated image hardware is available (e.g. graphics cards, digital signal processors, video compression chips, etc.), so that our work can start using existing hardware. In the course of the research school, we also explore and develop novel dedicated hardware. The Institute for Electronics Engineering covers the development of dedicated hardware to capture image and video data as well as the direct analog processing of image data on the photo sensor IC with focus on power and area savings at processing times comparable to digital implementations.