Computer Vision and more generally Perceptual Computing have great potentials to change the way we interact with computers and how machines such as robots perceive the world. Over the last three decades significant progress has been made in computer vision. Robustness of perception and vision algorithms however is a notorious problem and one of the major bottlenecks for industrial applications. At the same time there is little doubt that in the next decades small and inexpensive sensors will be developed and embedded in many devices. Therefore, computing may have access to and can be enhanced by ubiquitous sensors and sensing. My hypothesis is that the integration of multiple features and sensors facilitates robustness in environments of realistic complexity.

The talk will touch upon three areas of research, which are motivated by fundamental issues of computer vision and perceptual computing. The first area is concerned with the robustness of perception where I will discuss an approach for the robust recognition of objects through computer vision techniques. The second area deals with performance characterization and optimization of content-based image retrieval. The results highlight the importance and the potential to look at image retrieval from a vision system's oriented perspective. The third and last area briefly summarizes research to use distributed sensing devices to assist a human proactively during furniture assembly.