Computer Vision

Our research interests cover all aspects of computer vision. Computer vision generally interprets images which represent a 3D environment. Images can be captured with optical systems (i.e. video camera) or special sensors such as range sensors or medical imaging systems which capture the internal structure of 3D bodies (i.e. CAT, NMR). The results of interpretation depend on the goal or the role of computer vision in the overall system (i.e. identification of a person on a photography or visual navigation of a mobile robot). The interpretation results can be the segmentation of an image into semantic units, a geometric description of the scene, object tracking parameters, identification of classification of objects on the image etc. The main reasons why 3D interpretation of images is difficult are: - when a 3D scene is projected on a 2D image plane, - segmentation of an image into meaningful parts depends on the recognition of those parts, - understanding of human and biological vision is also not complete Goals of our research program are: - identification of models and algorithms which are beside the image information crucial for image interpretation - development of computer vision systems for solving specific problems (i.e. reconstruction of CAD models from images, content-based search of image databases, object recognition).