K.U.Leuven > ESAT > PSI > Visics > Research

Research

Research in the industrial image processing group (VISICS) is structured in several research teams. These teams specialise on certain topics within the realm of computer vision. The team structure is dynamic and depending on needs arising or fading out, new teams will be formed or old ones dissolved. At this moment, we have teams concentrating on the following subjects:

Reconstruction and recognition of shapes

This research includes the segmentation and recognition of 2D and 3D objects in an image (sequence) and the 3D reconstruction of scenes from one or more images. The analysed scene is also used to provide visual guidance to autonomous guided vehicles and space robots.

Multimedia and image communication

Different aspects of storing and retrieving multimedia data are covered: colour (filtering and representation), image/video compression (including MPEG activities), and virtual reality aspects (e.g., talking head animation).

Surveillance

This includes improving the visual quality of video and the tracking of objects (e.g., persons) in a scene, as well as the development of interactive tools for video-based metrology. Forensic applications are also covered.

Remote-sensing

Research on aerial and satellite images (VIS, IR and SAR) includes postprocessing (e.g., noise removal), multispectral analysis, classification (detection of crops, forests, coast lines,...) and detailed 3D terrain reconstruction.

Visual inspection

A number of versatile software packages for image processing have been developed for several platforms (workstations as well as PCs). Several visual inspection solutions are or have been developed in production (e.g., IC package inspection, textile inspection,...) and agriculture (e.g., weed selective spraying, pig carcass inspection,...). Filtering techniques to improve the quality for inspection are also implemented.

Image processing hardware

The hardware relates to optics for image input (lenses, colour filters,...), VLSI integration (design of chips for edge detection, non-linear diffusion and MPEG coding), and the implementation of an autonomous guided vehicle.

Content-Based Image Retrieval and Image Understanding

The goal of this research is the development of intelligent systems for image-understanding. In our view, visual intelligence is linked to the ability of the system to reach beyond mere pixel-based characteristics and autonomously access aspects of the image content. The range of applications of automatic image-interpretation and -understanding encompasses both industrial projects such as fabric inspection, as well as more generic methodologies, such as computer-generated image annotations.