Uppsala University

Computerized image analysis


Computerized image analysis is concerned with the extraction of information from images. A toolbox based on knowledge from mathematics, signal processing, computer science and physics is used for this. Applications are found in medicine and many other sectors of society.

Human vision has a remarkable capability to rapidly interpret and recognize objects in images. Computers on the other hand are very good at counting and measuring. In computerized image analysis we are trying to give computers the capability of pattern recognition and to combine that with measurements and calculations. Image analysis is today a mature science capable of creating solutions to many problems of practical interest.

Methodologically image analysis is truly multidisciplinary. It can be seen as applied mathematics and as multidimensional signal processing. Images are big data matrices and discreet geometry and topology are important theoretical tools to correctly interpret these. The methods are implemented as computer programs so computer science is important. Knowledge about image properties from physics and optics is often essential. The solution of many practical problems require application specific expertise so many projects are carried out in close collaboration with other experts.

Computer graphics and visualization are the inverse of image analysis; images are created rather than analysed. But visualization is often needed when working with images, particularly when the images show the 3D interior of objects such as is the case with medical tomographic images. Stereoscopic display techniques combined with haptic interaction provide powerful tools for exploration of 3D images.

At the Centre for Image Analysis we are mixing basic research about algorithms for image analysis and visualization with application oriented methods development particularly in the fields of medicine and environmental analysis. We are particularly interested in volume images, multispectral images and image sequences, i.e. images with more than two dimensions.

Examples of current projects are the recognition of viruses in electron microscopic images, the study of the fibre structure in paper and composite materials, analysis of the integration between implants and bone tissue using microtomographic images, analysis of the 3D chromatine structure in cell nuclei on PAP-smears for early detection of cervical cancer, detection and quantiative analysis of signals from individual molecules in fluorescence microscopy images of cells, interactive analysis of the uptake and washout of contrast media in magnetic resonance angiographic images of breasts.


Read more »

Contact:
Ewert Bengtsson
Gunilla Borgefors
Ingela Nyström
Stefan Seipel
Ingrid Carlbom


Departments:
Centre for Image Analysis


Partners:
Department of Information Technology
Department of Mathematics
Department of Genetics and Pathology


Keywords:
Image analysis, image processing, computer graphics, visualization.


The following undergraduate programs at Uppsala University are related to this research:
» Master of Science programme in Engineering Physics
» Master of Science programme in Information Technology Engineering
» Master of Science programme in Bioinformatics Engineering
» Master of Science programme in Molecular Biotechnology Engineering
» Master programme in Mathematics
» Master programme in Computer Science
» Master programme in Applied Biotechnology
» Master programme in Human-Computer Interaction
» Master programme in Computational Science

  Two examples of biomedical image analysis. At the top interactive haptical control of liver segmentation in a CT volume based on a deformable simplex mesh model. At the bottom stained (red and green) biomolecules can be automatically quantified per cell after segmentation.
Two examples of biomedical image analysis. At the top interactive haptical control of liver segmentation in a CT volume based on a deformable simplex mesh model. At the bottom stained (red and green) biomolecules can be automatically quantified per cell after segmentation.

Search research