Our department has a strong research program in the areas of image processing and image-based pattern recognition. Many results and algorithms we developed are generic in their nature and find their use in numerous application areas. This includes digital photography, surveillance systems, image forensics, biomedicine, remote sensing, astronomy and art conservation.
On this page, we summarize the most important problems we work or worked on. Each area contains a short text explaining in brief results we achieved including links to related journal papers.
Image registration is a process of matching and overlaying two or more images of the same scene. It is one of the most important image processing operations in medical imaging, remote sensing, surveillance, robot vision, quality inspection, and in many others. Image registration is a necessary step when performing image fusion, detecting changes, and importing the images into information systems.
The main publications can be found here
During last fifteen years we contributed significantly to the theory of moments invariants. This includes derivation of complete systems of invariants with respect to affine transformation, with respect to blurring by kernels with various types of symmetries and also invariants to combined degradations. These results are explained in more detail under the following links.
Our experience with moments and moment invariants gained from years of research in this area resulted in a recent book
covering the current state-of-the-art and presenting the latest developments in this field. Read more here.
In recent years, we work on algorithms removing a wide variety of degradations that are common in digital imaging.
We consider mainly noise, blurring and insufficient resolution.
The main research topics are
We have been involved in several cultural heritage applications. Nowadays, art restorers and conservators use various visual sensors for better analysis of old artworks and modern image processing methods can facilitate their work. The following list presents areas where our team was involved
We are active in developing mathematical and computational algorithms capable of detecting the traces of tampering in digital images. Below are short descriptions of some of our past and current work in this field.
Manual processing of microscopy images is very tedious and prone to errors, therefore we develop a new automated methods. Main tasks in processing are segmentation of cells from the background, segmentation of individual cells and their tracking over time.
We have been involved in many projects with other universities, research institutes and commercial companies. Here are several examples of the results of such cooperation.