Využití obrazových segmentačních technik pro rekonstrukci 3D modelů objektů z obrazu

The main goal of the project is analysis and enhancement of present image
segmentation techniques suitable for 2D, 2.5D and 3D object models based on their
image. Particular segmentation techniques are tested on a system for augmented
multi-user video communication. The experiments show that the present 2.5D relief
reconstruction approach is not robust enough. It is due to the input image
segmentation errors. A segmentation quality is directly influencing the quality
of reconstructed user model and also communication naturalness.
Wide use of
the segmentation methods allow to re-use the results of the project also for the
medical data analysis. Segmented images from CT/MR data will be processed by
existing system for 3D models design of bones and tissue. The aim is to
automatize whole segmentation process.
The students can cooperate on project
through their years and diploma projects. We will encourage their creativity by
establishing the fundamental tool for image processing and segmentation. Further
development will go on in collaboration with students.