Multidimensionale Segmentierung in Bildfolgen und Quantifizierung dynamischer Prozesse

In this interdisciplinary work I developed digital image processing methods for quantitative analysis of dynamics. The applications focused on biology and medicine. Confocal microscopes can resolve biological structures by the use of fluorescent markers. Due to a low signal to noise ratio the processing of noise reduction techniques was an important task. I developed a segmentation method in 2D and 3D based on deformable models. In 2D, I was able to show that the attraction range of the active B-spline contour could be increased in combination with a special external field. This improvement to the classical parametric active contour is especially important when the initialization of the curve is far beyond the object. In 3D, I introduced a method for simultaneously segmenting multiple objects in one image and adapted this approach to the special case of cell division. A cluster algorithm was applied to assign the extracted edge points to the different objects. With these methods, I was able to quantify the volume and area expansion of the membrane over time. To track multiple segmented objects over time, I have developed a particle-tracking algorithm, based on a Fuzzy decision kernel. Several applications show the benefit of the developed methods.