Texture based Automatisation of grayvalue image Segmentation

To compare two brains with each other, one first has to map one onto the other. This is a very tedious task, because one has to manually label the neuropiles prior to calculate a corresponding transform. An automated segmentation is difficult, as certain regions can be labelled correctly only with good knowledge of the neuroanatomy of the brain, and there is a good amount of noise which cannot be technically reduced. Furthermore, the exact geometrical form of the neuropiles varies vastly between two different brains. We are working towards an automated procedure to analyse the texture - meaning statistical properties of the gray values - which allows a mapping between two different brains.