For the analysis of learning processes and the underlying changes of the shape of excitatory synapses (spines), 3-D volume samples of selected dendritic segments are scanned by a confocal laser scanning microscope. The images are unsharp because of the (direction dependent) resolution limit. A deconvolution on image data is not sufficient for the resolution needed. Therefore a parametric model is used to reconstruct the dendrite and the spines. The parameter estimation of model is done in a two step approach. First, rough center axes of dendrite and spines are found by a growing model which can be adjusted interactively. In a second step the model parameters are optimized during an iterative process. To estimate the deviation between the microscope image and the model, the model is sampled with the same resolution as the microscope image and convolved by the microscope point spread function (psf). The result is a accurate model of dendrite and spines. The model fitting process is comparable with a deconvolution but with a limited number of model parameters and stable results without strong distortions by PSF. The associated presynaptic terminal can be detected in a second image channel inside a region of interest (roi) on spine position. Morphological features of spines from geometrical model and from second channel roi are combined for statistical analysis.