The application of the 3D mean-shift filter to 3D Computed Tomography Data enables the segmentation of internal traits. Specifically in maize seeds this approach gives the opportunity to separate the internal structure, for example the volume of the embryo, the cavities and the low and high dense parts of the starch body. To evaluate the mean-shift filter, the results were compared to the usage of a median-smoothing filter. To show the relevance of the mean-shift extended image pipeline an automatic assessment of biological relevant samples was conducted. As data sets maize seeds of 16 different genotypes were used to segment the three different parts of the seed, embryo and the structure due to density within the starch body.