Figure 3: Using the original image (a), a DSA image (b) is computed. After median filtering (c), the remaining motion artifacts from the catheter have vanished. Only the boundary at the diaphragm remained. Afterwards, all pixels are weighted by a sigmoid function (d) to get a more homogeneous value distribution inside the contrasted area. Edges are extracted using derivatives of Gaussian with a large kernel size (e). Finally, the similarity to the rendered edges (f) is evaluated.