Shape Recognition Based on Projected Edges and Global Statistical Features

Figure 3

The distributions of the vertical and horizontal moments on the Hungarian Forint Banknote pattern dataset (for details, see Section 8.). Classes 1–9 represent different patterns from banknotes; other irrelevant shapes are denoted as class 0. The figure shows that these values do not contain enough discriminative power to classify the patches but provide a good guide to filter and reject obviously different shapes.