F-score test distributions. The parameters of each tool were optimised for F-score on each of the ten images, and evaluated on the remaining nine images. Boxes extend from first (Q1) to third (Q3) quartiles of the results, with median values marked; whiskers extend to the furthest F-score less than 1.5 * (Q3 − Q1) from each end of the box. (a) F-scores grouped by image evaluated. (b) F-scores grouped by image used to optimise parameters.