Figure 9.

Results comparison of Precision, Recall, and F1-score for using adaptive thresholds. Zhu [19] applied the graph cut method instead of simplified thresholding. LTP [22] algorithm is obtained by applying a lower threshold value, i.e., ${Td}_{seg}=0.35$. Our proposed scheme can obtain comparative result (precision = 0.9894, recall = 1, F1-measure = 0.9990). The other comparative methods used are stated as LBP [20], Shi15 [14], Zeng [23], Shi14 [16], Su [17], Basar21 [21], Zhuo [18].