Figure 15

(a–d) Show the cloud plots of AUC for every image in dataset A–D respectively. Each point in the plot corresponds to a detection performance in terms of AUC of an image for a certain technique. The paired t-test p-values between each technique are shown in Table 6. It is observed that in dataset A and D, which contains noisy images, the MPAV technique is significant better than M^SSAE, M^SPV, and M^MPV techniques. In dataset B and C, there is no significant different between the M^MPAV and the M^SSAE.