Figure 3

Performance criteria and segmentation accuracy of four segmentation methods on synthetic images and ground truth image datasets. (A–D) True positive (TP) rate and Precision on four kinds of noisy synthetic images using three segmentation methods and the wavelet method. Error bars indicate standard deviation. (A) Samples of synthetic images (N = 20) Gaussian white noise and TP rate and Precision when applying point-wise method, Otsu method, DS method and wavelet-based segmentation method. (B) Samples of salt and pepper noise synthetic images (N = 20) and TP rate and Precision for four segmentation methods. (C) Intensity attenuation noisy synthetic image samples (N = 20) and TP rate and Precision for four segmentation methods. (D) Samples of overlapping synthetic image samples (N = 20) and TP rate and Precision for four segmentation methods. (E) Segmentation accuracy when applying Otsu method and wavelet-based segmentation method on three Fluo-N2DH-GOWT1 sample images. (F) True positive rate and Precision on five ground truth images when using point-wise method, Otsu method, DS method and wavelet-based segmentation method.