Fig 1.
The flowchart of the proposed method.
Fig 2.
Superpixel image segmentation.
(a) original image. (b) result of superpixel segmentation.
Fig 3.
Gradients mapping and curvature mapping.
(a)Original image. (b) Gradients mapping. (c) Curvature mapping.
Fig 4.
Results of structure learning.
(a) Result of superpixel segmentation. (b) Graph structure based on adjacent superpixel nodes. (c) Structure learning result based on the proposed method where λ = 6.
Fig 5.
(a) Original image. (b) Otsu quantitatively image. (c)-(f) Four thresholded binary images obtained from (b). (g)-(j) Three features (area, mean intensity, and fractal dimension) from (c)-(f) respectively.
Table 1.
Parametric description.
Fig 6.
Results of HGG image segmentation in BRATS data set.
(a) Original image. (b) Ground truth. (c) Superpixel segmentation. (d)-(f) Three demonstrations of segmentation results from the proposed method.
Fig 7.
Results of LGG image segmentation in BRATS data set.
(a) Original image. (b) Ground truth. (c) Superpixel segmentation. (d)-(f) Three demonstrations of segmentation results from the proposed method.
Fig 8.
Results of HGG image segmentation in HNPPH data set.
(a)&(d) Original image. (b)&(e) Ground truth. (c)&(f) Segmentation results from the proposed method.
Fig 9.
Results of LGG image segmentation in HNPPH data set.
(a)&(d) Original image. (b)&(e) Ground truth. (c)&(f) Segmentation results from the proposed method.
Table 2.
The quantitative evaluation of the proposed approach on data sets.
Fig 10.
The comparison of our approach with three related works for segmentation in BRATS data set.
The first column: the original images. The second column: ground truth. The third column: the proposed method. The fourth column: the method of Thiruvenkadam et al. The fifth column: the method of Zhao et al. The sixth column: the method of Gu et al.
Fig 11.
The comparison of our approach with three related works for segmentation in HNPPH data set.
The first column: the original images. The second column: ground truth. The third column: the proposed method. The fourth column: the method of Thiruvenkadam et al. The fifth column: the method of Zhao et al. The sixth column: the method of Gu et al.
Table 3.
The quantitative comparison of our approach with three related works for segmentation on the two data sets.