Presentation is loading. Please wait.

Presentation is loading. Please wait.

Tooth Segmentation on Dental Meshes Using Morphologic Skeleton Kan WU School of Software Tsinghua University, P. R. of China M.Eng. Li CHEN Ph.D. CAD/Graphics.

Similar presentations


Presentation on theme: "Tooth Segmentation on Dental Meshes Using Morphologic Skeleton Kan WU School of Software Tsinghua University, P. R. of China M.Eng. Li CHEN Ph.D. CAD/Graphics."— Presentation transcript:

1 Tooth Segmentation on Dental Meshes Using Morphologic Skeleton Kan WU School of Software Tsinghua University, P. R. of China M.Eng. Li CHEN Ph.D. CAD/Graphics 2013, Hong Kong Jing LI Department of Orthodontics Peking University School and Hospital of Stomatology, P. R. of China Ph.D. Yanheng ZHOU Ph.D. Computers & Graphics

2 Background our work

3 Contribution significantly reduced user interaction avoid complex mesh feature estimation an applicable pipeline for dental mesh segmentation experiments on various clinical cases of different tooth shapes and various levels of crowding problems

4 Problems of Current Work intensive interaction affected by feature disturbance not sufficiently accurate Kumar et al Kronfeld et al Shape

5 A Good Dental Segmentation Approach Should locate teeth area automatically separate adjacent teeth automatically less dependent on complex feature estimation smoothed & fitted boundary morphologic skeleton morphologic skeleton

6 Why Morphologic Skeleton insensitive to feature missing & disturbance simplified approximation of mesh features easy separation of adjacent objects ACCURACY EFFICIENCY REDUCED INTERACTION

7 Dental Mesh Segmentation Pipeline

8 1 st Step: Locating Teeth Parts automatic plane cutting region-growing skeletonization original mesh

9 PCA-based plane initialization energy field 1 st Step: Locating Teeth Parts – (1)Estimating Cutting Plane

10 (1)Estimating Cutting Plane – PCA-based Plane Initialization Kronfeld et al., 2010 set of feature vertices barycentric point eigenvector corresponding to the smallest eigenvalue

11 weighted distance feature points connected to v (1)Estimating Cutting Plane – Energy Field

12 skeleton 1 st Step: Locating Teeth Parts – (2)Morphologic Skeletonization curvature threshholding connectivity filtering morphologic operation skeletonization

13 original morphologic skeleton (Rossl et al., 2000) improved morphologic skeleton 1 st Step: Locating Teeth Parts – (2)Morphologic Skeletonization

14 seed points skeleton 1 st Step: Locating Teeth Parts – (3)Region-Growing

15 cut 2 nd Step: Separating Teeth

16 valid cut discarded cut 2 nd Step: Separating Teeth – Various Scenarios

17 2 nd Step: Separating Teeth – Results

18 3D contours 2D contours sampled 2D contourssampled 3D contours interpolated 3D contours 3 rd Step: Smoothing Tooth Contours

19 Direction Change Measure Length Change Measure center point for contour middle point 3 rd Step: Smoothing Tooth Contours – 2D Sampling

20 Direction Change Measures Length Measures sign(x) = 1 if x > 0, otherwise -1 3 rd Step: Smoothing Tooth Contours – 2D Sampling

21 Results – Mild Tooth Crowding original model cutting plane estimation skeletonization & region-growing separating & contour smoothing

22 Results – Moderate Tooth Crowding original model cutting plane estimation skeletonization & region-growing separating & contour smoothing

23 Results – Severe Tooth Crowding original model cutting plane estimation skeletonization & region-growing separating & contour smoothing

24 Results

25

26

27 Kumar et al Kronfeld et al our approach Comparative Results – Published Approaches

28 when user interaction is not sufficiently accurate enough 3Shape Software our approach Comparative Results – 3Shape Software

29 Accuracy Evaluation – Mean Errors The mean errors that compare our results to manually labeled ground truth. The unit is mm.

30 the distribution of particular error values across all segmented boundary vertices. The blue, yellow, red lines indicate the ranges of [0, 0.25], [0.25, 0.5], [0.5, 1.5], respectively. Accuracy Evaluation – Error Distribution

31 User Interaction Evaluation Time consumed by user interactions. The blue and yellow lines indicate manual boundary completion and additional seed adding, respectively. The unit is s

32 Limitations a dental mesh benchmark GPU accelerating completely eliminate user interaction Future Work user interaction still needed no GPU accelerating

33 Demo

34 Kan WU THANK YOU Li CHEN Jing LI Yanheng ZHOU


Download ppt "Tooth Segmentation on Dental Meshes Using Morphologic Skeleton Kan WU School of Software Tsinghua University, P. R. of China M.Eng. Li CHEN Ph.D. CAD/Graphics."

Similar presentations


Ads by Google