Presentation is loading. Please wait.

Presentation is loading. Please wait.

Boundary Extraction in Natural Images Using Ultrametric Contour Maps Pablo Arbeláez Université Paris Dauphine Presented by Derek Hoiem.

Similar presentations


Presentation on theme: "Boundary Extraction in Natural Images Using Ultrametric Contour Maps Pablo Arbeláez Université Paris Dauphine Presented by Derek Hoiem."— Presentation transcript:

1 Boundary Extraction in Natural Images Using Ultrametric Contour Maps Pablo Arbeláez Université Paris Dauphine Presented by Derek Hoiem

2 What is segmentation?

3 Segmentation is a result

4 What is segmentation? Segmentation is a result Segmentation is a process Woman Face

5 What is segmentation? Segmentation is a result Segmentation is a process Segmentation is a guide

6 Segmentation as a Guide Multiple Segmentations

7 Segmentation as a Guide Multiple Segmentations Hierarchy of Segmentations

8 Key Concepts/Contributions Hierarchical segmentation by iterative merging Ultrametric dissimilarities Thorough evaluation on BSDS

9 Hierarchical Segmentation λ 3 Region ImageDendrogram Contour Image

10 Ultrametric Contour Map Ultrametric –Definition: D(x,y) <= max{ D(x,z), D(z,y) } The union R 12 of two regions R 1 and R 2 must have >= distance to adjacent region R 3 than either R 1 or R 2 λ

11 Ultrametric Contour Map

12 Region Dissimilarity 1.D c (R 1, R 2 ): mean boundary contrast –contrast(x) = max L*a*b* diff within radius of x 2.D g (R 1, R 2 ): mean boundary gradient –gradient(x) = Pb(x) 3.D a (R 1 ): Area + α 3 Scatter (in color space) D(R 1, R 2 ) = [D c (R 1, R 2 ) + α 1 D g (R 1, R 2 )] · min{ D a (R 1 ), D a (R 2 ) } α2α2 Learned Parameters: x i = 4.5 α 1 = 5 α 2 = 0.2 α 3 = 0

13 Examples Contrast Contrast + Gradient Contrast + Gradient + Region

14 Algorithm Summary Create Initial Contours: –Extrema in gray channel form regions –Assign pixels to regions based on above ultrametric Iteratively merge regions –Keep adjacency/distance matrix

15 Comparison Martin et al. (Pb) Canny edge detector Hierarchical watersheds (using MFM for gradient) [Najman and Schmitt 1996] Variational (global energy minimization)

16 Pb No Boundary Boundary [Martin Fowlkes Malik 2004] Oriented Edges Brightness Gradient Color Gradient Texture Gradient

17 Pb

18 Variational Method [Koepfler Lopez Morel 1994] Originally Wavelet-based Textons

19 Comparison MFM: Martin et al. (Pb) Canny: Canny edge detector WS: Hierarchical watersheds (using MFM for gradient) [Najman and Schmitt 1996] MS: Variational (global energy minimization) Edge-BasedRegion-Based

20 Comparison

21 Results

22

23 Best Results http://www.ceremade.dauphine.fr/~arbelaez/results-UCM/main.html

24 Best Results http://www.ceremade.dauphine.fr/~arbelaez/results-UCM/main.html

25 Best Results http://www.ceremade.dauphine.fr/~arbelaez/results-UCM/main.html

26 Best Results http://www.ceremade.dauphine.fr/~arbelaez/results-UCM/main.html

27 Median Results

28

29

30

31 Worst Results

32

33

34

35 Hierarchies vs. Multiple Segmentations

36 Revising Segmentation


Download ppt "Boundary Extraction in Natural Images Using Ultrametric Contour Maps Pablo Arbeláez Université Paris Dauphine Presented by Derek Hoiem."

Similar presentations


Ads by Google