Title: Some links between mincuts, optimal spanning forests and watersheds
1Some links between min-cuts,optimal spanning
forests and watersheds
2Outline
- Motivation
- Usual algorithms in graph
- Min-cuts
- Watersheds
- Shortest-path spanning forests cuts (SPSF cuts)
- Comparisons of results
- Links
- Link between watersheds and SPSF cuts
- Link between min-cuts and watersheds
- Conclusion
3- 1 -Motivationgraph segmentation
4Image segmentation
1 Motivation graph segmentation
An image
5Image segmentation
1 Motivation graph segmentation
A set of markers on the image
6Image segmentation
1 Motivation graph segmentation
How to find a good segmentation?
7Energy to minimize
- Let I be an image. Let s and t be two pixels of
I. - We denote by
- the value of the pixel s
- the neighbourhood of the pixel s
- the label given to the pixel s.
- Example of energy to minimize for the
segmentation
1 Motivation graph segmentation
8What about graphs?
- a set of nodes, denoted by V,
- a set of edges linking a couple of nodes, denoted
by E.
1 Motivation graph segmentation
- Edge weighted graph
- Application
9Image to graph
- Building of the graph
- Each pixel ? node
- Pixels s and t are neighbours ? edge es,t of
weight
1 Motivation graph segmentation
10Image to graph
1 Motivation graph segmentation
Edge-weighted graph
11Markers
1 Motivation graph segmentation
12Result frontier
1 Motivation graph segmentation
Link to energyminimization of the sum of the
dashed edges
13Result partition
1 Motivation graph segmentation
Link to energymaximization of the sum of the
bold edges
14- 2 -Usual algorithmsin graph segmentation
15Vocabulary on graphs
Marker M
Extension
Spanning extension
Maximal extension
Cut
Spanning forest
1 Graph segmentation
Correspondance with image
Spanning forest relative to Mspanning extension
relative to M from which you cant remove any
edge without loosing the spanning extension
property
Maximal extension relative to Mspanning
extension relative to M which cant be strictly
contained by another extension relative to M
Cut relative to M set of edges linking two
different connected components of a spanning
extension relative to M
Marker Msubgraph of G
Extension relative to Msubgraph of G for which
each connected component contains exactly one
connected component of M
Spanning extension relative to M extension
relative to M which contains all the nodes of G
16Min-cuts
- Minimizing the cuts weight Maximizing the
maximal extensions weight - Searching the maximum flow (Ford Fulkerson,
1962) Searching the min-cut of a marker with 2
connected components - Computing complexity polynomial
- Drawback for more than two connected components
in the marker, can only give an approximated
result through successive cuts (NP-complete)
2 Usual algorithms in graph segmentationa -
Min-cuts
17Watersheds
- Origin S. Beucher C. Lantuéjoul, 1979
- Searching a watershed Searching the cut induced
by a spanning forest of minimum weight relative
to the minima of the graph - Computing complexity linear (J. Cousty al.,
2007) - Drawback doesnt take into account all the edges
of the maximal extension
2 Usual algorithms in graph segmentationb -
Watersheds
18Shortest-path spanning forest cuts (SPSF cuts)
- Let M be a subgraph of G, x be a node of G\M and
p be a path in G from x to M. - We define
- Length of a path
- Connection value of x
- Definition
- A subgraph F of G is a SPSF if and only if F is
a spanning forest and for any node x of F there
exists a path p in F from x to M such that
P(x)P(p).
2 Usual algorithms in graph segmentationc
SPSF cuts
19Shortest-path spanning forest cuts (SPSF cuts)
- Origin J.K. Udupa al., 2002 A.X. Falcao
al., 2004 R. Audigier R.A. Lotufo, 2006 - Computing complexity quasi-linear
- Drawback doesnt take into account all the edges
of the maximal extension
2 Usual algorithms in graph segmentationc
SPSF cuts
20Remark about comparison
- Min-cuts are of low value whereas watersheds or
SPSF cuts are of high value. - In the objective to compare min-cuts with
watersheds or SPSF cuts working on the same data,
we consider a strictly decreasing function
applied to the weights of the graph before
computation of a watershed or a SPSF cut.
2 Usual algorithms in graph segmentationb -
Watersheds
21Comparisons of results
2 Usual algorithms in graph segmentationd -
Comparisons of results
Min-cut
Watershed / SPSF cut
Similar results between watershed and SPSF cut,
but differences with min-cut.
22Comparisons of results
2 Usual algorithms in graph segmentationd -
Comparisons of results
Markers
Min-cut
Watershed / SPSF cut
- "Best" min-cut
- Drawback of watershed / SPSF cut leak point
23Comparisons of results
2 Usual algorithms in graph segmentationd -
Comparisons of results
Markers
Min-cut
Watershed / SPSF cut
- "Best" watershed / SPSF cut
- Drawback of min-cut as a global minimum, the cut
is minimum with a few edges of high value rather
than lots of edges of low value
24- 3 -Links
25Link between watersheds and SPSF cuts
Theorem 1 A spanning forest of minimum weight
is a SPSF.
Theorem 2 A SPSF relative to the minima of the
graph is a spanning forest of minimum weight
relative to the minima of the graph.
3 Linksa - Link between watersheds and SPSF
cuts
Consequence A SPSF cut relative to the minima
of the graph is a watershed.
(found by us and independently by R. Audigier
al., 2007)
26Link between min-cut and watershed
- We denote by Pn the application of weight risen
to power n. - Theorem
- There exists a real number m such that for any n
m, any min-cut relative to the marker M for
Pn is the cut induced by a maximum spanning
forest relative to the marker M for Pn. - Consequently, if M is the maxima of the graph,
this min-cut is a watershed.
3 Linksa - Link between watersheds and SPSF
cuts
Remark The value of n acts like a smoothing
parameter for the cut.
27Link between min-cut and watershed Rise of the
power of graph weights for min-cuts
Watershed
Min-cut
3 Linksb - Link between min-cuts and watersheds
Rise to square min cut watershed
28Link between min-cut and watershed Rise of the
power of graph weights for min-cuts
Min-cut
Marker
3 Linksb - Link between min-cuts and watersheds
Power n 1,4
Power n 1
Watershed
Power n 3
Power n 2
29Intuitively
Let Cn be the min-cut for the weight at power
n. Weight of the cut
3 Links b - Link between min-cuts and
watersheds
Looks like n-norm
When n is rising you converge towards
infinity-norm
Consequently
30Intuitively
Min-cut
Watershed
3 Links b - Link between min-cuts and
watersheds
31- 4 -Conclusion
32Conclusion
- Equivalence between watersheds and SPSF cuts
relative to minima - Link from min-cuts to watersheds
- Rise of the power for min-cuts smoothing
parameter of the cut
4 Conclusion
- Watersheds fast but isnt global (leak points)
- Min-cuts slow but global minimum (for a marker
with 2 connected components) and has smoothness
parameter - Future work link from watersheds to min-cuts?
33Thank you for your attention!Any question?
Min-cut(power p 1)
Min-cut(power p 1,4)
Min-cut(power p 2)
Watershed, SPSF cut and min-cut(power p 3)
Marker
34Scheme of proof
- F spanning forest of maximum weight (MaxSF) for
the graph risen to power n - C min-cut for the graph risen to power n
- F MaxSF for the graph complementary of C risen
to power n - We just have to prove that P(F) P(F).
- Let p1,,pk be the different values of weight in
the graph with p1gtgtpk. - Let nh(X) be the number of edges of weight ph in
the subgraph X. - So we have and
- Which comes to proving that for any h
3 Links b - Link between min-cuts and
watersheds
35Leveling (thresholding)
- Let Xh be the subgraph of X composed with only
the edges of weight greater or equal to ph.
3 Links b - Link between min-cuts and
watersheds
G3 (p3 7)
G2 (p2 8)
G6 (p6 4)
G4 (p4 6)
36Lemma 1
- For any h, Fh is a spanning forest relative to
the markers for Gh.
3 Links b - Link between min-cuts and
watersheds
G3 (p3 7)
F3 (p3 7)
G6 (p6 4)
F6 (p6 4)
37Lemma 2
- For any h, Fh is a spanning forest relative to
the markers for Gh.
3 Links b - Link between min-cuts and
watersheds
G3 (p3 7)
F3 (p3 7)
G6 (p6 4)
F6 (p6 4)
38Lemma 3 and epilogue
- Lemma 3 Any spanning forest for a subgraph X has
a constant number of edges. - We deduce that n1(F1)n1(F1).
- By induction for all h, nh(Fh)nh(Fh).
- Consequently, P(F)P(F) and finally F is a
MaxSF!
3 Links b - Link between min-cuts and
watersheds