and variational formulations also exp osed two basic questions (1) What is the criterion
that one wants to optimize? and (2) Is there an ecient algorithm for carrying out the
optimization? Many an attractive criterion has b een doomed by the inability to nd
an eective algorithm to nd its minimum{greedy or gradient descenttype approaches
fail to nd global optima for these high dimensional, nonlinear problems.
Our approach is most related to the graph theoretic formulation of grouping. The
set of points in an arbitrary feature space are represented as a weighted undirected
graph
G
=(
V
;
E
), where the no des of the graph are the points in the feature space,
and an edge is formed b etween every pair of no des. The weight on each edge,
w
(
i
;
j
),
is a function of the similaritybetween no des
i
and
j
.
In grouping, we seek to partition the set of vertices into disjoint sets
V
1
;
V
2
; :::;
V
m
,
where by some measure the similarity among the vertices in a set
V
i
is high and across
dierent sets
V
i
,
V
j
is low.
To partition a graph, we need to also ask the following questions:
1. What is the precise criterion for a goo d partition?
2. How can such a partition be computed eciently?
In the image segmentation and data clustering community, there has been much
previous work using variations of the minimal spanning tree or limited neighborhood
set approaches. Although those use ecient computational metho ds, the segmentation
criteria used in most of them are based on local properties of the graph. Because
perceptual grouping is ab out extracting the global impressions of a scene, as wesaw
earlier, this partitioning criterion often falls short of this main goal.
In this paper we propose a new graph-theoretic criterion for measuring the go odness
of an image partition{ the
normalized cut
. Weintroduce and justify this criterion
in section 2. The minimization of this criterion can b e formulated as a generalized
eigenvalue problem; the eigenvectors of this problem can b e used to construct go od
partitions of the image and the pro cess can b e continued recursively as desired(section
3). In section 4 we show experimental results. The formulation and minimization of the
normalized cut criterion draws on a b ody of results, theoretical and practical, from the
numerical analysis and theoretical computer science communities{section 5 discusses
previous work on the spectral partitioning problem. We conclude in section 6.
2 Grouping as graph partitioning
A graph
G
=(
V
;
E
) can be partitioned into two disjoint sets,
A; B
,
A
[
B
=
V
,
A
\
B
=
;
,by simply removing edges connecting the two parts. The degree of dissimilarity
between these two pieces can b e computed as total weight of the edges that have been
removed. In graph theoretic language, it is called the
cut
:
cut
(
A; B
)=
X
u
2
A;v
2
B
w
(
u; v
)
:
(1)
The optimal bi-partitioning of a graph is the one that minimizes this
cut
value. Al-
though there are exp onential number of such partitions, nding the
minimum cut
of a
graph is a well studied problem, and there exist ecient algorithms for solving it.
Wu and Leahy[21] proposed a clustering metho d based on this minimum cut cri-
terion. In particular, they seek to partition a graph into k-subgraphs, such that the
3