Sunteți pe pagina 1din 16

Image Segmentation

5/7/2016

Fingerprint Image Segmentation

Region of interest (ROI) segmentation is performed to extract fingerprint region and


remove background region and other irrelevant information from the fingerprint image.
This stage is also referred to as foreground/background detection. ROI extraction is
necessary, so that only valid features are extracted during feature extraction stage. In ROI
segmentation, the non-overlapping block-wise approaches are generally used for fast
computation and easy decision making.

In one approach, variance of local intensity is first calculated and then compared against a
threshold defined by the global intensity variance of the image. If the local variance is
less than the global threshold, then block is classified as a background region, otherwise,
a foreground region. The gray level variance for a block of size TxT is defined as:
1
V k 2
T

T 1 T 1

Gi, j M i, j ,
2

i 0 j 0

where V(k) is the variance for block k, G(i, j) is the gray level value at pixel (i, j) and M(i, j)
is the mean gray level value for the block k.

5/7/2016

Fingerprint Image Segmentation

Figure 2.1: Images segmented based on intensity variance based method

5/7/2016

Region Segmentation
Find sets of pixels, R 1 , R 2 , , R n such that

Ri I
i
i j, Ri R j
All pixels in region i satisfy some constraint of
similarity.

5/7/2016

Segmentation and Grouping


Purpose:
Obtain a compact representation from an
image/motion sequence/set of tokens
Should support the type of application and the
criteria for grouping is based on the specific
application

5/7/2016

Segmentation and Grouping


Major approaches
Forming image segments by grouping (or clustering)
collect together tokens that belong together
Fitting a model or a fundamental matrix
Fitting line to collection of points and associating lines
associate a model with tokens
which model?
which token goes to which element?
Tokens: whatever we need to group (pixels, points,
surface elements, etc., etc.)
5/7/2016

Segmentation by clustering
Partitioning: top down segmentation
Decomposing large sets into smaller ones according
to their eligibility to selected model/criterion
Decompose an image into regions that have
coherent color or texture
Decompose a video sequence into segments of
video showing about the same stuff (static area)
from same view point
Tokens belong together since they lie on same object
Grouping: bottom up segmentation
Tokens belong together since they are locally coherent
5/7/2016

Image Segmentation:
Segmentation by Clustering

5/7/2016

Segmentation by clustering
Main idea: Cluster together (pixels, tokens, etc.)
that belong together
Simple Methods
Divisive clustering
split cluster along best boundary
Repeat
Agglomerative clustering
attach token to cluster it is closest to
repeat
5/7/2016

Segmentation by clustering

Dendrograms

A representation of structure of hierarchy of clusters

Dendrogram

A data set

Given a set of N items to be clustered:


1). Start by assigning each item to its own cluster, so that if you have N items, you now have N
clusters, each containing just one item.
2). Find the closest (most similar) pair of clusters and merge them into a single cluster, so that
now you have one less cluster.
3). Compute distances (similarities) between the new cluster and each of the old clusters.
4). Repeat steps 2 and 3 until all items are clustered into a single cluster of size N.
10

Clustering and segmentation by K-Means


Desirables
Choose a fixed number of clusters
Choose cluster centers and point-cluster allocations to
minimize error
Algorithm
fix cluster centers; allocate points to closest cluster
fix allocation; compute best cluster centers

5/7/2016

11

Clustering and segmentation by K-Means

5/7/2016

12

Image Segmentation by K-Means


Select a feature vector for every pixel (color, texture,
position, or combination of these etc.)
Define a similarity measure between feature vectors
(Usually Euclidean Distance).
Apply K-Means Algorithm.
Apply Connected Components Algorithm.
Merge any components of size less than some
threshold to an adjacent component that is most
similar to it.

5/7/2016

13

Image Segmentation by K-Means (example)


Image

Clusters on intensity

Clusters on color

K-means clustering using intensity alone and color alone

5/7/2016

14

Image Segmentation by K-Means (example)

K-means using
color alone,
11 segments.

5/7/2016

15

Image Segmentation by K-Means (example)

K-means using
color and
position, 20
segments
5/7/2016

16

S-ar putea să vă placă și