International Journal of Engineering Trends and Technology (IJETT) – Volume 9 Number 2 - Mar 2014
Clustering Pairing Consensus
Using Image Classification
P.Ramya#1, Ms. Reeba*2,
Postgraduate student, sathyabama university, Chennai
Faculty,SathyabamaUniversity,Chennai,India
Abstract- Clustering pairing consensus is an algorithm it proposed
the fundamental matrix. The CPC produces the matched region
cluster and it produces the M-estimation to compute a matrix
function .Finally the best one is chosen as the final model from all
estimation .The proposed Clustering Pairing Consensus (CPC)
algorithm has been demonstrated to be able to effectively estimate
matrix to avoid difficult image pairs
Keywords-— Wide base line match, matrix function ,clustering
pairing consensus
Introduction
Data mining is the process of analysing data. Clustering is a
collection of object , Wide base line stereo algorithms is used
to tolerate a large change in view between the images it
include the higher precision of depth measurement and
smaller number of images needed to completely cover an
object or scene in previous work a topological clustering
algorithm was proposed by using Gaussian method ,it filter
only mismatches. Gaussian method is less efficient then
matrix method Clustering pairing consensus algorithm is used
to find best matched image. Clustering pairing consensus is
produced matched region cluster using matrix function. Best
one is choosen as the final model from all the estimation
.
II.RELATED WORK
In [1] The epipolar geometry (EG) is an matrix function to find the
image point correspondences in the possible presence of a dominant
scene plane. A novel RANSAC-based algorithm is presented that
handles in a unified manner that following three classes of scenes:
1. all points belong to a single scene plane,
2. majority of points belong to a dominant plane and the rest is off
the plane,
3. minority or no points lie in a scene plane (a general scene).
In the first case, only a plane homography is computed, in the other
cases, a correct EG is computed .The EG estimation process was
analyzed and showed that the larger the number of related
correspondences, the higher probability that DEGENSAC finds the
solution. As a consequence, with the increase in the number of points
in a dominant plane the running time of DEGENSAC decreases. It
was demonstrated experimentally. DEGENSAC estimates both the
EG and the homography correctly and reliably in the presence of a
dominant plane.
In[2] Wide base line technique is used To scope with wider
baselines, the affine geometric deformations in the image should
fully be taken into account during the matching process. One
approach is to deform a patch in the first image in an iterative way,
until it more or less fits a patch in the search that is involved reduces
ISSN: 2231-5381
the practicality of this approach. In contrast, our method is based on
the extraction and matching of invariant regions, and hence works on
the two images separately, without searching over the entire image or
applying combinatorics. The wide base line In each image, local
image patches are extracted in an affine invariant way, such that they
cover the same physical part of the scene .These patches or ‘invariant
regions’ are matched based on feature vectors of moment invariants
In [3].The difference between object and scene is described by
invariant features of Distinctive images. The features are highly
distinctive, in the sense that a single feature can be correctly matched
with high probability against a large database of features from many
images it also describes the features for object recognition The
recognition proceeds by matching individual features to a database of
features from known objects using a fast nearest-neighbor algorithm
it identify the objects among cluster and occlusion while achieving
near real-time performance . the major stages of computation used to
generate the set of image features:
1. Scale-space extrema detection:It search the over all scale and
image locations using Gaussian function to find the accrate point of
the scale
2. Keypoint localization: The each location scale. Keypoints are
selected based on measures of their stability.
3. Orientation assignment: One or more orientations are assigned to
each keypoint location based on local image . All future operations
are performed on image data that has been transformed relative to the
assigned orientation, scale, and location for each feature, it providing
invariance to these transformations.
4. Key point descriptor: The local image are measured at the
selected scale in the region around each key point . These are
transformed into a key point
In [4] Balanced Exploration and Exploitation Model Search (BEEM)
algorithm that works very well especially for these difficult images .
The algorithm includes the following main features: (1) Balanced use
of three search techniques: global random exploration, local
exploration near the current best solution and local exploitation to
improve the quality of the model. (2) Exploits available prior
information to accelerate the search process. (3) Uses the best found
model to guide the search process, escape from degenerate models
and to an efficient stopping criterion. (4) Presents a simple and
efficient method to estimate the epipolar geometry from two SIFT
correspondences. (5) Uses the locality-sensitive hashing (LSH)
approximate nearest neighbor algorithm for fast putative
correspondences generation. The resulting algorithm when tested on
real images with or without de- generate configurations gives quality
estimations and achieves significant speedups compared to the state
of the art algorithms! Finally, the number of global exploration
http://www.ijettjournal.org
Page 93
International Journal of Engineering Trends and Technology (IJETT) – Volume 9 Number 2 - Mar 2014
iteration of the BEEM algorithm is very low as a result of the use of
the prior information and the 2-SIFT method
III .EXISTING WORK
In existing topological clustering algorithm is used. It found the
mismatches using Gaussian method. Gaussian method is less
efficient then matrix method. it filter only mismatch image but it no
find the exact image. The random sampling and consensus is most
widely used this technique is search the possible solutions which is
solely by inliers. The epipolar geometry estimation geometry with
wide base line is difficult.
IV. PROPOSED WORK
Clustering pairing consensus algorithm used in proposed system.
Clustering pairing Consensus using matrix function Clustering
pairing consensus is produced the matched region cluster using
matrix function The CPC avoid the difficult image pairs Best one is
chosen as the final model from all the estimation. In matrix function
first test the relation between the scale parameter and he capability of
eliminating mismatches and reserving correct matches using CPC
algorithm. We are using six image pairs to find the accuracy image
sing matrix fnction for example There are seven MRCs in row 2 of
after the original SIFT matches are topologically clustered including
three correct MRCs (C1, C2,C3) and four incorrect MRCs (C4, C5,
C6, C7). C1, C2 and most of inliers lie approximately on a plane and
are thus consistent with many solutions. In the CPC method, region
matches are no longer regarded as atomic unit, and MRC pairs are
used to generate a series are shown below In general, correct MRCs
include most of true matches, butincorrect MRCs are composed of
only false matches. It’s easy to understand, the matches belong to the
same MRCs have an extremely similar affine transformation, and the
mismatches have independent, random affine transformations.
Therefore, it’s impossible to happen that the mismatches satisfy the
topological constraint just like the correct matches because the
mismatch has little chances to offer a suitable affine transformation
in the 6D affine transformation space. Thus, larger MRCs are more
likely to include the correct matches. Another property is that the
correct MRCs have more possibility to roughly correspond to the
planes in scene, which is very useful for breaking degeneracy.
scale c decreases until c = 2. We can see that the image The
clustering pairing consensusr ‘‘Oase’’ with scale c = 6, 4 and 2. The
MRCs including no less than three match regions are filtered. Rows
(from up to down): The initial SIFT matches; The MRCs with scale c
= 6; The MRCs with scale c = 4; The MRCs with scale c = 2. From
the experimental results, we can see that the Rr value decreases as the
scale c decreases, while the Ar value increases as the
“ oase”
ISSN: 2231-5381
http://www.ijettjournal.org
Page 94
International Journal of Engineering Trends and Technology (IJETT) – Volume 9 Number 2 - Mar 2014
V.CONCLUSION
Clustering pairing consensus algorithm is used. The best match
image is chosen as final using matrix function. Using wide baseline
we compare the no of images using matrix function we get exact best
match image
VI.REFERENCES
1.Chum.O., Werner,T.,MataS,J.,2005,Two view geometry estimation
unaffected b a dominant plane. In:proc.IEEE conf.computer vision
and pattern recognition,pp.1:772-1:779
2.Subbarao,R.,Meer,p., 2007.Discontinuity preserving filtering over
analytic manifolds. In:proceeding of the IEEE conference on
computer vision and pattern recognition pp 1-6
3.Lowe,D.,2004,Distinctive image features from scale invariant key
points int.j.comput vision 2(60).91-110.
4.Tordoff,B.,Murray,D.,2002 Guided sampling and consensus for
motion
estimation.in:proc,seventh
European
conf.computer
vision.pp.1:82-1:96
5.Tuytelaars,T., van Gool.L.,2004,Matching widely separated views
based on affine invariant regions.int.j.comput.vision 59(1),61-85
ISSN: 2231-5381
http://www.ijettjournal.org
Page 95