Professional Documents
Culture Documents
Textbook Image Analysis and Recognition 11Th International Conference Iciar 2014 Vilamoura Portugal October 22 24 2014 Proceedings Part Ii 1St Edition Aurelio Campilho Ebook All Chapter PDF
Textbook Image Analysis and Recognition 11Th International Conference Iciar 2014 Vilamoura Portugal October 22 24 2014 Proceedings Part Ii 1St Edition Aurelio Campilho Ebook All Chapter PDF
https://textbookfull.com/product/image-analysis-and-
recognition-11th-international-conference-iciar-2014-vilamoura-
portugal-october-22-24-2014-proceedings-part-i-1st-edition-
aurelio-campilho/
https://textbookfull.com/product/image-analysis-and-
recognition-17th-international-conference-iciar-2020-povoa-de-
varzim-portugal-june-24-26-2020-proceedings-part-ii-aurelio-
campilho/
https://textbookfull.com/product/image-analysis-and-
recognition-17th-international-conference-iciar-2020-povoa-de-
varzim-portugal-june-24-26-2020-proceedings-part-i-aurelio-
campilho/
https://textbookfull.com/product/computational-intelligence-
international-joint-conference-ijcci-2014-rome-italy-
october-22-24-2014-revised-selected-papers-1st-edition-juan-
julian-merelo/
Image Analysis
and Recognition
11th International Conference, ICIAR 2014
Vilamoura, Portugal, October 22–24, 2014
Proceedings, Part II
123
Lecture Notes in Computer Science 8815
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Alfred Kobsa
University of California, Irvine, CA, USA
Friedemann Mattern
ETH Zurich, Zürich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
Oscar Nierstrasz
University of Bern, Bern, Switzerland
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbruecken, Germany
More information about this series at http://www.springer.com/series/7412
Aurélio Campilho Mohamed Kamel (Eds.)
•
Image Analysis
and Recognition
11th International Conference, ICIAR 2014
Vilamoura, Portugal, October 22–24, 2014
Proceedings, Part II
123
Editors
Aurélio Campilho Mohamed Kamel
Faculty of Engineering Department of Electrical and Computer
University of Porto Engineering
Porto University of Waterloo
Portugal Waterloo, ON
Canada
LNCS Sublibrary: SL6 – Image Processing, Computer Vision, Pattern Recognition, and Graphics
This is the 11th edition of the ICIAR series of annual conferences offering an
opportunity for the participants to interact and present their latest research in theory,
methodology, and applications of image analysis and recognition. ICIAR 2014, the
International Conference on Image Analysis and Recognition, was held in Vila Moura,
Portugal, October 22–24, 2014. ICIAR is organized by AIMI – Association for Image
and Machine Intelligence, a not-for-profit organization registered in Ontario, Canada.
For ICIAR 2014, we received a total of 177 full papers from 39 countries. Before
the review process all the papers were checked for similarity using a comparison
database of scholarly work. The review process was carried out by members of the
Program Committee and other reviewers. Each paper was reviewed by at least two
reviewers, and checked by the conference chairs. A total of 107 papers were finally
accepted and appear in the two volumes of this proceedings. We would like to sincerely
thank the authors for responding to our call, and we thank the reviewers for the careful
evaluation and feedback provided to the authors. It is this collective effort that resulted
in the strong conference program and high-quality proceedings.
Each year we attempt to focus on a specific topic for the keynote speeches and
conduct a panel discussion on the topic.
This year, the conference theme was focused on the topic “Sparse Representations
for Image Analysis and Recognition.” We were very pleased to include three out-
standing keynote talks on this topic: “Optimization Algorithms for Sparse Represen-
tations: Some History and Recent Developments” by Mário Figueiredo, Instituto
Superior Técnico Portugal; “Morphological Diversities in Astrophysics” by Jean-Luc
Starck, CosmoStat Laboratory, France; and “Sparse Stochastic Processes with Appli-
cation to Biomedical Imaging” by Michael Unser, Ecole Polytechnique Fédérale de
Lausanne, Switzerland. The keynote speakers also participated in the panel “Sparse
Representation for Image Analysis and Recognition: Trends and Applications.” We
would like to express our gratitude to the keynote speakers for accepting our invitation
to share their vision and recent advances in their areas of expertise, which are at the
core of the topics of the conference.
We would like to thank Khaled Hammouda, the webmaster of the conference, for
maintaining the Web pages, interacting with the authors, and preparing the
proceedings.
As all conferences, the success of ICIAR 2014 is attributed to the effort and work of
many people, including members of the Organizing Committee, staff, and volunteers.
We gratefully acknowledge their support and efforts.
We are also grateful to Springer’s editorial staff for supporting this publication in the
LNCS series. We also would like to acknowledge the professional service of Viagens
Abreu in taking care of the registration process and the special events of the
conference.
VI Preface
Finally, we are very pleased to welcome all the participants to ICIAR 2014. For
those who were not able to attend, we hope this publication provides a good view into
the research presented at the conference, and we look forward to meeting you at the
next ICIAR conference.
General Chairs
Aurélio Campilho University of Porto, Portugal
Mohamed Kamel University of Waterloo, Canada
Conference Secretariat
Viagens Abreu SA, Portugal
Webmaster
Khaled Hammouda Waterloo, Ontario, Canada
Advisory Committee
M. Ahmadi University of Windsor, Canada
P. Bhattacharya Concordia University, Canada
T.D. Bui Concordia University, Canada
M. Cheriet University of Quebec, Canada
E. Dubois University of Ottawa, Canada
Z. Duric George Mason University, USA
G. Granlund Linköping University, Sweden
L. Guan Ryerson University, Canada
M. Haindl Institute of Information Theory and Automation,
Czech Republic
E. Hancock University of York, UK
J. Kovacevic Carnegie Mellon University, USA
M. Kunt Swiss Federal Institute of Technology (EPFL),
Switzerland
VIII ICIAR 2014
Program Committee
A. Abate University of Salerno, Italy
M. Ahmed Wilfrid Laurier University, Canada
L. Alexandre University of Beira Interior, Portugal
J. Alirezaie Ryerson University, Canada
G. Andreu-Garcia Universitat Politècnica de València, Spain
H. Araújo University of Coimbra, Portugal
Emilio Balaguer-Ballester Bournemouth University, UK
T. Barata University of Coimbra, Portugal
J. Barbosa University of Porto, Portugal
J. Batista University of Coimbra, Portugal
R. Bernardes University of Coimbra, Portugal
A. Bezerianos National University of Singapore, Singapore
J. Bioucas Technical University of Lisbon, Portugal
I. Bloch Télécom ParisTech, France
T.D. Bui Concordia University, Canada
C. Busch Gjøvik University College, Norway
F. Camastra University of Naples Parthenope, Italy
J. Cardoso University of Porto, Portugal
G. Carneiro University of Adelaide, Australia
M. Coimbra University of Porto, Portugal
M. Correia University of Porto, Portugal
J. Debayle Ecole Nationale Supérieure des Mines
de Saint-Étienne, France
J. Dias University of Coimbra, Portugal
G. Doretto West Virginia University, USA
H. du Buf University of the Algarve, Portugal
J. Fernandez Centro Nacional de Biotecnología – CSIC, Spain
I. Fondón University of Seville, Spain
A. Fred Technical University of Lisbon, Portugal
G. Freeman University of Waterloo, Canada
ICIAR 2014 IX
Reviewers
Supported by
Biometrics
Periodic Background Pattern Detection and Removal for Cell Tracking . . . . 123
Tiago Esteves, ¢ngela Carvalho, Fernando Jorge Monteiro,
and Pedro Quelhas
Nerve Detection in Ultrasound Images Using Median Gabor Binary Pattern. . . 132
Oussama Hadjerci, Adel Hafiane, Pascal Makris, Donatello Conte,
Pierre Vieyres, and Alain Delbos
Computer-Aided Diagnosis
Grading Cancer from Liver Histology Images Using Inter and Intra Region
Spatial Relations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Micka˝l Garnier, Maya Alsheh Ali, Johanne Seguin, Nathalie Mignet,
Thomas Hurtut, and Laurent Wendling
3D Imaging
Adaptive Feature Selection for Object Tracking with Particle Filter . . . . . . . 395
Darshan Venkatrayappa, D¯sir¯ Sidib¯, Fabrice Meriaudeau,
and Philippe Montesinos
Robot Vision
Multiple Camera Approach for SLAM Based Ultrasonic Tank Roof Inspection . . . 453
Christian Freye, Christian Bendicks, Erik Lilienblum,
and Ayoub Al-Hamadi
Sparse Representation
Remote Sensing
Applications
1 Introduction
Analysis of naturally occurring human emotions has become the main focus of
recent research in the field of affective computing. Emotional analysis is consid-
ered a vital step towards building efficient and more realistic intelligent human-
computer interfaces. The focus is now directed towards recognition in terms of
dimensional and continuous description, rather than a small number of discrete
emotion categories. Numerical representation of emotions in a multi-dimensional
space is considered a more appropriate representation that can reflect the gra-
dated nature of emotions. Moreover, human natural affective behavior is mul-
timodal, subtle, and complex, which makes it challenging to map the affective
human state into a single label or discrete number of classes [4].
Facial expressions and speech are the two modalities most commonly used to
analyze emotions in human interaction. While facial expressions are considered
the major modality in human communication, according to [8], speech is the
fastest and most natural method of communication between humans.
c Springer International Publishing Switzerland 2014
A. Campilho and M. Kamel (Eds.): ICIAR 2014, Part II, LNCS 8815, pp. 3–11, 2014.
DOI: 10.1007/978-3-319-11755-3 1
4 R. Araujo and M.S. Kamel
The bulk of the approaches found in the literature are on supervised learning,
despite the fact that the labeling process is demanding. With the abundance
of data available in this domain and the burdensome nature of the labeling
process, it is understandable that unsupervised methods should be pursued more
intensively. However, purely unsupervised methods may not produce desirable
results due to the complexity of the problem at hand. This need to balance
the demands of process and accuracy of results motivated us to pursue a semi-
supervised approach.
In this paper, we extend our previously proposed Semi-Supervised Aligned
Cluster Analysis (SSACA) method [1] by using an exhaustive constraint propa-
gation approach and apply it to the AVEC audio-visual database.
2 Related Work
Very few works have applied unsupervised methods to emotion analysis in gen-
eral. This observation is true for both of the two most-used modalities: facial
expressions and speech. A possible reason for the shortage of unsupervised work
is due to the lack of the temporal aspect of the traditional clustering algorithms.
The bulk of the methods are supervised, with HMM being the most used method
in terms of audio. In terms of visual features, Relevance Vector Machine (RVM)
has shown very good results [7].
Some supervised works that tackle the problem as a dimensional and con-
tinuous emotion recognition are [11] [7] [3]. Wollmer et al. [11] have studied the
estimation of emotions from speech in the valence and activation dimensions
using Long- Short-Term Recurrent Neural Networks. Nicolaou et al. [7] have
proposed the use of Output-Associative Relevance Vector Machine (OA-RVM)
for dimensional and continuous estimation of emotions from facial expressions.
Grimm et al. [3] have compared the performance of Support Vector Regression,
Fuzzy k-Nearest Neighbor, and Rule-based Fuzzy Logic classifiers as estimators
of spontaneously expressed emotions in speech from three continuous-valued
emotion primitives.
In terms of unsupervised methods for emotion analysis, one recent publication
is the work of De la Torre et al. [2], who have proposed a temporal segmentation
method of facial gestures to cluster similar facial actions. Zhou et al. [13] have
examined facial events directly from naturally occurring videos, using tempo-
ral clustering. They use two algorithms for this task: Aligned Cluster Analysis
(ACA) and a multi-subject correspondence for matching expressions.
Both of these works, however, analyze only one modality, and they use either
categorical labels or action units. In the case of speech and unsupervised meth-
ods, most of the research is on speech segmentation for speech recognition or on
speech separation.
Recently, we proposed a semi-supervised method, SSACA [1], which uses
pairwise constraints in the form of must-links and cannot-links as a way to add
side information to help the clustering process, and to boost its performance
with minimal supervised information. We applied this method to a naturalis-
tic database, and the results showed improvements compared to the original
Audio-Visual Emotion Analysis Using Semi-Supervised Temporal Clustering 5
3 Model Description
The emotional behavior of a person can be treated as a time series, wherein
the specific emotion primitive being evaluated varies over time. The goal is to
factorize (segment) multiple time series into disjointed segments that belong to
k temporal clusters. Essentially, we have a temporal clustering problem. The
idea is to have frames within a segment that are similar to each other and non-
overlapping segments that belong to k temporal clusters.
k
m
2
gci ψ(X[si ,si+1 ) ) − zc = [ψ(Y1 , . . . , ψ(Ym ) − ZGF ,
2
Jaca (G, s) =
c=1 i=1
dist2ψ (Yi , zc )
Fields (HMRF), with squared Euclidean distance and a certain class of constraint
penalty function, can be expressed as a special case of the weighted kernel k -
means. SSACA, which is based on kernel k -means, may use the same framework
of HMRF semi-supervised clustering. Thus, we can write the SSACA objective
function as:
k
m
2
Jssaca (G, s) = gci ψ(X[si ,si+1 ) ) − zc − wij + wij (2)
c=1 i=1
x ,x ∈M x ,x ∈C
i j i j
dist2ψ (Yi , zc ) gi =gj gi =gj
Algorithm 1. SS DPSearch
parameter: nmax , k, nml , ncl
input: G ∈ {0, 1}k×ṁ , ṡ ∈ R(ṁ+1) , K ∈ Rn×n , T ∈ Rm×m , M ∈ Znml ×2 , C ∈ Zncl ×2
output: G ∈ {0, 1}k×m , ṡ ∈ R(m+1)
1: headTail = getHeadTails(M, C);
2: for v = 1 to n do
3: J(v) ← ∞;
4: if v ≥ headTail(:,1) and v < headTail(:,2) then
5: continue;
6: end if
7: if isTail(v) then
8: for j = 1 to ṁ do
9: Retrive directly from T(X[i,v] , Y˙j );
10: end for
11: c∗ ← arg minc distψ (X[i,v] , z˙c );
12: J ← distψ (X[i,v] , żc∗ );
∗
13: J([i, v]) ← J, g[i,v] ← e∗c , i∗[i,v] ← i;
14: else
15: for nv = 1 to min(nmax , v) do
16: { Same as DPSearch}
17: end for
18: end if
19: end for
{Perform backward segmentation}
UPSHOT OF FEUDALISM.