image analysis and recognition, 2 conf., iciar 2005(lncs3656, springer, 2005)

Lecture Notes in Computer Science 3656Commenced Publication in 1973Founding and Former Series Editors:Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board

David HutchisonLancaster University, UK

Takeo KanadeCarnegie Mellon University, Pittsburgh, PA, USA

Josef KittlerUniversity of Surrey, Guildford, UK

Jon M. KleinbergCornell University, Ithaca, NY, USA

Friedemann MatternETH Zurich, Switzerland

John C. MitchellStanford University, CA, USA

Moni NaorWeizmann Institute of Science, Rehovot, Israel

Oscar NierstraszUniversity of Bern, Switzerland

C. Pandu RanganIndian Institute of Technology, Madras, India

Bernhard SteffenUniversity of Dortmund, Germany

Madhu SudanMassachusetts Institute of Technology, MA, USA

Demetri TerzopoulosNewYork University, NY, USA

Doug TygarUniversity of California, Berkeley, CA, USA

Moshe Y. VardiRice University, Houston, TX, USA

Gerhard WeikumMax-Planck Institute of Computer Science, Saarbruecken, Germany

Mohamed Kamel Aurlio Campilho (Eds.)

Image Analysisand Recognition

Second International Conference, ICIAR 2005Toronto, Canada, September 28-30, 2005Proceedings

13

Volume Editors

Mohamed KamelUniversity of WaterlooDepartment of Electrical and Computer EngineeringWaterloo, Ontario N2L 3G1, CanadaE-mail: [email protected]

Aurlio CampilhoUniversity of PortoFaculty of EngineeringInstitute of Biomedical EngineeringRua Dr. Roberto Friaas, 4200-465 Porto, PortugalE-mail: [email protected]

Library of Congress Control Number: 2005932546

CR Subject Classification (1998): I.4, I.5, I.3.5, I.2.10, I.2.6, F.2.2

ISSN 0302-9743ISBN-10 3-540-29069-9 Springer Berlin Heidelberg New YorkISBN-13 978-3-540-29069-8 Springer Berlin Heidelberg New York

This work is subject to copyright. All rights are reserved, whether the whole or part of the material isconcerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting,reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publicationor parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965,in its current version, and permission for use must always be obtained from Springer. Violations are liableto prosecution under the German Copyright Law.

Springer is a part of Springer Science+Business Media

springeronline.com

Springer-Verlag Berlin Heidelberg 2005Printed in Germany

Typesetting: Camera-ready by author, data conversion by Scientific Publishing Services, Chennai, IndiaPrinted on acid-free paper SPIN: 11559573 06/3142 5 4 3 2 1 0

Preface

ICIAR 2005, the International Conference on Image Analysis and Recognition,was the second ICIAR conference, and was held in Toronto, Canada. ICIAR isorganized annually, and alternates between Europe and North America. ICIAR2004 was held in Porto, Portugal. The idea of oering these conferences came asa result of discussion between researchers in Portugal and Canada to encouragecollaboration and exchange, mainly between these two countries, but also withthe open participation of other countries, addressing recent advances in theory,methodology and applications.

The response to the call for papers for ICIAR 2005 was encouraging. From 295full papers submitted, 153 were nally accepted (80 oral presentations, and 73posters). The review process was carried out by the Program Committee mem-bers and other reviewers; all are experts in various image analysis and recognitionareas. Each paper was reviewed by at least two reviewers, and also checked bythe conference co-chairs. The high quality of the papers in these proceedings isattributed rst to the authors, and second to the quality of the reviews providedby the experts. We would like to thank the authors for responding to our call,and we wholeheartedly thank the reviewers for their excellent work, and for theirtimely response. It is this collective eort that resulted in the strong conferenceprogram and high-quality proceedings in your hands.

We were very pleased to be able to include in the conference program keynotetalks by two world-renowned experts: Prof. Anastasios (Tas) N. Venetsanopou-los, Dean of the Faculty of Applied Science and Engineering at the University ofToronto, Canada; and Prof. Jelena Kovacevic, Director of the Center for Bioim-age Informatics, Departments of Biomedical Engineering & Electrical and Com-puter Engineering at Carnegie Mellon University, USA. We would like to expressour sincere gratitude to each of them for accepting our invitations.

We would like to thank Khaled Hammouda, the webmaster of the conference,for maintaining the Web pages, interacting with the authors and preparing theproceedings; and Cathie Lowell for her administrative assistance. We would alsolike to thank the members of the Local Organizing Committee for their adviceand help. We also appreciate the help of the Springer editorial sta ChristineGunther, Anna Kramer, and Alfred Hofmann, for supporting this publication inthe LNCS series.

Finally, we were very pleased to welcome all the participants to this confer-ence. For those who did not attend, we hope this publication provides a goodview into the research presented at the conference, and we look forward to meet-ing you at the next ICIAR conference.

September 2005 Mohamed Kamel, Aurelio Campilho

ICIAR 2005 International Conference onImage Analysis and Recognition

General Chair

Mohamed KamelUniversity of Waterloo, [email protected]

General Co-chair

Aurelio CampilhoUniversity of Porto, [email protected]

Local Organizing Committee

Otman BasirUniversity of Waterloo, [email protected]

Alex BotIEEE Toronto Section, [email protected]

David ClausiUniversity of Waterloo, [email protected]

Mahmoud El-SakkaUniversity of Western Ontario, [email protected]

Paul FieguthUniversity of Waterloo, [email protected]

Rastislav LukacUniversity of Toronto, [email protected]

Kostas PlataniotisUniversity of Toronto, [email protected]

Hamid TizhooshUniversity of Waterloo, [email protected]

Webmaster

Khaled HammoudaUniversity of Waterloo, [email protected]

Organization VII

Supported by

Pattern Analysis and Machine Intelli-gence Group, University of Waterloo,Canada

Department of Electrical and Com-puter Engineering, Faculty of Engineer-ing, University of Porto, Portugal

INEB Instituto de EngenhariaBiomedica

IEEE Toronto Section

IEEE Kitchener-Waterloo Section

Advisory and Program Committee

M. Abdallah American University of Beirut, LebanonP. Abolmaesumi Queens University, CanadaR. Abugharbieh University of British Columbia, CanadaM. Ahmadi University of Windsor, CanadaM. Ahmed Wilfrid Laurier University, CanadaJ. Alirezaie Ryerson University, CanadaA. Amin University of New South Wales, AustraliaD. Androutsos Ryerson University, CanadaH. Araujo University of Coimbra, PortugalJ. Barron University of Western Ontario, CanadaO. Basir University of Waterloo, CanadaJ. Bioucas Technical University of Lisbon, PortugalA. Bot IEEE Toronto Section, CanadaB. Boubakeur University of Windsor, CanadaT. Bui Concordia University, CanadaM. Cheriet University of Quebec, CanadaD. Chiu University of Guelph, CanadaD. Clausi University of Waterloo, CanadaL. Corte-Real University of Porto, Portugal

VIII Organization

E. Dubois University of Ottawa, CanadaM. El-Sakka University of Western Ontario, CanadaR. Fazel University of Manitoba, CanadaM. Ferretti University of Pavia, ItalyP. Fieguth University of Waterloo, CanadaM. Figueiredo Technical University of Lisbon, PortugalA. Fred Technical University of Lisbon, PortugalG. Freeman University of Waterloo, CanadaL. Guan Ryerson University, CanadaM. Haindl Institute of Information Theory and Automation,

Czech RepublicE. Hancock University of York, UKE. Jernigan University of Waterloo, CanadaJ. Jorge INESC-ID, PortugalG. Khan Ryerson University, CanadaS. Krishnan Ryerson University, CanadaA. Krzyzak Concordia University, CanadaR. Laganie`re University of Ottawa, CanadaR. Lins Universidade Federal de Pernambuco, BrazilS. Lu Memorial University of Newfoundland, CanadaR. Lukac University of Toronto, CanadaJ. Marques Technical University of Lisbon, PortugalA. Mendonca University of Porto, PortugalJ. Orchard University of Waterloo, CanadaA. Ouda University of Western Ontario, CanadaA. Padilha University of Porto, PortugalP. Payeur University of Ottawa, CanadaF. Perales University of the Balearic Islands, SpainF. Pereira Technical University of Lisbon, PortugalN. Peres de la Blanca University of Granada, SpainE. Petrakis Technical University of Crete, GreeceP. Pina Technical University of Lisbon, PortugalA. Pinho University of Aveiro, PortugalJ. Pinto Technical University of Lisbon, PortugalF. Pla University of Jaume I, SpainK. Plataniotis University of Toronto, CanadaT. Rabie University of Toronto, CanadaP. Radeva Universitat Auto`noma de Barcelona, SpainL. Rueda University of Windsor, CanadaF. Samavati University of Calgary, CanadaB. Santos University of Aveiro, PortugalG. Schaefer Nottingham Trent University, UKP. Scheunders University of Antwerp, BelgiumJ. Sequeira Ecole Superieure dIngenieurs de Luminy, FranceM. Sid-Ahmed University of Windsor, Canada

Organization IX

J. Silva University of Porto, PortugalW. Skarbek Warsaw University of Technology, PolandB. Smolka Silesian University of Technology, PolandJ. Sousa University of Coimbra, PortugalC. Suen Concordia University, CanadaS. Sural Indian Institute of Technology, Kharagpur, IndiaG. Thomas University of Waterloo, CanadaH. Tizhoosh University of Waterloo, CanadaD. Vandermeulen Catholic University of Leuven, BelgiumA. Venetsanopoulos University of Toronto, CanadaM. Vento University of Salerno, ItalyE. Vrscay University of Waterloo, CanadaR. Ward University of British Columbia, CanadaM. Wirth University of Guelph, CanadaJ. Wu University of Windsor, CanadaJ. Yeow University of Waterloo, CanadaJ. Zelek University of Waterloo, CanadaX. Zhang Ryerson University, Canada

Reviewers

W. Abd-Almageed University of Maryland, USAA. Adegorite University of Waterloo, CanadaN. Alajlan University of Waterloo, CanadaB. Avila Universidade Federal de Pernambuco, BrazilT. Barata Instituto Superior Tecnico, PortugalE. Cernadas University of Vigo, SpainL. Chen University of Waterloo, CanadaS. Chowdhury University of Waterloo, CanadaM. Correia University of Porto, PortugalR. Dara University of Waterloo, CanadaA. Dawoud University of South Alabama, USAO. El Badawy University of Waterloo, CanadaI. El Rube University of Waterloo, CanadaJ. Glasa Slovak Academy of Sciences, SlovakiaV. Grau University of Oxford, UKC. Hong Hong Kong Polytechnic, Hong Kong, ChinaA. Kong University of Waterloo, CanadaJ. Martnez University of Jaume I, SpainB. Miners University of Waterloo, CanadaA. Monteiro University of Porto, PortugalF. Monteiro IPB, PortugalD. Oliveira Universidade Federal de Pernambuco, BrazilA. Picariello University of Naples, ItalyA. Puga University of Porto, Portugal

X Organization

S. Rahnamayan University of Waterloo, CanadaR. Rocha INEB Instituto de Engenharia Biomedica, PortugalM. Sabri University of Waterloo, CanadaF. Sahba University of Waterloo, CanadaA. Silva Universidade Federal de Pernambuco, BrazilB. van Ginneken Image Sciences Institute, NetherlandsC. Vinhais ISEP, PortugalD. Xi University of Waterloo, CanadaC. Yang National Dong Hwa University, TaiwanQ. Yu University of Waterloo, Canada

Table of Contents

Image Segmentation

Localization Scale Selection for Scale-Space SegmentationSokratis Makrogiannis, Nikolaos Bourbakis . . . . . . . . . . . . . . . . . . . . . . . 1

Image Segmentation for the Application of the Neugebauer ColourPrediction Model on Inkjet Printed Ceramic Tiles

P. Latorre, G. Peris-Fajarnes, M.A.T. Figueiredo . . . . . . . . . . . . . . . . . 9

FCM with Spatial and Multiresolution Constraints for ImageSegmentation

Adel Haane, Bertrand Zavidovique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Combined Color and Texture Segmentation Based on Fibonacci LatticeSampling and Mean Shift

Yuchou Chang, Yue Zhou, Yonggang Wang . . . . . . . . . . . . . . . . . . . . . . 24

Unsupervised Image Segmentation Using Contourlet Domain HiddenMarkov Trees Model

Yuheng Sha, Lin Cong, Qiang Sun, Licheng Jiao . . . . . . . . . . . . . . . . . 32

A Novel Color C-V Method and Its ApplicationLi Chen, Yue Zhou, Yonggang Wang . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

SAR Image Segmentation Using Kernel Based Spatial FCMXiangrong Zhang, Tan Shan, Shuang Wang, Licheng Jiao . . . . . . . . . . 48

Segmentation of Nanocolumnar Crystals from Microscopic ImagesDavid Cuesta Frau, Mara Angeles Hernandez-Fenollosa,Pau Mico Tormos, Jordi Linares-Pellicer . . . . . . . . . . . . . . . . . . . . . . . . 55

Image and Video Processing and Analysis

Mutual Information-Based Methods to Improve Local Region-of-InterestImage Registration

K.P. Wilkie, E.R. Vrscay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

Image Denoising Using Complex Wavelets and Markov Prior ModelsFu Jin, Paul Fieguth, Lowell Winger . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

XII Table of Contents

A New Vector Median Filter Based on Fuzzy MetricsSamuel Morillas, Valentn Gregori, Guillermo Peris-Fajarnes,Pedro Latorre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Image Denoising Using Neighbor and Level DependencyDongwook Cho, Tien D. Bui, Guangyi Chen . . . . . . . . . . . . . . . . . . . . . . 91

Time Oriented Video SummarizationChaoqiang Liu, Tao Xia, Hui Li . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

Shadow Removal in Gradient DomainZhenlong Du, Xueying qin, Hai Lin, Hujun Bao . . . . . . . . . . . . . . . . . . 107

Ecient Global Weighted Least-Squares Translation Registration inthe Frequency Domain

Je Orchard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Isotropic Blur Identication for Fully Digital Auto-focusingJeongho Shin, Sunghyun Hwang, Seong-Won Lee, Joonki Paik . . . . . . 125

Edge Detection ModelsQ.H. Zhang, S. Gao, Tien D. Bui . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Video Stabilization Using Kalman Filter and Phase CorrelationMatching

Ohyun Kwon, Jeongho Shin, Joonki Paik . . . . . . . . . . . . . . . . . . . . . . . . 141

Wavelet Image Denoising Using Localized Thresholding OperatorsM. Ghazel, G.H. Freeman, E.R. Vrscay, R.K. Ward . . . . . . . . . . . . . . . 149

Type-2 Fuzzy Image EnhancementP. Ensa, H.R. Tizhoosh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

A Multi-level Framework for Video Shot StructuringYun Zhai, Mubarak Shah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

All-in-Focus Imaging Using a Series of Images on Dierent Focal PlanesMark Antunes, Michael Trachtenberg, Gabriel Thomas, Tina Shoa . . 174

Skew Estimation and Correction for Form Documents Using WaveletDecomposition

Dihua Xi, Mohamed Kamel, Seong-Whan Lee . . . . . . . . . . . . . . . . . . . . 182

Scalable e-Learning Multimedia Adaptation ArchitectureMazen Almaoui, Konstantinos N. Plataniotis . . . . . . . . . . . . . . . . . . . . . 191

Table of Contents XIII

Highlight Detection and Removal Based on ChromaticityShu-Chang Xu, Xiuzi Ye, Yin Wu, Sanyuan Zhang . . . . . . . . . . . . . . . . 199

Digital Video Scrambling Using Motion Vector and Slice RelocationSang Gu Kwon, Woong Il Choi, Byeungwoo Jeon . . . . . . . . . . . . . . . . . 207

Weighted Information Entropy: A Method for Estimating the ComplexDegree of Infrared Images Backgrounds

Lei Yang, Jie Yang, Ningsong Peng, Jianguo Ling . . . . . . . . . . . . . . . . 215

Neural Network Adaptive Switching Median Filter for the Restorationof Impulse Noise Corrupted Images

Pavel S. Zvonarev, Ilia V. Apalkov, Vladimir V. Khryashchev,Irina V. Reznikova . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

A Shot Boundary Detection Method for News Video Based on RoughSets and Fuzzy Clustering

Xin-bo Gao, Bing Han, Hong-bing Ji . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

Image Enhancement via Fusion Based on Laplacian PyramidDirectional Filter Banks

Hai-yan Jin, Xiao-hui Yang, Li-cheng Jiao, Fang Liu . . . . . . . . . . . . . 239

Wavelet-Based Methods for Improving Signal-to-Noise Ratio in PhaseImages

Hector Cruz-Enriquez, Juan V. Lorenzo-Ginori . . . . . . . . . . . . . . . . . . . 247

Image Evaluation FactorsHongxun Yao, Min-Yu Huseh, Guilin Yao, Yazhou Liu . . . . . . . . . . . . 255

Monoscale Dual Ridgelet FrameTan Shan, Licheng Jiao . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

Description Selection Scheme for Intermediate Frame Based MultipleDescription Video Streaming

S. Pavan, G. Sridhar, V. Sridhar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270

Background Removal of Document Images Acquired Using PortableDigital Cameras

Andre R. Gomes e Silva, Rafael Dueire Lins . . . . . . . . . . . . . . . . . . . . . 278

Comparison of the Image Distortion Correction Methods for an X-RayDigital Tomosynthesis System

J.Y. Kim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286

XIV Table of Contents

Image and Video Coding

An Ecient Video Watermarking Scheme Using Adaptive Thresholdand Minimum Modication on Motion Vectors

Kyung-Won Kang, Kwang-Seok Moon, Gwang-Seok Jung,Jong-Nam Kim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

Lossless Compression of Correlated Images/Data with Low ComplexityEncoder Using Distributed Source Coding Techniques

Mortuza Ali, Manzur Murshed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302

Automatically Detecting Symmetries in Decorative TilesRafael Dueire Lins, Daniel Marques Oliveira . . . . . . . . . . . . . . . . . . . . . 310

A Fast Video Mixing Method for Multiparty Video ConferenceXin-Gang Liu, Kook-Yeol Yoo, Kwang-Deok Seo . . . . . . . . . . . . . . . . . . 320

Grayscale Two-Dimensional Lempel-Ziv EncodingNathanael J. Brittain, Mahmoud R. El-Sakka . . . . . . . . . . . . . . . . . . . . 328

Unequal Error Protection Using Convolutional Codes for PCA-CodedImages

Sabina Hosic, Aykut Hocanin, Hasan Demirel . . . . . . . . . . . . . . . . . . . . 335

Design of Tree Filter Algorithm for Random Number Generator inCrypto Module

Jinkeun Hong, Kihong Kim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343

Layer Based Multiple Description Packetized CodingCanhui Cai, Jing Chen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351

Extended Application of Scalable Video Coding MethodsZhi-gang Li, Zhao-yang Zhang, Biao Wu, Ying Zhang . . . . . . . . . . . . . 359

Accelerated Motion Estimation of H.264 on Imagine Stream ProcessorHaiyan Li, Mei Wen, Chunyuan Zhang, Nan Wu, Li Li,Changqing Xun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367

MPEG-2 Test Stream with Static Test Patterns in DTV SystemSoo-Wook Jang, Gwang-Soon Lee, Eun-Su Kim, Sung-Hak Lee,Kyu-Ik Sohng . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375

Speed Optimization of a MPEG-4 Software Decoder Based on ARMFamily Cores

Linjian Mo, Haixiang Zhang, Jiajun Bu, Chun Chen . . . . . . . . . . . . . . 383

MPEG-2 Test Stream with Static Test Patternsin DTV System

Soo-Wook Jang1, Gwang-Soon Lee2, Eun-Su Kim1,Sung-Hak Lee1, and Kyu-Ik Sohng1

1 School of Electronic Engineering and Computer Science,Kyungpook National University 1370, Sankyug-Dong, Buk-Gu, Daegu,

702-701, Korea{jjang, saeloum, shark2, kisohng}@ee.knu.ac.kr

2 Electronics and Telecommunications Research Institute,161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305-350, Korea

[email protected]

Abstract. MPEG-2 test stream for evaluation the static picture qualityof digital television (DTV) should meet both good picture quality andstable bit rate. In this paper, we present a method for generating a highquality test stream to evaluate the static picture quality in DTV receiver.The proposed method is suitable for encoding the static test pattern, suchas multiburst and crosshatch, and is based on user-dened quantizationand adaptive zero stung algorithm. The user-dened quantization issuitable for minimizing the quantization error, which is the reason ofdegradation of picture quality, and the adaptive zero stung algorithmis used to solve the overow of video buer verier (VBV) buer whileencoding process by MPEG-2 encoder. Experimental results show thatthe average PSNR and the bit rate of the proposed method have moreecient and stable than those of the conventional.

1 Introduction

The basic structure of current TV system is newly created by digital technology,delivering high quality video, audio, and data. As the DTV service becomes morewidely used than traditional analog TV system, how to measure a picture qualityin DTV becomes the main problem. So, the need for a reference test stream toevaluate the picture quality of DTV has substantially increased [1]. The teststreams are needed to perform the role traditionally taken by static analoguetest patterns and are must satised MPEG-2 regulation [2]. From equipmentmanufacture to system monitoring, the test stream are must guarantee the goodpicture quality and stable bit rate during the decoding process to evaluate thepicture quality [3], [4].

At the heart of coding in MPEG-2 is the discrete cosine transform (DCT).When the DCT is computed for a block of pels, it is desirable to represent thecoecients for high spatial frequencies with less precision, which is consideredthe spatial frequency response of the human visual system (HVS). This is done by

M. Kamel and A. Campilho (Eds.): ICIAR 2005, LNCS 3656, pp. 375382, 2005.c Springer-Verlag Berlin Heidelberg 2005

376 S.-W. Jang et al.

a process called quantization. However, if the spatial frequencies become higherand higher, quantization errors are increased. There is the reason of degradationof picture quality at high spatial frequencies.

For the bit rate control in MPEG-2, if more coding bit than the allocatedtarget number are exhausted, the remaining bit resource in the GOP is gettingsmaller, In such a case, insucient coding bits can be allocated to the picturesat the end of the GOP, which may result in severe degradation of picture qualityif buer overow. C.-T. Ahn et al. [5] proposed zero bit stung to prevent theVBV buer overow which can cause annoying picture quality degradation. Butthis method is used simply to prevent buer overow without improving picturequality, so that it is dicult to expect good picture quality. if the target numberof bits and the number of actual coding bits for each picture do not match well,than the degradation of picture quality and buer overow may occur at theend of the GOP. J. W. Lee et al. [6] proposed target bit matching for MPEG-2video rate control to solve this problem. This algorithm is based on accurate bitestimation, but there is only better than the MPEG-2 Test Model 5 (TM5)[7]algorithm for the complex and fast-moving sequence, while both algorithms yieldsimilar performances for the simple and slowly-moving sequence [6].

This paper proposed a new method for generating the reference test streamto evaluate the picture quality of DTV. The test stream is encoded the statictest pattern, such as multiburst and crosshatch, by using MPEG-2 encoder. Inorder to obtain the test stream with good picture quality, we propose a new user-dened quantization and an adaptive zero stung algorithm. The user-denedquantization table is suitable for minimizing the quantization error which is thereason of degradation of picture quality, and the adaptive zero stung algorithmthat determines the number of zero stung bits for preventing buer overowand maintaining good picture quality at the same time.

To evaluate the proposed method, we generate the test pattern stream us-ing MPEG-2 encoder with and without the proposed algorithm respectively.Experimental results show that the test pattern stream has a stable bit rateduring the decoding process, and the PSNR of the proposed method is about4 dB higher than that of the conventional cases. The proposed method has astable bit rate and good picture quality and it is suitable for evaluation a DTVreceiver.

2 The Proposed Algorithm

To obtain a test stream with good picture quality and stable bit rate, this paperproposed a new algorithm, which is suitable for encoding the static test pattern.The proposed method is shown in Fig. 1.

2.1 Proposed User-Dened Quantization

DCT and visually-weighted quantization of the DCT are key parts of the MPEG-2 coding system. Quantization is basically a process for reduction the precision

MPEG-2 Test Stream with Static Test Patterns in DTV System 377

of the DCT coecients. It is desirable to represent the coecients for highspatial frequencies with less precision, which is considered the spatial frequencyresponse of the HVS. However, in case of quantizing with default weighting table,if the spatial frequencies become higher and higher then quantization errorsincreased, which is the reason of degradation of picture quality and undesirableresult. The new user-dened quantization is proposed to solve this problem. Theproposed weighting table values are set by the minimum value so that there isno degradation of picture quality during quantization process, as shown in Fig.2. Moreover, proposed weighting table is transmitted to one part of sequenceheader and is available without change of decoder.

Proposed ZeroStuffing Decision

Bit Rate Control

Source Coder

ProposedWeighting Table

Target Bits

Picture

SignalActivity

Scaling Factor

ActualCoding Bits

Buffer

Buffer Status

Zero Stuffing

Bit Stream

Fig. 1. Block diagram for video compression by proposed algorithm

8 16 19 22 26 27 29 3416 16 22 24 27 29 34 3719 22 26 27 29 34 34 3822 22 26 27 29 34 37 4022 26 27 29 32 35 40 4826 27 29 32 35 40 48 5826 27 29 34 38 46 56 6927 29 36 38 46 56 69 83

(a)

8 8 8 8 8 8 8 8

8 8 8 8 8 8 8 8

8 8 8 8 8 8 8 8

8 8 8 8 8 8 8 88 8 8 8 8 8 8 8

8 8 8 8 8 8 8 8

8 8 8 8 8 8 8 8

8 8 8 8 8 8 8 8

(b)

Fig. 2. (a) Default and (b) proposed quantization weighting table


2.2 Proposed Zero Stung Algorithm

We propose a new algorithm that determines the number of zero stung bitsfor preventing VBV buer overow and maintaining the high picture quality atthe same time. The proposed method is shown in Fig. 3.

D = T - S

level Setting,Calculation QP and n

Next Video BufferStatus Prediction

n = n + 1( 0 < n < level )

Z bits Zero Stuffing

Target Bits andActual Coding Bits Check

No

Yes

Z = Rstuffing D

Rstuffing =

n

level

Buffer Overflow ?

Fig. 3. Flowchart of the proposed zero stung algorithm

For this operation, we can exploit the relationship between the number ofactual coding bits and the number of target bits. The dierence Di,p,b can bedetermined by the following relation.

Di,p,b = Ti,p,b Si,p,b (1)where T and S are target number of bits and actual coding bits respectively. Andsubscription i, p, and b correspond to a picture type I, P, and B respectively. Ifthe target number of bits and the number of actual coding bits for each picturedo not match well, then the degradation of picture quality and buer overowmay occur at the end of the GOP. However, if the number of zero stung bitsis allocated as much as Di,p,b to prevent buer overow, then the remaining bit


resource in the next GOP is getting smaller. As a result, insucient coding bitsare allocated to the pictures of the next GOP, and the severe degradation ofpicture quality may occur. Therefore we propose a formulation that determinesthe number of zero stung bits Zi,p,b and the initial buer fullness d

i,p,b0

Zi,p,b = Rstuffing Di,p,b (2)and

di,p,b0 = di,p,b0 + (Si,p,b + Zi,p,b Ti,p,b) (3)

respectively. The initial buer fullness is updated per each picture. The ratio ofzero stung Rstuffing is

Rstuffing =1

level n (4)

where level is set by the complexity of encoder, and n is calculated as follows:

1 < n < , n : interger (5)

=QP QPmin

QPmax QPmin level (6)

where the QP is the PSNR of current picture, the QPmin and the QPmax, theminimum and the maximum PSNR, are updated or maintained by PSNR ofcurrent picture, QP . We set the initial QPmin and QPmax to 20dB and 90dBrespectively. The setting up algorithm of QPmin and QPmax is shown in Fig. 4.If buer overow is predicted at the next picture, then we increase the numberof zero stung bits through updating the Rstuffing and we can prevent bueroverow. Fig. 5 illustrates relationship between n and QP .

Picture Quality Check

Q P max < Q P

QP max = Q PR

stuffing = 1

YesQP min < Q P < Q Pmax

0 < Rstuffing < 1

No

Yes

No

QP min = Q PR

stuffing = 0

Fig. 4. The setting up algorithm of QPmin and QPmax


n

Q P min Q P max0

level

level1

The PSNR of current picture, Q p

1

2

3

4

5

Fig. 5. Relationship between n and QP

3 Experiments and Results

We have tested the performance of proposed algorithm with two static testpatterns, multiburst and crosshatch as shown in Fig. 6. They are encoded byMPEG-2 encoder with and without the proposed algorithm. Bit rate, VBV buersize, and level are 12 Mbps, 9.78 Mbits, and 9, respectively. And TM5 is usedfor rate control.

(a) (b)

Fig. 6. (a) Multiburst (1280 720) and (b) crosshatch (1920 1080)

Fig. 7(a) and Fig. 8(a) show the PSNR of the test streams. From thosegures, we can see that the proposed method is about 4.5 dB more ecientthan the conventional cases. Fig. 7(b) and Fig. 8(b) plot the VBV buer state


Frame number0 10 20 30 40 50 60 70 80

PSN

R [d

B]

54

56

58

60

62

64

66

ProposedAhn et al.MPEG-2 TM5

(a)

Frame number0 10 20 30 40 50 60 70 80

PSN

R [d

B]

54

56

58

60

62

64

66


(b)

Fig. 7. (a) PSNR and (b) VBV buer state of multiburst pattern encoded

Frame number0 10 20 30 40 50 60 70 80

PSN

R [d

B]

56

58

60

62

64

66

68

70


(a)

Frame number0 10 20 30 40 50 60 70 80

VBV

buf

fer o

ccu

panc

y, [M

bps]

0.0

2.0e+6

4.0e+6

6.0e+6

8.0e+6

1.0e+7

1.2e+7

1.4e+7

1.6e+7

1.8e+7


(b)

Fig. 8. (a) PSNR and (b) VBV buer state of crosshatch pattern encoded

of the test streams. In both gures, the proposed test stream has a very stablebit rate, while large uctuation or overow is occurred in the others. Therefore,we have not experienced any buer overow with the proposed algorithm whilehigh picture quality is to be maintained. The average PSNR of proposed andconventional methods are shown in Table 1.

4 Conclusions

High quality test and measurement in the MPEG domain is required to evaluatethe static picture quality of DTV receiver. Therefore, high quality test stream isessential if stable bit rate is to be maintained. This paper proposes a new methodfor generating a high quality test stream to evaluate the static picture quality ofDTV receiver. Moreover the proposed method is suitable to compression of thestatic test pattern. In order to obtain the test stream with good picture quality,we propose a new user-dened quantization for minimizing the quantization


Table 1. The average PSNR of the proposed method and the conventional methods

PSNR[dB]Video Rate[Mbps]Test Images

Proposed Ahn et al. [5] MPEG-2 TM5

Multiburst

Crosshatch

12

12

62.8

65.4

58.7

58.6

58.6

62.7

error and an adaptive zero stung algorithm that determines the number ofzero stung bits for preventing buer overow and maintaining good picturequality at the same time. To evaluate the proposed method, we generate thetest pattern streams by using MPEG-2 encoder with and without the proposedalgorithm. Experimental results show that the proposed test pattern stream hasa stable bit rate during the decoding process, and the PSNR of the proposedmethod is about 4 dB higher than that of the conventional cases. The proposedmethod has a stable bit rate and good picture quality and it is suitable forevaluation a DTV receiver.

References

[1] W. Sohn and J. H. Kim: System Test of Digital DBS System for Video and AudioSignals. IEEE Transactions on Broadcasting, Vol. 45, No. 2, (1999) 187-191

[2] K. D. McCann: Testing and Conformance Checking in the Digital Television En-vironment. IEE International Broadcasting Convention, No. 428, (1996) 331-336

[3] A. N. Rau: Automated Test System for Digital TV Receivers. IEEE InternationalConference on Consumer Electronics, (2000) 228-229

[4] C. M. Kim, B. U. Lee, and R. H. Park: Design of MPEG-2 Video Test Bitstreams.IEEE Transactions on Consumer Electronics, Vol. 45, No. 4, (1999) 1213-1220

[5] C. T. Ahn and H. S. Chang: A Parallel Processing Architecture for HDTV En-coding System. International Workshop on HDTV96, (1996) 162-168

[6] J. W. Lee and Y. S. Ho: Target Bit Matching for MPEG-2 Video Rate Con-trol. IEEE Region 10 International Conference on Global Connectivity in Energy,Computer, Communication, and Control, Vol. 1, (1998) 66-69

[7] ISO/IEC JTC1/SC29/WG11. Test Model 5, Draft, (1993)

Front matterChapter 1IntroductionOutline of the Segmentation SchemeSelection of the Localization ScaleScale Description FunctionScale Selection Approaches

Experimental Results and Conclusion

Chapter 2IntroductionThe Neugebauer Color Prediction Model and the Experimental ProcedureSegmentation Algorithms to Estimate Dot AreaObtaining the ``Groundtruth" SegmentationsResults and DiscussionSegmentation ResultsAssessing the Neugebauer Model

Conclusion

Chapter 3IntroductionPyramidal RepresentationSpatial Multiresolution ConstraintExperimentsConclusion

Chapter 4IntroductionConstruction of Color Component and Texture ComponentColor Space Sampling Based on Fibonacci LatticeConstruction of Texture Component by Fuzzy HomogeneityConstruction of Color Component by Peer Group Filtering

Joint Color-Texture Segmentation Based on Mean ShiftMean Shift Based ClusteringJoint Color-Texture Segmentation

Experimental ResultsConclusionsReferences

Chapter 5IntroductionContourlet Domain Hidden Markov Trees ModelContourlet ApproximationContourlet Domain Hidden Markov Trees Model

Unsupervised Image Segmentation Using Contourlet Domain Hidden Markov Trees ModelSupervised Bayesian SegmentationUnsupervised Image Segmentation Using Likelihood Disparity

Simulation Results and AnalysisConclusion and DiscussionReferences

Chapter 6IntroductionClassical C-V MethodColor C-V MethodA Novel Color C-V MethodChoice of Color Space

Applications and Experimental ResultsConclusionReferences

Chapter 7IntroductionWavelet Energy Measures for Texture FeatureKernel SFCM for SAR Image SegmentationExperiments and AnalysisConclusionReferences

Chapter 8IntroductionImage PreprocessingImage BinarizationContour ExtractionColumns SegmentationDiscussion and Conclusion

Chapter 9IntroductionMathematical PreliminariesSome Simple Examples Tailored to the Registration Problem

Proposed Methods to Improve ROI RegistrationMethod 1: Weighted Mutual InformationMethod 2: Mutual Information of Weighted Distributions

Results of an ROI-Based Registration ExperimentSummary and Concluding Remarks

Chapter 10IntroductionThe Denoising MethodThe Complex Wavelet TransformThe a Prior ModelsExperimental Results and Discussions

Chapter 11IntroductionAn Appropriate Fuzzy MetricComputational Analysis

Image FilteringClassical Vector Median Filter \cite{Astola,Plataniotis}Proposed Vector Median Filter

Experimental ResultsAdjusting the K ParameterComparing Performances

Conclusions

Chapter 12IntroductionShrinkage Approaches for DenoisingMethod 1Method 2Method 3

Experimental Analysis and EvaluationConclusion

Chapter 13IntroductionTime Oriented Video SummarizationTOVS Based on Key FramesTOVS Based on Idea Motion CompensationTOVS Based on Maximum Decoding Length

Simulation ResultsConclusion

Chapter 14IntroductionRelated WorkEstimation of Irradiance and ReflectanceImplementation of Shadow RemovalComputation of $K_d$ The Coarse Shadow Matte ExtractionSmoothing the Coarse Shadow Matte in Gradient Domain

Conclusion

Chapter 15IntroductionTheoryCorrelation CoefficientWeighted Sum of Squared DifferencesIntensity RemappingAlgorithmic Complexity

MethodsResults and DiscussionConclusions and Future Work

Chapter 16IntroductionIsotropic Blur Identification2D Isotropic PSF ModelDetection of Feasible EdgesLeast Squares PSF Estimation

Experimental ResultsConclusions

Chapter 17IntroductionPiecewise Constant ApproximationROF Model with Gradient TermROF Model with High Order DerivativeLinear Approximations of MS and ROF ModelsExperimental ResultsConclusions

Chapter 18IntroductionLMV Estimation by Phase CorrelationPhase CorrelationLMV Estimation

FMV Prediction Using Adaptive Kalman FilterExperimental ResultsConclusionsReferences

Chapter 19Introduction Wavelet Thresholding for Image Denoising Conventional Thresholding OperatorsStandard Wavelet Thresholding MethodsRemarks

Localized Wavelet Thresholding OperatorsEnhancement Using Cycle SpinningExperimental ResultsBefore Cycle SpinningAfter Cycle SpinningAdditional Experimental Results

Concluding Remarks

Chapter 20IntroductionFuzzy Image DefinitionFuzzy HyperbolizationLocally Adaptive Image EnhancementType-2 Fuzzy SetsProposed TechniqueExperimental ResultsConclusionReferences

Chapter 21IntroductionProposed FrameworkTransition Boundary InitializationIllumination Artifact RemovalDetermining Transition TypeGradual Transition Boundary Determination

System Evaluation ResultsConclusions

Chapter 22IntroductionImage AcquisitionEdge Detection as a Focusing MetricThe Proposed Method AlgorithmVariable ThresholdingImage Subdivision and Edge SummationDetecting In-focus RegionsWeighted Voting and Final Pixel SelectionCreating an All-in-Focus Image

ConclusionsReferences

Chapter 23IntroductionForm Image DecompositionConstruction of Non-orthogonal Wavelet with Adjustable Rectangle SupportWavelet Decomposition of Skew Form Document Image

Determination of the Skew AngleExperiments and Conclusions

Chapter 24IntroductionRelated WorkProposed SystemResultsConclusions

Chapter 25IntroductionReflection ModelHighlight DetectionThe Best Fit WindowDiffuse Chromaticity Estimation

Non-white IlluminantExperimentsConclusionsAcknowledgementsReferences

Chapter 26IntroductionProposed Scrambling MethodsMotion Vector RelocationSlice Relocation

Experimental ResultsConclusion

Chapter 27IntroductionTheory Foundation: Butterworth High Pass FilterUsing Weighted Information Entropy to Estimate the Complex Degree of Infrared Images' BackgoundsA New IdeaDiscussion of the Validity

Adaptively Detect the Small Target in the Video SequencesExperiments and ResultsConclusionsReferences

Chapter 28IntroductionImpulse Noise Detection and RemovingImpulse DetectionNoise Filtering

Simulation ResultsConclusionReferences

Chapter 29IntroductionBasic Concepts of Rough SetsShot Boundary Detection Scheme Based on RS and FCMFuzzy $c$-Mean AlgorithmFeature Selection Based on RS and FCMRules Generation


Chapter 30IntroductionImage Enhancement via Fusion Based on the DFB and LPLaplacian Pyramid (LP) DecompositionA Construction Method for DFBThe Image Fusion Algorithm Based on LPDFB

Numerical Experiments and AnalysisThe Evaluation Standard of the Effect of Image FusionNumerical ExperimentsExperiments Analysis

ConclusionReferences

Chapter 31IntroductionMaterials and MethodsSimulated ImageMeasurement ParametersAnalysis of Noisy Complex ImagesDe-noising AlgorithmsThreshold Calculation

ResultsDiscussion and ConclusionsReferences

Chapter 32IntroductionRelated WorkImage Quality Measure (IQM)Analysis of IQM

Image Evaluation FactorsThe Contour-Volume FactorThe Noise-Rate FactorThe Uniform Intensity-Distribution Factor

Computational EfficiencyConclusionAcknowledgmentsReferences

Chapter 33IntroductionDual Ridgelet Frame and Dual Monoscale Ridgelet FrameImage Denosing ApplicationConclusionReferences

Chapter 34IntroductionIntermediate Frame Based Multiple DescriptionsProblem of Selection of Description That Best Suits Given ChannelResults and DiscussionConclusionReferences

Chapter 35IntroductionDocument FeaturesThe New AlgorithmFile Format ConversionRegion SplittingAxes Drawing and ScanningMarginal Region DefinitionBorder Detection

Results ObtainedConclusions and Lines for Further WorkAcknowledgementsReferences

Chapter 36IntroductionSystem Configuration and Image DistortionDistortion Correction by Using a Distance Ratio FunctionIntensity Distortion CorrectionShape Distortion Correction

Distortion Correction by Using a General Polynomial ModelIntensity Distortion CorrectionShape Distortion Correction

Comparison of Distortion Correction PerformancesConclusionsReferences

Chapter 37IntroductionState of the Art in Video Watermark on Motion VectorsProposed Video Watermark SchemeThe Principle for Watermark EmbeddingWatermark Detection Approach

Experimental ResultsConclusionsAcknowledgementsReferences

Chapter 38IntroductionPreliminariesAn Asymmetric Lossless Compression SchemeExperimental ResultsConclusionsReferences

Chapter 39IntroductionDetecting SymmetriesA New AlgorithmResults ObtainedAlgorithm LimitationsConclusionsAcknowledgementsReferences

Chapter 40IntroductionConventional Video Mixing MethodsProposed Hybrid Mixing TechniqueSimulation ResultsComparison Between Pixel-Domain and Bitstream-Domain Transcoder MethodPerformance Comparison of Various Mixing Methods


Chapter 41IntroductionThe GS-2D-LZ SchemeSearch and Encoding StrategiesData Structures Defined


Chapter 42IntroductionPrincipal Component AnalysisCalculating EigenfacesRecognition

UEP Using Convolutional Codes for PCA Coded ImagesUEP of Projection Coefficients Using Convolutional Codes

Results and DiscussionFace Recognition of 80 Received Coded Images (Projection Vectors)


Chapter 43IntroductionFramework of RNG in Crypto ModuleThe Filter Model of H/W Random Number GeneratorFilter Algorithm in Conventional ModelProposed Filter Algorithm for Tree Model


Chapter 44IntroductionThe Layer Based Multiple Description CodingThe Framework of the Layer Based Multiple Description CodingThe Layer Based Multiple Description Subband CodingThe Simulation Results by Layer Based Multiple Description Subband Coding and Discussion

The Layer Based Multiple Description Packetized CodingThe Realization of the Layer Based Multiple Description Packetized Coding FrameworkExperimental Results by the Layer Based Multiple Description Packetized Coding

Concluding RemarksAcknowledgementsReferences

Chapter 45IntroductionH.264/AVC-Based FGS CodingVideo Bitstream SwitchingCombination of FGS Coding and Video Bitstream SwitchingExperimental ResultsConclusionsAcknowledgementReferences

Chapter 46IntroductionUMHexagonS Motion Estimation Algorithm [6]Imagine Stream Processor [8]Implementation of H.264 Motion EstimationResult AnalysisConclusionReferences

Chapter 47IntroductionThe Proposed AlgorithmProposed User-Defined QuantizationProposed Zero Stuffing Algorithm

Experiments and ResultsConclusionsReferences

Chapter 48IntroductionSpeed Optimization TechniquesConverting YUV to RGB565 Using Multiplierless Integer TransformMemory Access Optimizing


Chapter 49IntroductionImage FeaturesMatching CandidateStables Marriages MatchingEliminating OutliersGlobal Image TransformationExperimental ResultsConclusion

Chapter 50IntroductionPerceptual GroupingThe Proposed ApproachObject GroupingEnvelope Detection


Chapter 51IntroductionOur Approach and Related Work

Deformation ModelPrototype Template RepresentationParametric TransformationsProbabilistic Constraints

Objective FunctionExperimental ResultsConclusions and Future Work

Chapter 52IntroductionMulti-scale Triangle-Area Representation (MTAR)TAR SignaturesMTAR ImagesMaxima and Minima Points

MatchingExperimental ResultsConclusions

Chapter 53IntroductionFeatures ExtractionTrack-AbilityEntropyDCT Edge Density

Segmentation MethodExperimental ResultsExample 1Example 2 and 3Verification of RMA


Chapter 54IntroductionShape Representation Based on Beam Angle Statistics (BAS)Dynamic Warping with PenaltyCyclic Sequence ComparisonExperimentsConclusionReferences

Chapter 55IntroductionMethodsVelocity FieldsImplementationExperiments

ResultsDiscussion and Conclusions

Chapter 56IntroductionMethodSimilarity IndexHierarchical Classification of Digital Images

Index Performance with Synthetic DataEvaluation of the Index Performance

Results with Image DataImage Classification and ClusteringAnalysis

Conclusions

Chapter 57Motivation and Related WorkCurvature Features for Face DetectionExperimentsSemi Synthetic DataReal DataDiscussion

Conclusion

Chapter 58IntroductionMorphological Modeling of Object Planar ShapesLocalization of Skeleton VerticesPiecewise-Linear Skeletonization by the Vertex Growing AlgorithmLocally-Adaptive Segmentation of Object RegionsExperimental ResultsConclusionReferences

Chapter 59IntroductionObject OntologyOntology-Based Object RecognitionExperimental ResultsConclusionAcknowledgmentsReferences

Chapter 60IntroductionStatistical Object ModelFeature VectorsObject AreaDensity Functions of the Feature Vectors

Localization and ClassificationObject Density ValueRecognition Algorithm

Experiments and ResultsImage Data BaseLocalization and Classification Rates

Conclusions

Chapter 61IntroductionThe Dual-Tree Complex Wavelet TransformThe Inter-coefficient ProductDetermination of Feature Orientation from Neighbour CoefficientsFeature Orientation Calculations for All SubbandsDefinition of ICP

Results and InterpretationConclusions

Chapter 62Introduction and MotivationDescription of the Computer Vision SystemThe Use of Circle Detection on EllipsesFormation of an Ellipse Based on CirclesThe Hough TransformDetecting Circles in Ellipsoidal Traces

Results Obtained Using a Real CylinderConclusionsReferences

Chapter 63IntroductionSampled Active Contour ModelPerformance of S-ACMThe Problem of Original S-ACMProposal MethodComparison

Automatic Lip DetectionBase Point DetectionSet Up ROIBinalizationApply S-ACMFor Other Frame

Lip ReadingDetect Speech PeriodVisual FeaturesRecognition

ExperimentsConclusion

Chapter 64IntroductionImage Normalization by Piecewise Affine MappingSubspace Method for Mouth ClassificationLDA for Mouth ClassificationExperimental ResultsConclusion

Chapter 65IntroductionHahn Polynomials and MomentsHahn PolynomialsHahn Moments of ImageComputational Aspects


Chapter 66IntroductionClustering Gabor Filtering Results of Object ImagesGabor Filtering and Gabor Energy MapClustering Gabor Feature Vectors

Classification of Artificial / Natural Object ImagesSum of Sector Power DifferenceEnergy of Edge Direction HistogramClassification of Object Images

Experimental Results and DiscussionsConclusionsReferences

Chapter 67IntroductionIndividual Code ExtractionCode Sequence Block ExtractionIndividual Code ExtractionPicture Area Extraction

Enhanced Fuzzy RBF Network for Recognition of PassportsPerformance EvaluationConclusionReferences

Chapter 68IntroductionContour Trees for Continuous Scalar FieldsRegion-Based Contour Trees from Digital ImagesDefinition of Isosurfaces in Digital ImagesDefinition of the Transitions of Isosurfaces Region-Based Contour TreesIntroducing the Isosurfaces Surrounding Whole ImagesConstruction of Region-Based Contour TreesExperimental Result

ApplicationsSelection of IsosurfacesImage SegmentationImage Filtering

Conclusion

Chapter 69Introduction and BackgroundThe Proposed AlgorithmExperimental ResultsSummary and Conclusions

Chapter 70IntroductionPerceptual Features for CBIRBase GET ClassPredominant GETParallel GET PairJoint GET

Image Content Representation and Similarity MeasureFeature Histogram RepresentationSimilarity Measure

Experiments and EvaluationConclusions

Chapter 71IntroductionSystem OverviewVideo SegmentationMotion Region DetectionFuzzy Motion Region Classification SystemIndexing and RetrievalResultsConclusionReferences

Chapter 72IntroductionProposed ApproachMIL-Based SVMSGlobal-Feature-Based SVMsFusion Approach

Experimental ResultsCategorization ResultsValidation of the Proposed MethodSensitivity to the Number of Categories


Chapter 73IntroductionModified Edge-Based GFDIgnoring Textured Regions Using ETCExperimental ResultsConclusionsReferences

Chapter 74IntroductionParameter Estimation for Statistical Distance MeasuresInteractive Retrieval with Relevance FeedbackExperimental Setup and ResultsConclusion

Chapter 75IntroductionBackgroundThe MethodEnergy MinimizationExperimental ResultsConclusions

Chapter 76IntroductionBrief Overview of Registration in 3D ScanningOverview of Profile Registration

Local Optimisation for 3D Profile RegistrationProfile Matching and Point CorrespondenceLocal Optimisation Methodology

The 3D Planar Profile Registration AlgorithmResults and DiscussionReferences

Chapter 77IntroductionExtraction of the Edge-Direction DistributionText Rotation in 3D and Transform Model for the EDDText-Pose Estimation MethodResultsConclusions

Chapter 78IntroductionSystem DescriptionCalibrationReconstructionGeometrical ReconstructionSpectral Reconstruction

Results and DiscussionConclusion

Chapter 79IntroductionStereo Matching AlgorithmExperimental ResultsConclusionAcknowledgmentsReferences

Chapter 80IntroductionRelated Geometrical ConceptsThe Voronoi Diagram and the Delaunay TriangulationEuclidean \alpha-Shapes

Anisotropic ConceptsGlobal AnisotropyThe Anisotropic Diagrams and Triangulations Based on the Distance $d_Q$The Anisotropic \alpha-Shapes Based on the Distance $d_Q$

Detection of StructuresDetection of Structures in a Given DirectionRemoving Non-significant StructuresDetection of Polyhedral Structures

Conclusion

Chapter 81IntroductionMorphological Edge DetectorThresholding AlgorithmMorphological Gradient Histogram AnalysisProposed Algorithm and Computation Complexity

ExperimentsConclusionsReferences

Chapter 82IntroductionBasic NotionsModelling of ImagesFuzzy SetsBinary Morphology, Grey-Scale Morphology Based on the Threshold Approach and on the Umbra ApproachFuzzy Mathematical MorphologyColour Morphology

New Vector Ordering in the RGB Colour SpaceDefinition of New Maximum and MinimumNew Vector Morphological Operators for Colour ImagesExperimental ResultsConclusion

Chapter 83IntroductionPreliminariesNotations on 3D Euclidean GeometryNotations on 3D Digital Geometry and Definition of Digital Convex Polyhedron

Decompositions of 3D Digital Convex PolyhedronsDecomposition Condition of Digital Convex Polyhedron

Decomposition of Convex Structuring Element into Neighborhood Structuring ElementsDecomposition of Convex Structuring Elements into a Set of BasesNeighborhood Decomposition of 3D Convex Structuring Elements and Cost FunctionDecomposition Examples


Chapter 84IntroductionVector Median Based FiltersProposed Filtering DesignExperimental ResultsConclusion

Chapter 85IntroductionNonparametricsApplication to Color IndexingColor Space Embedding

ExperimentationConclusions

Chapter 86IntroductionDemosaicing Using Taylor Series ExtrapolationStage 1 - Green Plane ExtrapolationStage 2 - Classifier


Chapter 87IntroductionConventional Colorimetric Characterization of Digital CameraProposed Colorimetric Characterization of Digital CameraColorimetric Characterization of Ideal Color CameraProposed Adaptive Colorimetric Characterization of Digital Camera with the White Balance

Experiments and ResultsConclusions

Chapter 88IntroductionThe Color Constancy AlgorithmLikelihood FunctionColor CoordinatesColor Change ModelMapping EstimationMapping Selection

Experiments and ResultsConclusions

Chapter 89IntroductionSkin-Color Models2D Gaussian Model in RGB Color Space (RG-GM)2D Gaussian Model in HSV Color Space (HS-GM)1D Lookup Table Based on Hue Histogram (H-LT)Adaptive Hue Lookup Table Model (AH-LT)


Chapter 90IntroductionGabor-Like Hermite Filter BankCartesian Hermite FiltersKrawtchouk FiltersSteered Hermite FiltersGabor-Like Hermite Filters

Texture Feature Extraction and IndexingExperimental Results on Textured ImagesHandwriting Document IndexingSimilarity ComputationPractical Results and Evaluation on Handwriting Documents


Chapter 91IntroductionRotation Invariant Feature ExtractionGabor Function Based Multi-channel Filtering ModelRotation Invariant Features

Feature Extraction via Steerable Gabor FiltersSteerable FilterSteerable Approximations of Gabor FiltersFast Feature Extraction Using Basis Functions

Experimental ResultsConclusionReferences

Chapter 92IntroductionMultiresolution HistogramsSupport Vector MachinesLearning SVMs with Multiresolution HistogramsExperimentsDataExperimental SetupExperimental Results

Conclusion

Chapter 93IntroductionFeatures ExtractionMomentsFractal Dimension

Support Vector Machines (SVMs)ExperimentsConclusionsReferences

Chapter 94IntroductionTheoryImplementationExperimentsConclusions

Chapter 95IntroductionPrevious WorkOur MethodMarkovian Segmentation$K$-Nearest-Neighbor-Based FusionExperimental ResultsDiscussion

Chapter 96IntroductionOutline of the AlgorithmBackground Subtraction by MoGOptical Flow ComputationExperimental ResultsConclusionsReferences

Chapter 97IntroductionRelated WorkOverview

Video-Based AnalysisAutomatic Projective RectificationWater Surface Normal

Non-photorealistic Rendering ApplicationResultsConclusion and Future Work

Chapter 98IntroductionResearch BackgroundDesign DiscussionUnderlying AssumptionsImage GrabbingTraining and Calculating Initial Thresholds

The AlgorithmResultsReferences

Chapter 99IntroductionProposed AlgorithmDecision for Illumination ChangeThe Extraction ModeDouble-ThresholdingVOP Extraction


Chapter 100IntroductionMoving Object Detection and Shadow EliminationTrackingData AssociationSegment Generation and Trajectory Maintenance


Chapter 101IntroductionHead ModellingPerceptually Uniform Color SpaceHead RepresentationElliptic FittingEllipse Modelling

Head TrackingExperimental ResultsConclusions

Chapter 102Introduction and BackgroundOverview of the ApproachThe Proposed Tracking ModelThe Object ModelVelocity Estimation ModelVelocity Measurement Model

Object Parameter Updating and Occlusion ReasoningExperimental ResultsConclusions

Chapter 103Introduction3D Arm Tracking Using a Particle FilterSingular Movement Detection and Tracking Recovery Through Physically Invalid SamplesExperimental ResultsConclusion

Chapter 104IntroductionPredictive Estimation to Track Multi-targetsTrack InitializationPredictive Estimation Using Occlusion Activity DetectionData Association Between Moving Blobs and Real Targets


Chapter 105IntroductionProposed Model-Based Cell TrackingCell Image ModelProbability of Cell Boundary $P_cb$Probability of Cell Interior $P_ic$Probability of Uniformity of Cell Boundary $P_cdf$HSC Tracking

ResultsConclusions and Discussions

Chapter 106IntroductionThe Proposed SolutionThe Pre-processing StageThe Fuzzy Region Growing StageThe Threshold Selection StageThe Boundary Extraction Stage

ResultsConclusionsAcknowledgementsReferences

Chapter 107IntroductioncDNA Microarray Imaging BasicsRoot Signals Based Segmentation FrameworkVector Median FilterRoot Signals


Chapter 108IntroductionAdaptive Ellipse MethodParameter EstimationDiagonalizationComputing the Radius


Chapter 109IntroductionPreprocessingPixel-Level Exudate RecognitionApplication of Pixel Level Exudate Recognition on Whole Retinal ImageConclusionsReferences

Chapter 110IntroductionROI IdentificationTexture Feature ConstructionSecond Order StatisticsHigher Order Statistics

Feature SelectionParticle Swarm Optimization (PSO)ANT Colony Optimization

ClassificationResults and DiscussionConclusionsReferences

Chapter 111IntroductionEntropyComputing Conditional EntropyData DescriptionExperimentsExperimental Survey

Results and DiscussionConclusion

Chapter 112IntroductionHepatic Tumor SegmentationLiver SegmentationVessel EliminationComposite Hypotheses

Experiments and AnalysisConclusionsReferences

Chapter 113IntroductionBinary Mathematical Morphology(MM)Proposed ApproachThresholdingMorphological FilteringConditional ErosionInitial Snake/Seed Points Extraction

Experimental ResultsDiscussionConclusion

Chapter 114IntroductionMulti-scale Line-Structure SegmentationLine-Structure Detection with the Hessian MatrixModified Scale-Specific Line-Structure DetectionBayesian Identification of InstrumentsFinal Identification

Results and DiscussionConclusionsAcknowledgementsReferences

Chapter 115Introduction and MotivationsPrevious WorkActive Contour ModelsProposed Hybrid ModelSequential MCE ThresholdingChan-Vese Active Contour with Fixed Intensity Constants

ResultsConclusions

Chapter 116IntroductionMaterials and MethodsGeometrical ModelModel DeformationOriented Maps

Model RegistrationResultsConclusions

Chapter 117IntroductionMammogram ArtifactsStripe SuppressionWeighted Median FilterStripe ExtractionStripe Restoration

Experimental ResultsConclusionReferences

Chapter 118IntroductionDetection of Lanes and BandsBand CharacterizationBand Shape CharacterizationParameter Estimation

Feature Extraction and ClassificationConclusions and Future WorkReferences

Chapter 119IntroductionA Weighted Energy Maximization ApproachGridding Algorithms


Chapter 120IntroductionImpulsive Noise RemovalProposed Noise Detection AlgorithmSimulation ResultsConclusion

Chapter 121IntroductionMethodsPre-processingDetection of Potential Microcalcifications (Signals)Classification of Signals into Real MicrocalcificationsDetection of Microcalcification ClustersClassification of Microcalcification Clusters into Benigns and Malignants

ResultsConclusions and Future Work

Chapter 122IntroductionBackground: Level Set MethodMaterialsMethodsResults and Conclusions

Chapter 123IntroductionHistogram SamplingSVM-Based Wound Segmentation in the Color Feature SpaceComputation of Feature Vectors

Experiments and ResultsSingle Multi-dimensional Histogram Versus Multiple 1-D HistogramsComparison of Different Sampling TechniquesExamples of Wound Segmentation

Conclusion and Future WorkReferences

Chapter 124IntroductionFace Detection and Pose Estimation Using Geometrical RelationshipSynthesizing Deformable Face and Normalizing PosesMapping Input Image to Facial Model Using Extracted FeaturesTransforming Input Face to Frontal Face

Empirical AnalysisResults and Analysis


Chapter 125IntroductionYCgCr Color SpaceColor-Based SegmentationMorphological-Based SegmentationFace DetectionFeature ExtractionBest-Fit Ellipse Model


Chapter 126IntroductionFacial Feature ExtractionOrientation AssignmentGabor WaveletsLocal EntropyFacial Feature ModelsDetection Algorithm


Chapter 127IntroductionProposed Stereo Vision SystemDisparity Compensation of Stereo ImagesScaling of the Face Images According to the DistanceRange-Based Pose Estimation Using Optimized 3D Information

Pose Estimation and Face RecognitionExperimental ResultsConclusionsReferences

Chapter 128IntroductionSystem FrameworkGeneric LearnerMultiple Base Generic LearnersCombine Base Learners -- Level 1 Combination

Specific LearnerCombine Generic and Specific Learners -- Level 2 CombinationExperimentsExperiment SetupResults and Analysis

Conclusion

Chapter 129IntroductionDataAsymmetry in Frequency DomainThe Asymmetry Biometrics

Feature AnalysisDiscriminative Feature Sets

ResultsHuman IdentificationExpression ClassificationIllumination VariationsPhase-Only Images

Discussion

Chapter 130IntroductionFeatures and Feature Vector StructureAn `Ad Hoc' Distance Function (A, B) Experimental ResultsConclusions

Chapter 131Previous Work on Palmprint RecognitionOptimal Trade-Off Correlation Filter ClassifiersPalmprint Data SetPalmprint Verification Experiments and ResultsExperiment Set 1: Verification of 100 ClassesExperiment Set 2: Verification of 400 Classes

Conclusions

Chapter 132IntroductionAdvanced Correlation FiltersEyeglass Detection and Removal in Thermal ImageryPerformance ComparisonSummaryReferences

Chapter 133IntroductionIris SegmentationGabor Wavelet Iris EncodingAdvanced Correlation Filter DesignTesting and ResultsConclusionsReferences

Chapter 134IntroductionBackgroundFingerprint VerificationAttack PointsPrevious Approaches

Secure and Efficient Transmission of Fingerprint ImagesChallenge-Response ProtocolImage-Based Selective Encryption

Implementation Details and Performance EvaluationConclusionsAcknowledgementReferences

Chapter 135IntroductionDichotomy Model and Dichotomy TransformationFeature Extraction and Distance ComputationComparative Experimental ResultsReferences

Chapter 136IntroductionFacial Component DetectionFacial Region DetectionEye Region DetectionEyebrow Region DetectionMouth Region Detection

FCP ExtractionResults of FCP Extraction ExperimentConclusionReferences

Chapter 137IntroductionDatabase Based on the Dimensions of EmotionPCA Representations of Facial ExpressionsRecognition PerformanceDiscussionReferences

Chapter 138IntroductionRegion-Based Super-ResolutionFacial Feature ExtractionResults and DiscussionsConclusionsReferences

Chapter 139IntroductionFace TrackingSearch Region EstimationStochastic Skin Color ModelingLocating Face Region

Facial Feature TrackingEye TrackingLip Corners TrackingNostril Tracking

Results and DiscussionsConclusionsReferences

Chapter 140IntroductionFramework of Steganography ModelAnalysis and Related WorksUsing the Stego-key to Determine the Embedding PositionsScrambling the Embedded Message to Randomize the Hidden InformationScrambling the Cover Image

The Major ProblemThe Proposed Key Generation UnitBasic NotationsSecurity Attributes RequirementsThe Hidden Key-Agreement Protocol, Stego-KAThe Major Features of the Proposed Model


Chapter 141IntroductionPreliminariesAspect Ratio of the Recovered ImageThe Previous ARIVSS Schemes

The Proposed ARIVSS SchemesBasic ConceptEncoding AlgorithmThe Mapping Pattern $M_a,b$

ComparisonsConclusionReferences

Chapter 142IntroductionLiterature SurveyDetecting Steganographic ContentEmbeddings That Mimic Image StatisticsConclusions

Chapter 143IntroductionThe Basic VSS Scheme and EVSS SchemeThe VSS Scheme with Random Shadow ImagesThe EVSS Scheme with Black and White Shadow Images

The Proposed EVSS Scheme with High-Quality Shadow ImagesBasic ConceptAn H-EVSS Scheme with a General Access StructureImprove the Contrast of the Recovered SecretImprove the Contrast of the Shadow Image

The Contrasts of Recovered Image and Shadow Image for H-EVSS SchemesConclusionReferences

Chapter 144IntroductionSteganography and Steganalysis on Digital ImagesLSB EncodingVisual SteganalysisQuantitative Steganalysis

Proposed MethodResultsMessage EmbeddingVisual TestsRS Steganalysis

Conclusions

Chapter 145IntroductionRadar SystemsPreliminaries of Density FunctionsEstimation of Target Density FunctionSummary and Conclusion

Chapter 146IntroductionThe NUC Algorithm for Infrared Video SequencesScribner's Neural Network for NUCNUC Method with Ghosting Reduction Capabilities

Performance Evaluation with Real Infrared Image SequencesConclusions

Chapter 147IntroductionRelated Work

Performance Evaluation on Image Processing OperationsImage Processing Operations SelectedThe Implementations on CPU and GPUTest Conditions

Evaluation ResultsCPU vs. GPUEffects of Shading Languages and Profiles

Conclusions and Future WorkReferences

Chapter 148IntroductionProposed Background ModelCamera Motion EstimationBackground Image Updating

System ImplementationExperimental ResultsConclusions

Chapter 149IntroductionRelated WorksSystem ArchitectureAlgorithm ImprovementsExperimental ResultsConclusionsReferences

Chapter 150IntroductionProposed AlgorithmPerformance ComparisonImplementation ResultConclusionsReferences

Chapter 151IntroductionIterative Global Optimal Sampling (IGLOS)Problem FormulationIterative Approach for Solving Problem 1Optimal Sampling: Inlier Selection

ImplementationInitialization

Experimental ResultsSummary and Conclusions

Chapter 152IntroductionDigital Image Warping for Display Distortion CorrectionExperimental ResultsConclusionAcknowledgmentsReferences

Chapter 153IntroductionProposed Motion Estimation AlgorithmInitial Stage with Zero Motion DetectionPredictive Stage with Early Termination StrategyRefined Stage by Small Diamond Search

Complexity-Controllable SchemeExperimental ResultsConclusions

Back matter

image analysis and recognition, 2 conf., iciar 2005(lncs3656, springer, 2005)

Documents

springer science business

mitchellstanford university

kleinbergcornell university

vardirice university

iciar conference

german copyright law

computer engineeringwaterloo

data conversion