image analysis and recognition, 2 conf., iciar 2005(lncs3656, springer, 2005)
DESCRIPTION
Image recognitionTRANSCRIPT
-
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
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
-
378 S.-W. Jang et al.
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
-
MPEG-2 Test Stream with Static Test Patterns in DTV System 379
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
-
380 S.-W. Jang et al.
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
-
MPEG-2 Test Stream with Static Test Patterns in DTV System 381
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
ProposedAhn et al.MPEG-2 TM5
(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
ProposedAhn et al.MPEG-2 TM5
(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
ProposedAhn et al.MPEG-2 TM5
(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
-
382 S.-W. Jang et al.
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
Experimental ResultsConclusionsReferences
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
ConclusionReferences
Chapter 41IntroductionThe GS-2D-LZ SchemeSearch and Encoding StrategiesData Structures Defined
Experimental ResultsConclusionsReferences
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)
ConclusionReferences
Chapter 43IntroductionFramework of RNG in Crypto ModuleThe Filter Model of H/W Random Number GeneratorFilter Algorithm in Conventional ModelProposed Filter Algorithm for Tree Model
Experimental ResultsConclusionsReferences
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
Experimental ResultsConclusions
Chapter 49IntroductionImage FeaturesMatching CandidateStables Marriages MatchingEliminating OutliersGlobal Image TransformationExperimental ResultsConclusion
Chapter 50IntroductionPerceptual GroupingThe Proposed ApproachObject GroupingEnvelope Detection
Experimental ResultsConclusions
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
ConclusionsReferences
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
Experimental ResultsConclusions
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
ConclusionsReferences
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
ConclusionReferences
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
Experimental ResultsConclusion
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)
Experimental ResultsConclusion
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
ConclusionReferences
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
Experimental ResultsConclusions
Chapter 100IntroductionMoving Object Detection and Shadow EliminationTrackingData AssociationSegment Generation and Trajectory Maintenance
Experimental ResultsConclusion
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
Experimental ResultsConclusionsAcknowledgementsReferences
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
Experimental ResultsConclusion
Chapter 108IntroductionAdaptive Ellipse MethodParameter EstimationDiagonalizationComputing the Radius
Experimental ResultsConclusions
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
Experimental ResultsConclusion
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
ConclusionReferences
Chapter 125IntroductionYCgCr Color SpaceColor-Based SegmentationMorphological-Based SegmentationFace DetectionFeature ExtractionBest-Fit Ellipse Model
ConclusionsReferences
Chapter 126IntroductionFacial Feature ExtractionOrientation AssignmentGabor WaveletsLocal EntropyFacial Feature ModelsDetection Algorithm
Experimental ResultsConclusionsAcknowledgementsReferences
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
ConclusionReferences
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