SPRINGER BRIEFS IN APPLIED SCIENCES AND TECHNOLOGY  COMPUTATIONAL INTELLIGENCE Kailash Jagannath Karande Sanjay Nilkanth Talbar Independent Component Analysis of Edge Information for Face Recognition SpringerBriefs in Applied Sciences and Technology Computational Intelligence Series Editor Janusz Kacprzyk For furthervolumes: http://www.springer.com/series/10618 Kailash Jagannath Karande Sanjay Nilkanth Talbar Independent Component Analysis of Edge Information for Face Recognition 123 Kailash Jagannath Karande SanjayNilkanth Talbar Electronics andTelecommunication Electronics andTelecommunication Engineering Engineering SKNSinhgad College ofEngineering ShriGGS InstituteofEngineering Pandharpur and Technology Maharashtra Nanded India Maharashtra India ISSN 2191-530X ISSN 2191-5318 (electronic) ISBN 978-81-322-1511-0 ISBN 978-81-322-1512-7 (eBook) DOI 10.1007/978-81-322-1512-7 SpringerNewDelhiHeidelbergNewYorkDordrechtLondon LibraryofCongressControlNumber:2013940094 (cid:2)TheAuthor(s)2014 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionor informationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purposeofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthe work. Duplication of this publication or parts thereof is permitted only under the provisions of theCopyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the CopyrightClearanceCenter.ViolationsareliabletoprosecutionundertherespectiveCopyrightLaw. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexempt fromtherelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. While the advice and information in this book are believed to be true and accurate at the date of publication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityfor anyerrorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,with respecttothematerialcontainedherein. Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Dedicated to my beloved daughter Kanishka - Kailash J. Karande Preface Anedgeinanimageisacontouracrosswhichthebrightnessoftheimagechanges abruptly. In image processing, an edge is often interpreted as one class of sin- gularities.Inafunction,singularitiescanbecharacterizedeasilyasdiscontinuities where the gradient approaches infinity. However, image data is discrete, so edges inanimageareoftendefinedasthelocalmaximaofthegradient.Thesedefinitions become the way to lead the research work in this book. Edge detection is an important taskinimageprocessing.Itisamaintoolinpatternrecognition,image segmentation, and scene analysis. An edge detector is basically a high pass filter that can be applied to extract the edge points in an image. Edge information plays vital role in many applications of image processing area.Tothebestofourknowledge,thereishardly anyreportedresearch workon facerecognitionusingedge information asfeatures forface recognitionwithICA algorithms. Here, we have used edge detection as a feature extraction method to extract edges information from facial images. The independent components are extracted from edge information. These independent components are used with classifierstomatchthefacialimagesforrecognitionpurpose.Inthisresearchwork we have explored Canny and LOG edge detectors as standard edge detection methods. Oriented Laplacian of Gaussian (OLOG) method is explored to extract the edge information with different orientations of Laplacian pyramid. Multiscale Wavelet model for edge detection is also proposed and explored in this research work to extract edge information. This book will give future direction for the PG studentsandresearchersintheareaofImageProcessingandPatternRecognition. vii Acknowledgments It is a privilege for me to have been associated with Prof. Sanjay N. Talbar, my guide, during myresearch workand writing ofthisbook. Itis with great pleasure thatIexpressmydeepsenseofgratitudetohimforhisvaluableguidance,constant encouragement, motivation, support, and patience throughout this research work. Hiscontinuousinspirationhelpedlotformypersonaldevelopmentandshapedmy career as a passionate teacher. IexpressmygratitudetoProf.M.N.Navale,President,SPSPMforhisconstant encouragement and strong support during the completion of this book. He is the main source of inspiration for me to broaden my thinking. I wish to express my deepest sense of gratitude to my parents and all family membersfortheirmoralsupportandblessings,whichenabledmetocompletethis task.MyheartfulthanksgotoNisha,mywifeforherpatienceandunderstanding, and to my daughter Kanishka who had to miss many affectionate hours during writing of this book. Finally, I would like to thank all those who have helped directly or indirectly during the writing of this book. Kailash J. Karande ix Contents 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Face Subspace. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Edge Detection as Feature Extraction Methods. . . . . . . . . . . . . 3 1.4 Technical Challenges Involved in Face Recognition . . . . . . . . . 4 1.5 Introduction to ICA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.6 ICA Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.7 FastICA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.8 Kernel ICA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.9 Sphering of Data Using KPCA . . . . . . . . . . . . . . . . . . . . . . . . 10 1.10 Further Processing Using ICA. . . . . . . . . . . . . . . . . . . . . . . . . 12 1.11 Distance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.12 Experimental Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.12.1 Face Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2 Canny Edge Detection for Face Recognition Using ICA. . . . . . . . . 21 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.1.1 Canny Edge Detection. . . . . . . . . . . . . . . . . . . . . . . . . 21 2.1.2 Preprocessing by PCA. . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 ICA Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 Results with Canny Edge Detector. . . . . . . . . . . . . . . . . . . . . . 25 3 Laplacian of Gaussian Edge Detection for Face Recognition Using ICA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.1.1 Laplacian Edge Detector . . . . . . . . . . . . . . . . . . . . . . . 35 3.1.2 Laplacian of Gaussian Edge Detector . . . . . . . . . . . . . . 36 3.1.3 Preprocessing by PCA. . . . . . . . . . . . . . . . . . . . . . . . . 37 3.2 ICA Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.3 Results with LOG Edge Detector . . . . . . . . . . . . . . . . . . . . . . 40 xi xii Contents 4 Oriented Laplacian of Gaussian Edge Detection for Face Recognition Using ICA. . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.1 Oriented Laplacian of Gaussian. . . . . . . . . . . . . . . . . . . . . . . . 49 4.2 Preprocessing by PCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.3 ICA Algorithms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.4 Results with OLOG Edge Detector . . . . . . . . . . . . . . . . . . . . . 53 5 Multiscale Wavelet-Based Edge Detection for Face Recognition Using ICA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.1 Multiscale Wavelet-Based Edge Detection . . . . . . . . . . . . . . . . 63 5.2 Continuous Wavelet Transforms . . . . . . . . . . . . . . . . . . . . . . . 64 5.3 Scale of a Wavelet Function. . . . . . . . . . . . . . . . . . . . . . . . . . 65 5.4 Preprocessing by PCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 5.4.1 ICA Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 5.5 Results with Multiscale Wavelet Edge Detector . . . . . . . . . . . . 68 6 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 About the Authors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Abbreviations and Symbols a(x,y) Edge Direction r Standard Deviation d(x,y) Distance Operator g(x,y) Local Gradient Cos Cosine Distance FastICA FastICA Algorithm FRT Face Recognition Techniques ICA Independent Component Analysis JADE Joint Approximate Diagonalization of Eigenmatrices KICA Kernel ICA Algorithm L1 Manhattan Distance L2 Euclidean Distance LEM Line Edge Map LOG Laplacian of Gaussian MAH Mahalanobis Distance OLOG Oriented Laplacian of Gaussian PCA Principle Component Analysis xiii