CONCEALING TATTOOS Darijan Marčetić darijan.marcetic@fer.hr Faculty of EE and Computing
PRESENTATION TOPICS 1. Introduction 2. Tattoo identification 3. Tattoo de-identification 4. Conclusion Literature
1. INTRODUCTION Laws in democratic countries protect privacy Technology development resulted with governmental and corporate invasion of privacy Massive installation of surveillance equipment at public places Sensible personal information stored at corporate servers Private information can be easily misused Many corporate services like Google maps or Facebook are illegal because private information is not protected according to the law Development of Next Generation Identification (NGI) The aim of COST Action IC1206 is to develop de-identification methods for privacy protection in multimedia content as is required by the law and EU directives
1. INTRODUCTION Biggest companies with certificate for NGI ACABIO, Inc. ARH, Inc. Advanced Livescan Tecnologies, Inc. Antheus, Inc. Automation Designs & Solutions, Inc. Aware, Inc. BI2 Technologies, LLC BioEnable Technologies Pvt Ltd Biometrics4All, Inc. Biometrika srl BSI2000, Inc. Cogent Systems, Inc. Comnetix Computer systems, Inc. Computer Deductions, Inc. Corvus Integration, Inc. Cross Match Tehnologies, Inc. DataWorks Plus, LLC DBA Systems, Inc. Dermalog Identification Systems, Gmbh Digital Biometrics, Inc. DigitalPersonas, Inc. Exegenetics, Inc. FieldPrint, Inc. FingerMatrix, Inc. Fulcrum Biometrics, Inc. Futronic Technology Company, Ltd Green Bit S.p.A. Griaule Biometrics Heimann Biometrics Systems Hongda Opto-Electron Co, I Hunter Systems Group Hyndai Information Technology Company I/O Software, Inc. IAFIS Program Office IBIOS Private Ltd. IBIS Corporation ID Networks Identicator Identification International Identix, Inc. idsoftware, Inc. IISL, Ltd. ImageWare Systems, Inc. Imaging Technologies, Inc. Improvision Research Corporation Integrated Biometric Technology (IBT), Inc. Integrated Biometric Solutions (P) Ltd. (IBIOS) Integrated Biometric, LLC ITALDATA Ing. dell'idea ITouch Biometrics, LLC Jobin L-1 Identity Solutions Lockheed Martin Information Systems Lumidigm, Inc. M2SYS,LLC Mantra Softech (India) Pvt MaxID Corp MaxVision, LLC Mentalix, Inc Mobizent, LLC MORPHO, Inc. Motorola, Inc. National Background Check, Inc. NEC Solutions (America) NEC Technologies, Inc. Nitrogen Co. Ltd. North American Morpho North Grumman Corp. P.P.H.U. STANIMEX s.c. Papilon Biometrics Papilon Systems Ltd. PRC, Inc. PrideRock Holding Company, Inc. Printrak International, Inc. Sagem Defense Securite Sagem SA Sagem Securite SecuGen Corp. Secure Outcomes Inc. Shriraj Software Solutions (S3INDIA) Smartmatic Internationa Corporation Smiths Heimann Biometrics Gmbh Sonda, Inc. Spex Forensics, Inc. Starttek Engineering, Inc. Suprema, Inc. Titan Systems Corporation Tutis Technologies, Ltd. Ultra-Scan Corporation Union Community Co., Ltd. UPEK, Inc. Vertical Screen, Inc. Visionics Corporation Wiseassist Knowledge Solutions Pvt. Ltd. WYSE Biometrics Systems Pvt Ltd. Zerco Systems International, Inc. Zeum, Inc.
1. INTRODUCTION The main focus of this presentation is on tattoo de-identification Tattoos have been used for more than 5000 years as a way to express personal belief and as a confirmation of membership into some groups 14 % of all people in the USA, 32 % of people of age 25-29 and 38% of age 30-39 have at least one tattoo Systems for removing scars, marks and tattoos (SMT) do not exist at the present time in the scientific literature Google service Maps does blur faces but tattoos are not removed from other body regions (neck, chest, arms, legs), and therefore privacy is violated
1. INTRODUCTION Tattoos are indexed based on their position on the body by National Crime Information Centre (NCIC) into 31 main categories (hand, head, ) i 71 subcategories (forehead, right index finger, ) ANSI/NIST-ITL.1-2011 standard classifies tattoos into 8 classes and 70 subclasses
2. TATTOO IDENTIFICATION Systems for tattoo identification: Tattoo-ID, A.K. Jain at al. 2012. Michigan State University i MorphoTrak FASTID D. Manger 2012., Fraunhofer Institute for Optotronics, System Technologies and Image Explotation IOSB B. Heflin, W. Scheirer, T.E. Boult, 2012. Securics Inc i University of Colorado Springs S. T. Acton, A. Rossi 2008., Platinum Solutions, Univerity of Virginia
2. TATTOO IDENTIFICATION Tattoo-ID, A.K. Jain at al. Michigan State University i MorphoTrak Tatto localization is manuall Scale Invariant Feature Transform (SIFT) features are used, Lowe 2004. Identification is performed by matching SIFT features
2. TATTOO IDENTIFICATION FASTID D. Manger, Fraunhofer Institute for Optotronics, System Technologies and Image Explotation IOSB FAST and efficient international disaster victim IDentification (FASTID) Project http://www.interpol.int/projects/fastid (FP7/2007-2013) Tattoos are described with bag of words model Words are coded with SIFT features
2. TATTOO IDENTIFICATION Too large number of SIFT features Picture in the example has 1551
2. TATTOO IDENTIFICATION SIFT features are located only on parts of tattoos
2. TATTOO IDENTIFICATION SIFT features are not very reliable in real world conditions No correspondence in the example with just few pixels slightly changed in the query image Manual identification has superior quality compared to an automatic identification 12
3. TATTOO DE-IDENTIFICATION Proposed system for tattoo de-identification: Real time performance is critical for acceptability of the system ROI from the video stream is estimated based on movement, skin colour and texture analysis Resolution pyramid can be used to enhance performance The gaol is to conceal both location and appearance of a tattoo Automatic or manual identification should not be possible based on de-identified tattoo images
3. TATTOO DE-IDENTIFICATION Tattoo de-identification problems: Automatic unsupervised localization Finding tattoos and separation from skin Methods for tattoo canceling Feature selection Real time performance, hardware implementation Quality and naturalness estimation Distinguishing tattoos from clothes, jewelry,
3. TATTOO DE-IDENTIFICATION Commercial application of tattoo de-identification: Google streets blurs faces and license plates but identities of persons are not protected because tattoos are not removed Surveillance hardware should support de-identification in order to be in compliance with the law Information about tattoo location and appearance must be removed Retrieving tattoo appetence from de-identified tattoos must be supported in emergency situations related with criminal activity
4. CONCLUSION Tattoo de-identification and privacy protection: Tattoos can be used for person identification Privacy protection must include concealing tattoos No system for tattoo de-identification is recorded in scientific literature It is necessary to enhance existing methods and develop new ones for de-identifying tattoos in complex scenes in real time
LITERATURE Tattoo identification: A. K. Jain, J.-E. Lee and R. Jin, Tattoo-ID: Automatic Tattoo Image Retrieval for Suspect & Victim Identification, Proc. of Pacific-Rim Conference on Multimedia (PCM), pp. 256-265, Hong Kong, December 2007. J-E. Lee, A. K. Jain and R. Jin, Scars, Marks and Tattoos (SMT): Soft Biometric for Suspect and Victim Identification, in Proc. Biometric Symposium, BCC, September, 2008. A. K. Jain, J.-E. Lee, R. Jin, and N. Gregg, Content Based Image Retrieval: An Application to Tattoo Images, IEEE ICIP, Nov., 2009. A. K. Jain, R. Jin, and J.-E. Lee, Tattoo Image Matching and Retrieval, IEEE Computer, Vol. 45, No. 5, pp. 93-96, May, 2012. Heflin, B., Scheirer, W., & Boult, T. E. (2012). Detecting and classifying scars, marks, and tattoos found in the wild. Paper presented at the 2012 IEEE 5th International Conference on Biometrics: Theory, Applications and Systems, BTAS 2012, 31-38. Acton, S.T. & Rossi, A. 2008, "Matching and retrieval of tattoo images: Active contour CBIR and glocal image features", Proceedings of the IEEE Southwest Symposium on Image Analysis and Interpretation, pp. 21. D. Manger, 2012., Large-Scale Tattoo Image Retrieval, 2012 Ninth Conference on Computer and Robot Vision, pp. 454-459.