Description of Biometric Data Interchange Format Standards INCITS M1 Technical Editors: Creed Jones, Colin Soutar, Jim Cambier, Michael McCabe, Paul Griffin, Rod Beatson, Samir Tamer 1
Introduction International Committee for Information Technology Standards (INCITS) is the standards-developing body for US IT standards. M1 is the INCITS committee for biometrics, established November 2001. M1 serves as the US TAG (Technical Advisory Group) for ISO/IEC JTC1/SC37 Biometrics subcommittee, established June 2002. 2
M1 Program of Work data interchange formats common file formats application program interfaces application profiles performance testing and reporting 3
Motivation for Interchange Standards Exchange of biometric data in nonproprietary format among multiple vendors/applications Compile biometric databases for use in evaluating multiple algorithms Produce enrollment databases that enable re-enrollment using future algorithms or algorithm enhancements 4
Relationship to Other Standards Data compliant with interchange formats are embedded in CBEFF (NISTIR 6529-A) Biometric Data Block Certain of the formats reference ANSI/NIST-ITL 1-2000 Standard Data Format for the Interchange of Fingerprint, Facial, & Scar mark & Tattoo (SMT) Information Compliance assured with ANSI/INCITS 358-2002 standard (the BioAPI Specification) 5
M1.1 Task Group on Biometric Data Interchange Formats Finger Minutiae Format for Data Interchange Finger Pattern Based Interchange Format Iris Interchange Format Finger Image Based Interchange Format Face Recognition Format for Data Interchange Signature/Sign Image Based Interchange Format Hand Geometry Interchange Format 6
Project 1564-D - Finger Minutiae Format for Data Interchange - INCITS 378 Creed Jones Associate Professor of Computer Science, Seattle Pacific University Strategic Engineer, Sagem Morpho Inc., INCITS M1.1 Task Group Chairman 7
Finger Minutiae Format for Data Interchange Represents a fingerprint in terms of the ridge endings and splits these are called minutiae Standard contains: Precise definitions of minutiae Data Format Definition Agrees in concept with most law enforcement and AFIS fingerprint technologies Backward-compatible with AAMVA fingerprint format 8
Minutia Location Placement is defined carefully for compatibility with ANSI/NIST standard Bifurcation Ridge Endings 9
Finger Minutiae Data Format Allows more than one finger per record Allows more than one view per finger CBEFF-compliant Minutia location, angle and quality Optional Extended Data area for: Core / delta information Ridge count information Opaque data and future use 10
Finger Minutiae Data Format Compact size Example a two-finger record with 27 & 22 minutiae occupies 342 bytes Extensible Extended data area Designed for Interoperability 11
Project 1563-D - Finger Pattern-Based Interchange Format - INCITS 377 Colin Soutar, Ph.D. Chief Technology Officer, Bioscrypt Inc., INCITS M1 Vice-Chairman 12
Finger Pattern Based Data Interchange Format Based on Robust Commercial Algorithm Implementations Small Area/Slide Format Sensors Data Reduction Reduction in Resolution Cellular Representation Data Format Definition 13
Cellular Representation -1 Pattern broken down into cells Sample Cell Finger Pattern 14
Cellular Representation - 2 Each Cell represented by a sinusoidal waveform. Sample Cell θ δ λ 15
Example Original Image 128x128 x 8 bits Reduction in resolution to 200 dpi (96x96 x 8 bits) Cellular Representation 14x16 x 10 bits (280 bytes + 28 bytes of quality information) ~ 50:1 reduction in data storage requirements (would be 200:1 for 500 dpi image). Header Overhead - 43 bytes 16
Project 1576-D - Iris Interchange Format - INCITS 379 James L. Cambier, Ph.D. VP Engineering and Chief Technology Officer, Iridian Technologies Inc. INCITS M1.1 Vice-Chair 17
Iris Image Interchange Scope Image capture requirements (informative) Data formats Cartesian (rectilinear) image Polar Image Compliance Conformity to data format Multiple image quality levels (per CBEFF) 18
Iris Image Capture Recommendations Resolution Grayscale Density Illumination Contrast Visible Iris Pixel Aspect Ratio Image Scale Optical Distortion Noise Image Orientation Presentation Image Storage Format...to assure consistent performance and interoperability 19
Iris Format Options Data Storage Options Compressed/uncompressed Variable image size Lossy/lossless compression Variable intensity depth Monochrome/color Multiple quality levels Orientation Cartesian Format Polar Format Scan type Occlusions Expected iris diameter Boundary extraction Rotation angle Transformation type 20
Iris Image Formats Pupil Boundary + + r=0 r=n Cartesian 20K (compressed) θ=0 r=0 M θ=0 1 2 3 Iris Boundary θ=m r=n Polar 2K (uncompressed) 21
Project 1577-D - Finger Image Based Data Format for Data Interchange - INCITS 378 R. Michael McCabe Computer Scientist, Information Technology Laboratory, National Institute of Standards and Technology 22
Finger Image Based Data Interchange Format Standard Developer: INCITS M1 BIOMETRICS & JTC1/SC 37 Generic Image standard for applications requiring: Exchange of raw or processed fingerprint images Not limited by the amount of resources Computer identification and verification functions Human comparison of fingerprint image details Non-proprietary storage technique for processed image data Does not rely on pre-established definitions to reduce data to features 23
Image Standard Features Storing, recording, and transmitting information from one or more fingers or palm image areas Defines the content, format, and units of measurement for the image data CBEFF compliant structure Consists of a variety of mandatory/optional items Scanning resolution (125-1000 ppi) Grayscale levels ( 2-256) Compressed / uncompressed images Certification (EFTS/F) 24
Order of Scanned Lines 25
Finger/Palm Image Record Format Fixed-length general record header Single Record for each finger/multifinger or palm Fixed-length finger header Compressed / uncompressed image data Features: Multiple fingers / palms in a record Fourteen Finger Position codes / 16 Palm codes Multiple views Provisions for four compression approaches 26
Project 1565-D - Face Recognition Format for Data Interchange Paul Griffin, Ph.D. Chief Technology Officer, Identix Inc. Kirsten Nobel, Ph.D. Sales Engineer, Identix Inc. 27
Facial Data Interchange Format Preamble: 5 Objectives of Draft Facial Standard By establishing standard formats for facial images, we will: Allow interoperability among facial recognition vendors Minimize the amount of data necessary to store and thereby encourage adoption of biometrics in applications where interoperability is vital Facilitate the use of face information with applications that have limited storage Ensure that enrolled images will meet a quality standard needed for face recognition Improve system throughput by saving the intermediate data instead of the raw data 28
Facial Data Interchange Format Overview Paul Griffin, CTO - Technical Editor Work began March, 2002 National INCITS M1 doc quite mature. Mirrored in ISO SC37 draft standard, developed in collaboration with Passports Australia and the ICAO NTWG. Specifies standard interchange format for facial image (as opposed to template). Goal: interoperability at level of a human readable image, no larger than necessary to preserve information needed for optimal biometric match performance if image later used to create template. Through SC37, solicited comments from face vendor community and card manufacturers/id system vendors. Beginning to receive input. Encourage others interested parties to contribute. 29
Facial Data Interchange Format Highlights: Specifies format of CBEFF-compliant header and data blob. Deals with compression level and type (JPEG, JPEG2000), plus specifics related to facial image quality, including resolution, pose, expression, lighting, etc. Informative Annexes outline an example format, the FaceIt canonical image and provide quantitative analyses of effects of various compression methods on resultant facial match performance included in another Annex. 30
Facial Data Interchange Format Document Outline 1. Introduction scope, related standards activities, normative references, definitions, etc. 2. General record format Header - CBEFF compliant Landmarks text info. associated with face, plus pose angle Image data ( blob ) supported image formats, compression types, data conventions. Definitions of (a) full (b) canonical and (c) cropped image types. 3. Image Acquisition Requirements (e.g. pixel aspect ratio, depth, image resolution, scanner resolution, color profile, etc.) 31
Facial Data Interchange Format Informative Annexes Document Outline (cont d.) A. Photography Best Practices B. Digital Attributes of Full and FaceIt Canonical Image C. Image Compression (for color and grayscale; JPEG vs. JPEG2K) Contact for any questions, technical contributions, etc.: Kirsten R. Nobel Sales Engineer, Prof. Services Group kirsten.nobel@identix.com 32
Project 1603-D - Signature/Sign Image Interchange Format Rod Beatson President, Transaction Security Inc 33
Typical Signature/Sign Biometric Enrollment Process Digitizer - n samples for Enrollment Form N Template Feature Data Values At Least Three Samples No Compatible? Yes Fj:- ( j = 1..N ) (Mj, Sj) are Mean & Standard Deviation of Feature j Typically: Template Data Values Consist of Means & Standard Deviations So The Template Data Format Is Different From The Feature Data Format 34
Typical Signature/Sign Biometric Matching Process One Sample: Form N Feature Data Values Typically: Digitizer One Sample For Matching No Compare With Template Data OK? Yes Update Template Feature Data Values Fj:- ( j = 1..N ) (Mj, Sj) are Mean & Standard Deviation of Feature j Sample Feature Data Values Are Compared with The Template Using Means & Standard Deviations 35
Objective Of Signature/Sign Biometric Interchange Data #Standard Data From Many Different Digitizers With Different Resolutions And Drivers:- Off-The-Shelf Digitizers, Tablet PC, POS Signature Capture Devices, Pocket PC s, Palms, Phone Enabled PDA s etc. Send/Receive! Template From System A"System B! Sample Feature Data From System A"System B! Raw Sample And Sample Feature Data! Raw Sample, Feature Data & Template Data! Combinations Of the Above! Compare Sample from Vendor A Algorithms With Template Created By Vendor B Algorithms 36
Signature Sign/Biometric Interchange Data Format! Record Header (14 Bytes) e.g. Format Identifer Total Record Length CBEFF PID No. of Templates in Record! Feature Data (Mandatory For Most Applications) Digitizer Data (ID, Resolution, etc.) Common Feature Sample Data Private Feature Sample Data! Template Data (If Present) Means/SD s Common Feature Template Data Private Feature Template Data! Raw Signature/Sign Data (If Present) Number of Bytes in Image N samples of (X,Y,T,P) Data 37
Comparing Vendor A Feature Data With Vendor B Template Signature/ Sign Sample Capture Algorithm A Template Generation Standard Header Common Algorithm A Template Proprietary Data Data Signature/ Sign Sample Capture Algorithm A Matching Matching Results Signature/Sign Sample Capture Algorithm B Matching Matching Results 38
Hand Geometry Interchange Format Samir Tamer Technology Physicist, Recognition Systems 39
Hand Geometry New M1 project, approved August 2003 Hand geometry measures the size and shape of the hand Most widely deployed biometric in Access Control and Time & Attendance markets (Frost&Sullivan World Biometric Report 2002) 40
Hand Geometry Project goals Vendor independence Extensibility Small dataset size Proposed interchange format Hand silhouette (not image) Compressed x,y data 41
Hand Geometry 42
Summary of Project Status Project 1564-D 1563-D 1576-D 1577-D 1565-D 1603-D Topic Finger Minutia Format Finger Pattern Format Iris Image Format Finger Image Format Face Image Format Signature/Sign Format Hand Geometry Format Status Public Review Public Review Public Review Committee Draft Committee Draft Committee Draft Proposed Project 43