analysis: type 10 vs. ansi 385
TRANSCRIPT
Analysis: Type 10 vs. ANSI 385
Paul Griffin, Ph.D.Chief Technology OfficerIdentix
April 2005
2© 2005 Identix Inc. All rights reserved.
Topics/Questions
• Type 10s can consist of face images denoted by the “FAC” IMG entry.
• ANSI 385 encodes face image data and factors present in the image and during image creation.
• What is the overlap between type 10 and ANSI 385?
• What additional capabilities would be achieved by adding more fields to type 10?
• What is in 385 that is a candidate for inclusion into a new type 10?
• Is there a logical “fit” between type 10 and 385 or should 385 be added whole elsewhere?
Image Information
4© 2005 Identix Inc. All rights reserved.
Image Information
Type 10• Photo Date• Width/Height• Scale Units• Color Space
– YCC – RGB 888– Grayscale
• Compression– JPEG – JPEGL (Lossless)
ANSI 385• Photo Date (from CBEFF)• Width/Height• Not present• Color Space
– YUV 422 (Specific YCC)– RGB 888 – Grayscale
• Compression– JPEG– JPEG2000
5© 2005 Identix Inc. All rights reserved.
Image Info in 385 not in Type 10
• Source Type • Some usefulness in – Unknown Photo identifying the video vs. – Digital Camera film to compensate for – Scanner log-gamma differences – Video Frame Analog between video and still – Video Frame Digital photography– etc • Interlacing artifacts can
• Device Type be removed from video – Vendor ID when source is identified
• Quality • Interlaced video frame – Placeholder intended to capture can be explicitly
represent expected match forbidden using performance application profiles (PIV).
xy
z
11.5
11.4
11.2
10.2
10.4
10.10
10.810.6
2.14
7.1
11.6 4.6
4.4
4.2
5.2
5.4
2.10
2.122.1
11.1
Tongue
6.26.4 6.3
6.1Mouth
8.18.9 8.10 8.5
8.3
8.7
8.2
8.8
8.48.6
2.2
2.3
2.6
2.82.9
2.72.5 2.4
2.12.12 2.11
2.142.10
2.13
10.610.8
10.4
10.2
10.105.4
5.2
5.3
5.1
10.1
10.910.3
10.510.7
4.1 4.34.54.6
4.4 4.2
11.111.2 11.3
11.4
11.5
x
y
z
Nose
9.6 9.7
9.14 9.13
9.12
9.2
9.4 9.15 9.5
9.3
9.1
Teeth
9.10 9.11
9.8
9.9
Feature points affected by FAPsOther feature points
Right eye Left eye
3.13
3.7
3.9
3.5
3.1
3.3
3.11
3.14
3.10
3.12 3.6
3.4
3.23.8
Feature Points
7© 2005 Identix Inc. All rights reserved.
Why include feature points?
• “Minutia” of the face– Required for all face matching algorithms (at least
the eye positions at minimum for 2D matching)– Can be automatically determined with good quality
(mainly frontal face) data.– Can be difficult to determine in low-res video or with
significant pose• Future Forensic Face Applications
– Submitting non-frontal non-high-quality face images for searching will require feature point determination on the client
– Similar to minutia markup of latents with investigator workstations
8© 2005 Identix Inc. All rights reserved.
On Forensic Face
• Applications are being developed where face images are used for lookup either fused with finger or stand-alone investigations.
• Significant manipulation of the face image may be required – sharpening, gamma correction, noise removal, frame merging, pose correction.
• Landmarks and pose information are likely to be submitted along with the face image data to aid the search system(s).
• Functionality requires additions to ANSI/NIST-ITL-2000 – feature points, new pose angles.
Facial Information
10© 2005 Identix Inc. All rights reserved.
Differences in Philosophy
Type 10 ANSI 385• Can denote “unique” • Encapsulates permanent
permanent physical properties of the face characteristics of body image discernable by per NCIC code manual examination of the
– Moles photograph– Scars – Eye color– Tattoos – Hair color– Missing body parts – Observed Gender– Eye color • Encapsulates temporary – Hair color face properties associated
• Encapsulates some with face match temporary characteristics performance
– Pose – Expression– Hat/scarf – Blinking– glasses – 3D pose angles
– glasses
11© 2005 Identix Inc. All rights reserved.
Permanent Physical Characteristics
• Type 10, via NCIC code manual, allows for a much richer encoding of these characteristics
– Unlikely that 385 can contribute new indicators– Organization could be improved to allow for easier
input by users…
12© 2005 Identix Inc. All rights reserved.
Temporary Image Characteristics
• For the face, 385 offers a richer parameterization of these indicators due to focus on face match performance
– Teeth visible– Mouth open– Smile (closed jaw)– Smile (mouth open)– Raised eyebrows– Eyes looking away from camera– Squinting– Frowning
• Useful indicators for forensic face applications
13© 2005 Identix Inc. All rights reserved.
Pose angles in 385
• Superset of Type 10– Encoding difference need
to be addressed• 3D pose angles in 385
– Yaw: same angle as subject pose in type 10
– Pitch: difficult issue for face match algorithms
– Roll: easier to compensate for using image processing
• Will typically be required manual input for forensic face applications
On Quality
15© 2005 Identix Inc. All rights reserved.
Activity/Status
Resolution (Image Size)Under Exposured (Brightness)Good Contrast (Entropy)Well Focused (Sharpness)Others (Half-tone patterns, …)
• NIST (Grother) is working to develop face quality methodology
• Identix and others are developing quality metrics
• Placeholders in ANSI and ISO formats
• However, not ready to mandate a universal face quality score for frontal face images
• And not ready to certify face quality modules Head Viewpoint (Pose)
Face Presency (Faceness)Detectable Eyes (Confidence)
Others (Natural Expression, Liveness, …)
Face Geometry (Size, Position)Wearing Eyeglasses (Clear eyes,
Glare)High Resolution (Texture)Proper Lighting (Face &
Background)
On Face Image Types
17© 2005 Identix Inc. All rights reserved.
Full Frontal Face Image Type
• A single “application • Similar concept to 381 profile”, the frontal image finger image data levels type encapsulates most 31 requirements with requirements for superior WSQ 15:1 and appendix Fmug shot capture • Similar to Mike McCabe’s
– The PIV application profile photo mug shot best starts with this and practices improves by mandating compression and source – Lists Detailed constraints requirements for scene,
• This is not present in lighting, subject, format.
Type 10, which relies – Compression requirements extremely upon other “best important
practice” documentation
Recommendations
19© 2005 Identix Inc. All rights reserved.
Options to Extend Type 10 (part 1)
1. Add color JPEG2000 as an allowed image format1. Superior human examination quality for
fixed compression compared to JPEG2. Add source type to aid in artifact removal and
gamma (gain) correction.3. Add quality measure with algorithm source
identifier4. Add feature point(s) Facial Image Attribute
1. Essential for face search applications.2. Possible use for skull forensics
20© 2005 Identix Inc. All rights reserved.
Options to Extend Type 10
5. Add 3D pose angle set.1. Essential for face investigator workstations
6. Add Frontal Mug Shot Application Profile1. Including compression limits2. Including capture source requirements3. Make into the improved Mug Shot capture standard
7. Add Temporary Face Image Attributes1. Expression, blinking, etc.2. Helpful to face match algorithms.
Thank You. Questions?