`(12) Patent Application Publication (10) Pub. No.: US 2010/0066822 A1
`Steinberg et al.
`(43) Pub. Date:
`Mar. 18, 2010
`
`US 2010.0066822A1
`
`(54) CLASSIFICATION AND ORGANIZATION OF
`CONSUMER DIGITAL MAGES USING
`WORKFLOW, AND FACE DETECTION AND
`RECOGNITION
`
`(75) Inventors:
`
`Eran Steinberg, San Francisco, CA
`(US); Peter Corcoran, Galway
`(IE); Petronel Bigioi, Galway (IE):
`Mihai Ciuc, Bucuresti (RO);
`Stefanita Ciurel, Bucuresti (RO);
`Constantin Vertran, Bucuresti
`(RO)
`
`Correspondence Address:
`Tessera/FotoNation
`Patent Legal Dept.
`3025 Orchard Parkway
`San Jose, CA 95134 (US)
`
`(73) Assignee:
`
`FotoNation Ireland Limited,
`Galway (IE)
`
`(21) Appl. No.:
`
`12/554.258
`
`(22) Filed:
`
`Sep. 4, 2009
`
`Related U.S. Application Data
`(63) Continuation-in-part of application No. 10/764,335,
`filed on Jan. 22, 2004, now Pat. No. 7,587,068.
`Publication Classification
`
`(51) Int. Cl.
`(2006.01)
`H04N 7/8
`(2006.01)
`G06K 9/00
`(52) U.S. C. ... 348/77:382/118; 382/224; 348/E07.085
`(57)
`ABSTRACT
`A processor-based system operating according to digitally
`embedded programming instructions performs a method
`including identifying a group of pixels corresponding to a
`face region within digital image data acquired by an image
`acquisition device. A set of face analysis parameter values is
`extacted from said face region, including a faceprint associ
`ated with the face region. First and second reference face
`prints are determined for a person using reference images
`captured respectively in predetermined face-portrait condi
`tions and using ambient conditions. The faceprints are ana
`lyzed to determine a baseline faceprint and a range of vari
`ability from the baseline associated with the person. Results
`of the analyzing are stored and used in Subsequent recogni
`tion of the person in a Subsequent image acquired under
`ambient conditions.
`
`deni & Rei Adrin Query Browser
`\cd.ie
`wide ?ocie iiie
`C
`
`Msg, slide Show Publisher FaceTools
`Oce vice
`viaduie
`viciule
`
`se teace scies
`
`1040
`
`1050
`
`1060 /
`
`Of
`
`1080
`
`o cc or
`
`c oxic to :
`
`oo :
`
`a
`
`Work Cecises
`
`110
`
`190
`
`A.
`
`i
`
`at:
`image
`Det.
`Dei.
`Myle | Module
`is
`
`
`
`18O
`W
`
`refSofia
`cit
`Collection
`
`-
`
`|
`
`--
`
`r
`F.
`3C
`s
`3.
`Recog. Dale |-1150
`Nor
`Module
`tile -
`Module
`f
`*
`e ---,
`is-n
`-
`s
`image
`Classification
`Catabase
`
`y
`
`image
`Classification 1 O
`8
`3
`
`60
`
`c
`
`W
`
`- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
`
`Petitioner Apple Inc. - Ex. 1054, p. 1
`
`
`
`Patent Application Publication
`
`Mar.18, 2010 Sheet 1 of 29
`
`US 2010/0066822 A1
`
`Oo
`oSS
`=
`
`esegeleq
`
`|
`
`euibug
`foRa|NEPameggmememmogmmoyoansgocnespunammeng°70T3me:|ainpow||sinpoy
`
`
`
`
`
`UGHEIYISSEIDOZLLEebeuyJfgOBE
`ainpoy|feinpoy||einpoyy1|sjooy
`
`
`osar=6z01fogor|osor|ovar\oeorl\ozo
`
`
`soe,|jueusiignd||Mmougepi
`le£E
`SeOINOS1OSE"|ebeu|ebeuyi
`
`OLLE-SANPOW]BODAOLIOAR
`
`
`_-|uoqeayisseigf-————Mabel
`
`pooj2UOSIOcIpuleXy
`
`ainpoyy)(ainpow‘Booey||tion2yoq|aoe;||aoe3084
`afeuypjf|A“f5x£7é
`
`Petitioner Apple Inc. - Ex. 1054, p. 2
`
`Petitioner Apple Inc. - Ex. 1054, p. 2
`
`
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 2 of 29
`
`US 2010/0066822 A1
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 3
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 3 of 29
`
`US 2010/0066822 A1
`
`
`
`
`
`
`
`
`
`
`
`?)z)
`
`the r- r - - - - - - - - a ran m r - - - - - - - - - - - - - - - - - - - - - - - - - - - - a ran
`
`Petitioner Apple Inc. - Ex. 1054, p. 4
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 4 of 29
`
`US 2010/0066822 A1
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`|-------}T___
`
`k
`wer
`van
`N
`
`r
`
`s
`xx
`
`Petitioner Apple Inc. - Ex. 1054, p. 5
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 5 of 29
`
`US 2010/0066822 A1
`
`w
`va
`
`ce A. s
`
`xx xx xx xx xx xx xx xx xx xx sex saw
`
`xa
`cy
`N
`
`rex res exa e ex: text ex; we sex w w was a *
`
`Y
`
`w w w w w w me.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ex .
`
`. .
`
`.
`
`- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - r
`
`Petitioner Apple Inc. - Ex. 1054, p. 6
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 6 of 29
`
`US 2010/0066822 A1
`
`Face Detection Workflow (Detection Viodie)
`
`Wait for image from Main
`Workflow (idle State)
`
`
`
`
`
`
`
`
`
`Locate Face Pixel Groupings
`ar Custer Groupings to form XXX XXXX-XXX
`Complete Face Candidate Regions
`
`A.
`
`Scar image with Face Feature Pre-Filter
`. ocate & Mark Face Candidate Regions
`
`3150
`3160-1
`Custer Remaining Face Pixel Groupings
`to Form Partial Face Candidates
`
`Pass Auto-Recognition List to
`workflow Module and return to idle
`
`
`
`
`
`Pass Marua & Auto-Recognition lists
`to Workflow Module and return to die J
`Y-3200
`
`Pass raining list to workflow Module
`
`
`
`FG. 3
`
`Petitioner Apple Inc. - Ex. 1054, p. 7
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 7 of 29
`
`US 2010/0066822 A1
`
`Core Syster Workflow Workflow iodie
`
`Wait for Next image
`*1101WEWE.
`
`Send image to Face Detection Module
`
`
`
`
`
`4.50
`B. Get next Face on Auto List
`X
`
`
`
`4155-
`X
`t
`Serid to face Recog. Mod.
`
`Prompt User for Face identity
`
`X Got Faceprint - ra Record Data
`470- Yesy
`437
`NUpdate Database
`Add to Faceprint Search list
`
`
`
`NO
`A? Face
`Region to "Un
`identified" list
`
`Yes
`
`’ Yes T-m-
`
`
`
`
`
`
`
`
`
`Search Faceprint Database 4210
`for the "N" closest entries
`42201. --
`ce
`X Same Fo class?
`to...Prs
`Yes --
`4230
`4250
`
`
`
`
`
`No.
`Add to Manual list
`4398
`No.
`smresar
`Sane identity car
`Yes
`
`
`
`FIG. 4(a)
`
`Petitioner Apple Inc. - Ex. 1054, p. 8
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 8 of 29
`
`US 2010/0066822 A1
`
`Core System Workflow (Workflow lodie )
`
`4230
`
`
`
`Same FP class?
`Yes
`harr:
`Air Faceprint to Face CaSS
`4240
`427C Link image to Known Person Data)
`X
`428O- Crossink Person to image Data
`429-1 Print visg. to ressage Module
`4300-
`ca
`ser
`NC
`ar
`sr. r. r
`awe
`-last Facepring Wax
`Yes
`
`WAM
`
`Same identity
`Yes
`Add Faceprint to closest Face
`Class for this identity
`
`"4260
`
`439
`TFrom Fig4(a)
`: Add Faceprint to most
`frequent Face Class
`for this identity
`"ggers".
`.
`. .
`.
`.
`Get next Faceprint in
`the Searc list
`X
`
`M
`
`430 load wa? a Recognition list
`4320- Get next Face on Manual list
`A330- Prompt ser for FaCe identity
`
`
`
`4340-
`435
`
`XM
`User Responds cer
`Yes
`if Faceprint is valid create a new face class
`X
`s'.
`and add to identity, Crosslink if age and
`dentity data, Pring Msg. to message module Add Face Region
`f
`to "r-identified"
`4360-Jpdate Database
`list
`4370 ast Entry de
`4380-
`Yes
`Return to Main workflow idle
`
`
`
`FIG. 4(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 9
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 9 of 29
`
`US 2010/0066822 A1
`
`Face Norialization Mode (Voie A
`
`Wait for Face Region from Main
`Workflow (die State
`
`SO2
`
`
`
`X
`
`
`
`53
`
`5efite2 C Apply Face Feature location Filter
`
`
`
`Determine Face
`Orientation & Pose
`
`Mark Face Features (Eye, Nose & Mouth)
`X
`50.24
`
`
`
`
`
`s
`-
`al Semi-Frontal
`5040PN Se-1
`
`
`
`NO s:
`
`
`
`Apply 2-D Transforms to
`b generate Frontai Face
`Reson
`
`- Half-Profile
`505 w
`s Face- x
`
`8 Estate Stretc. Factor or
`xxxxxxxxxxxxxxxxxxxxxacacacacaca
`
`Main Workflow Mode
`and return to ide
`
`Map Face Region to 3.
`Normalized Face Model
`Rotate Frontal Position-5056
`
`X
`
`Regiof
`
`
`
`
`
`
`
`Riga.
`
`pply 2D transforms for
`lumiration aid Scale
`
`p Pass Nortalized Frofia
`Face Region to Workflow
`Module and return to idle
`
`
`
`FIG. 5(a)
`
`Petitioner Apple Inc. - Ex. 1054, p. 10
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 10 of 29
`
`US 2010/0066822 A1
`
`Face Normalization Modie (Mode 8)
`
`5 O.
`
`Workflow (die State)
`
`
`
`
`
`5120
`
`
`
`
`
`
`
`FreFitec
`Yes. >Apply Face Feature locatin Fiter
`No
`eternie Face
`Orientation & Pose s
`5130
`
`Mark Face Features Eye, Nose & O
`524.
`
`)
`
`
`
`-
`Map onto 3D Face Model
`s seriental
`Beard Generate Multi-View
`51407 NaCee-
`- Fans
`NC
`5145
`Apply Normalization
`Fiters for Scaling & k
`
`
`
`
`
`
`
`Orientatio? E. 5 1 5
`
`Pass Normalized Rotated
`Face Region to Workflow
`Wode with Pose Rotation
`ata, retu" to die
`
`FIG. 5(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 11
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 11 of 29
`
`US 2010/0066822 A1
`
`FIG. 6(a)
`
`FIG. 6(b)
`
`FIG. 6(c)
`
`
`
`FIG. 6(d)
`
`Petitioner Apple Inc. - Ex. 1054, p. 12
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 12 of 29
`
`US 2010/0066822 A1
`
`
`
`FIG 7(d)
`
`FIG. 7(e)
`
`Fig. 7(f)
`
`Petitioner Apple Inc. - Ex. 1054, p. 13
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 13 of 29
`
`US 2010/0066822 A1
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 14
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 14 of 29
`
`US 2010/0066822 A1
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 15
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 15 of 29
`
`US 2010/0066822 A1
`
`
`
`
`
`@ZETTJ uee wieqe ||
`
`Petitioner Apple Inc. - Ex. 1054, p. 16
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 16 of 29
`
`US 2010/0066822 A1
`
`
`
`FIG. 9(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 17
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 17 of 29
`
`US 2010/0066822 A1
`
`FIG. 10(a)
`
`
`
`10120
`
`FIG. 10(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 18
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 18 of 29
`
`US 2010/0066822 A1
`
`
`
`FIG. 11(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 19
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 19 of 29
`
`US 2010/0066822 A1
`
`s
`
`y-3-
`a
`*
`
`es
`
`-
`
`-
`
`f
`
`FC
`4
`
`es
`
`i
`f
`f
`Y&
`* -
`><
`11220
`
`as r-e-r-retire .
`re
`t
`--
`3A
`---.S.
`FC
`*s
`".
`“s
`W
`'
`f
`is . .
`.
`.
`. . .
`.
`i--------- a- - -:
`if
`;
`s
`FC ; :
`3B v.
`i;
`if
`e -- ... -- - -
`
`r
`
`Ps
`
`sa
`
`as Y are
`
`FC
`3
`*
`Y
`
`a
`
`Petitioner Apple Inc. - Ex. 1054, p. 20
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 20 of 29
`
`US 2010/0066822 A1
`
`1200
`
`FIG. 12(a)
`
`N-12130
`
`x12232
`NY-12230
`
`>12242
`JCFC,
`
`112d20
`FC
`
`
`
`
`
`
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 21
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 21 of 29
`
`US 2010/0066822 A1
`
`
`
`13052
`
`13056
`13050
`
`13010
`13060
`
`FIG. 13(a)
`
`Petitioner Apple Inc. - Ex. 1054, p. 22
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 22 of 29
`
`US 2010/0066822 A1
`
`.
`
`"
`
`13120-N
`
`new ID,
`1
`i v
`/ 9d p
`-1340
`7
`a
`',
`, /
`1
`it
`f
`f
`f;
`fi
`
`--
`i/
`f
`
`-13130
`
`2
`
`\
`W.
`W.
`
`f
`
`1
`
`f
`W
`Y -- 1
`
`f-13110
`
`A.
`
`FIG. 13(b)
`
`Petitioner Apple Inc. - Ex. 1054, p. 23
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 23 of 29
`
`US 2010/0066822 A1
`
`1
`
`
`
`as
`
`wa Exm a
`
`Y?
`
`3. S.
`N
`
`- 13210
`
`N \
`
`v
`
`f
`
`YN -13220
`
`N-13246
`N13240
`
`FIG. 13(c)
`
`Petitioner Apple Inc. - Ex. 1054, p. 24
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 24 of 29
`
`US 2010/0066822 A1
`
`^ A
`
`xx
`
`- r a
`
`few
`
`1
`
`FIG. 13(d)
`
`
`
`FIG. 13(e)
`
`Petitioner Apple Inc. - Ex. 1054, p. 25
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 25 of 29
`
`US 2010/0066822 A1
`
`
`
`
`
`E?IEREEDT???Tº TOEGIE:55:
`
`Petitioner Apple Inc. - Ex. 1054, p. 26
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 26 of 29
`
`US 2010/0066822 A1
`
`
`
`A3
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 27
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 27 of 29
`
`US 2010/0066822 A1
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 28
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 28 of 29
`
`US 2010/0066822 A1
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 29
`
`
`
`Patent Application Publication
`
`Mar. 18, 2010 Sheet 29 of 29
`
`US 2010/0066822 A1
`
`
`
`Petitioner Apple Inc. - Ex. 1054, p. 30
`
`
`
`US 2010/0066822 A1
`
`Mar. 18, 2010
`
`CLASSIFICATION AND ORGANIZATION OF
`CONSUMER DIGITAL IMAGES USING
`WORKFLOW, AND FACE DETECTION AND
`RECOGNITION
`
`CROSS-REFERENCE TO RELATED
`APPLICATIONS
`0001. This application is a continuation in part (CIP) of
`U.S. patent application Ser. No. 10/764,335, filed Jan. 22.
`2004, which is one of a series of contemporaneously-filed
`patent applications including U.S. Ser. No. 10/764,339, now
`U.S. Pat. No. 7,551,755, entitled, “Classification and Orga
`nization of Consumer Digital Images using Workflow, and
`Face Detection and Recognition': U.S. Ser. No. 10/764,336,
`now U.S. Pat. No. 7,558,408, entitled, “A Classification Sys
`tem for Consumer Digital Images using Workflow and User
`Interface Modules, and Face Detection and Recognition':
`U.S. Ser. No. 10/764,335, entitled, “A Classification Data
`base for Consumer Digital Images': U.S. Ser. No. 10/764,
`274, now U.S. Pat. No. 7,555,148, entitled, “A Classification
`System for Consumer Digital Images using Workflow, Face
`Detection, Normalization, and Face Recognition'; and U.S.
`Ser. No. 10/763,801, now U.S. Pat. No. 7,564,994, entitled,
`'A Classification System for Consumer Digital Images using
`Automatic Workflow and Face Detection and Recognition'.
`
`BACKGROUND
`0002 1. Field of the Invention
`0003. The invention relates to digital image processing,
`particularly to the field of automatic or semiautomatic group
`ing and classification of images in a database or image col
`lection and based on the occurrence of faces in the images and
`the identification and classification of Such faces.
`0004 2. Description of the Related Art
`0005. The techniques of face detection and face recogni
`tion are each being explored by those skilled and a great many
`advancement have been made in those respective fields in
`recent years. Face detection has to do with the problem of
`locating regions within a digital image or video sequence
`which have a high probability of representing a human face.
`Face recognition involves the analysis of such a “face region'
`and its comparison with a database of known faces to deter
`mine if the unknown “face region' is sufficiently similar to
`any of the known faces to represent a high probability match.
`The related field of tracking involves face or identity recog
`nition between different frames in a temporal sequence of
`frames. A useful review of face detection is provided by Yang
`et al., in IEEE Transactions on Pattern Analysis and Machine
`Intelligence, Vol. 24, No. 1, pages 34-58, January 2002. A
`review of face recognition techniques is given in Zhanget al.,
`Proceedings of the IEEE, Vol. 85, No. 9, pages 1423-1435,
`September 1997.
`0006. Other related art refers to the grouping, classifica
`tion, management, presentation and access to collections of
`digital images in databases, file-systems or other storage
`mechanisms, being based on image content, global image
`parameters, or image metadata. Such content based
`approaches analyze the image content using spatial color
`distribution, texture, shape, object location and geometry, etc.
`However they do not explicitly teach to utilize face recogni
`tion in conjunction with these techniques, or to initially detect
`faces in their images, prior to applying a recognition process.
`It is recognized in the present invention that an advantageous
`
`system that provides automation in the detection, recognition
`and classification processing of digital images would be
`highly desirable.
`0007. None of the prior art references that are cited
`throughout the description below provides this feature. Many
`of the classification techniques described are applied to entire
`images and they do not teach to detect faces in an image, or to
`perform recognition of Such faces. Many of these references
`concentrate on methods storing or accessing images using
`databases, but they do not employ in conjunction with these
`methods the advantageous image processing techniques
`described by inventors of the present invention.
`0008. Some of the medical applications provide classifi
`cation and archiving of images into particular groups that are
`associated with a single customer. For example, the customer
`may be a patient and the classification may be particularly
`related to medical diagnosis or treatment applications where
`a large amount of image data (X-rays, Ultrasound scans, etc)
`which is related to a single patient may be gathered. However,
`these do not utilize face recognition as a means to compile or
`manage this image data, i.e., a user is expected to categorize
`the image according to the associated patient.
`0009. Further references available in the literature of the
`related art describe multi-format transcoding applications for
`visual data. Others describe means for constructing digital
`photo albums. These references do not, however, teach to use
`image processing techniques in the management or access of
`the data.
`0010. At this point we note that the present invention is
`presented primarily in the context of collections of consumer
`digital images which would be generated by a typical user of
`a digital camera. Such an image collection is in a constant
`state of growth as new sets of images are added every time the
`user off-loads pictures from the camera onto his computer.
`Because the image set is in a constant state of flux, it is often
`not practical to perform database-wide sorting, grouping or
`management operations every time a few images are added to
`the collection, because this would put an excessive load on the
`users computer. Much of the related art literature describes
`how to function with and operate on a large static image
`collection. Thus when a sizeable batch of new images is
`added, as will oftenhappen when a camera is offloaded, these
`related art teaching do not describe how to perform significant
`image processing and database-wide testing to determine
`similarities between new and existing database images and
`then group and store the new images before the user can
`access and enjoy his pictures. In reality the application of
`image processing techniques, or of other image-related tools
`is understood by the inventors in the present invention as
`being an ongoing process for collections of consumer images
`and for the design of these tools, where possible, to operate as
`automated or semi-automated background processes for
`applications in consumer imaging.
`0011. There is a very compelling need for new and
`improved tools to manage collections of images. More par
`ticularly, there is a need for tools, which can manage and
`organize image collections which are in a constant state of
`change and growth. It is also important that these tools can
`manage and organize Such ad-hoc collections using methods,
`which are easily understandable by the layman and, where
`possible that Such tools can function in semi- or fully-auto
`
`Petitioner Apple Inc. - Ex. 1054, p. 31
`
`
`
`US 2010/0066822 A1
`
`Mar. 18, 2010
`
`matic modes so that their work of cataloging and organizing
`is almost imperceptible to the end-user.
`
`BRIEF DESCRIPTION OF THE DRAWINGS
`0012 FIGS. 1(a)-1(c) shows an overview of the principle
`components of the invention;
`0013 FIG. 1(a) is an outline of the main system compo
`nents implemented as a computer program.
`0014 FIG. 1(b) shows an alternative embodiment in
`which certain aspects of the system of the preferred embodi
`ment, including the face detection and recognition modules,
`are implemented within an image capture appliance Such as a
`digital camera, while the remaining aspects are implemented
`as a computer program on a desktop computer.
`0015 FIG.1(c) shows an embodiment wherein the system
`is entirely implemented within a digital camera.
`0016 FIG. 2(a) describes an embodiment of a main sys
`tem database.
`0017 FIG. 2(b) gives additional detail of the face recog
`nition data component of the database.
`0.018 FIG.3 describes a main face detection workflow in
`accordance with a preferred embodiment.
`0019 FIGS. 4(a)-4(b) describe a core system workflow in
`accordance with a preferred embodiment.
`0020 FIGS. 5(a)-5(b) shows a face normalization work
`flow in accordance with a preferred embodiment.
`0021
`FIGS. 6(a)-6(e) illustrate shows some of the differ
`ent ways that a face candidate region, obtained from the
`detection module, can be distorted; these distortions should
`be corrected by the normalization module:
`0022 FIG. 6(a) shows a frontal candidate region which is
`incorrectly oriented and must be rotated into an upright posi
`tion prior to applying the face recognition module;
`0023 FIG. 6(b) is a frontal candidate region which is of a
`reduced size and must be enlarged prior to applying the rec
`ognition module;
`0024 FIG. 6(c) is a correct frontal face candidate region
`which does not require either orientation or size correction;
`0025 FIGS. 6(d) and 6(e) illustrate two non-frontal face
`candidate regions which require pose normalization in addi
`tion to size and orientation normalization.
`0026 FIGS. 7(a)-7(f) illustrate how a 3-D model can be
`applied to model a range of face candidate regions:
`0027 FIGS. 7(a)-7(c) illustrate how a simple 1-D scaling
`of a normalized face model can be used to model the majority
`of face candidate regions with good accuracy;
`0028 FIGS. 7(d)-7(f) illustrate how a 2-D face candidate
`region can be mapped onto Such a 3-D normalized face model
`with 1-D scaling along the horizontal axis.
`0029 FIGS. 8(a)-8(b) illustrate how three face regions
`(FR1, FR2 and FR3) may be mapped to faceprints (FP1, FP2,
`and FP3) in a 3-component face space.
`0030 FIG. 8(c) illustrates multiple face regions extracted
`from digital images that have subtle pose, orientational, illu
`mination and/or size distortions to be adjusted automatically
`upon detection in a normalization process in accordance with
`a preferred embodiment prior to automatic or semi-automatic
`face recognition processing:
`0031
`FIG. 9(a) shows a graphical representation of how
`multiple, distinct, face classes, formed from collections of
`closely collocated faceprints can be used to define a unique
`region in face space which is associated with a particular
`person's identity.
`
`0032 FIG. 9(b) illustrates two such identity spaces with
`their associated face classes and faceprints.
`0033 FIG. 10(a) illustrates how a new faceprint creates a
`new face class for a person's identity when it is located at a
`distance further than a certain R, from an existing face
`class.
`0034 FIG.10(b) illustrates how a new faceprint extends or
`grows an existing face class when it is within a distance R
`from the existing face class.
`0035 FIG.11(a) illustrates how an identity region associ
`ated with one person can grow to overlap with the identity
`region of another person.
`0036 FIG.11(b) describes how these overlapping identity
`regions can be separated from each other by shrinking the two
`identity regions into their component face classes.
`0037 FIG. 11(c) illustrates a face class shrinking opera
`tion in accordance with a preferred embodiment.
`0038 FIG. 12(a) shows a face class which has grown over
`time to incorporate a relatively large number of faceprints
`which exhibit localized clustering.
`0039 FIG. 12(b) illustrates explicitly how these faceprints
`are clustered.
`0040 FIG. 12(c) shows how each local cluster can be
`replaced by a single clustered face class which is composed of
`a centre faceprint location in face space and a cluster radius,
`R.
`FIG. 13(a) describes the recognition process where
`0041
`a newly detected faceprint lies in an region of face space
`between two “known identity regions.
`0042 FIG. 13(b) shows how, once the recognition process
`has associated the new faceprint with one of the two known
`identity regions, ID, that identity region then grown to
`include the new faceprint as a new face class within ID.
`0043 FIG. 13(c) shows a similar situation to FIG. 13(a)
`but in this case it is not clear which of the two identity regions
`should be associated with the new faceprint and the system
`must ask the user to make this determination.
`0044 FIG. 13(d) illustrates the case where the user
`chooses ID.
`004.5
`FIG. 13(e) illustrates the case where the user
`chooses ID.
`0046 FIGS. 14(a)-14(d) show a variety of aspects of the
`user interface to the main workflow module.
`0047 FIG. 15(a) illustrates a faceprint associated with an
`acquired facial image:
`0048 FIGS. 15(b) and 15(c) illustrates how a VAR vector
`can be used to align higher order components of flash and
`non-flash feature vectors, or faceprints.
`0049 FIG. 15(d) illustrates a combined, illumination-nor
`malized feature vector for slab and non-flash faceprints.
`0050 FIG. 15(e) illustrates a local baseline faceprint
`determined from a flash image and a radius of variability
`determined from the image in ambient illumination.
`
`INCORPORATION BY REFERENCE
`0051 What follows is a cite list of references each of
`which is, in addition to that which is described as background,
`the invention summary, the abstract, the brief description of
`the drawings and the drawings themselves, hereby incorpo
`rated by reference into the detailed description of the pre
`ferred embodiments below, as disclosing alternative embodi
`ments of elements or features of the preferred embodiments
`not otherwise set forth in detail below. A single one or a
`combination of two or more of these references may be con
`
`Petitioner Apple Inc. - Ex. 1054, p. 32
`
`
`
`US 2010/0066822 A1
`
`Mar. 18, 2010
`
`sulted to obtain a variation of the preferred embodiments
`described in the detailed description herein:
`0.052
`U.S. Pat. Nos. RE33682, RE31370, 4,047,187,
`4,317,991, 4,367,027, 4,638,364, 5,291,234, 5,488,429,
`5,638,136, 5,710,833, 5,724,456, 5,781,650, 5,812, 193,
`5,818,975, 5,835,616, 5,852,823, 5,870,138, 5,911,139,
`5,978,519, 5,991,456, 6,072,904, 6,097,470, 6,101,271,
`6,128,397, 6,148,092, 6,188,777, 6,192,149, 6,249,315,
`6,263,113, 6,268,939, 6,282,317, 6,301,370, 6,332,033,
`6,349,373, 6,351,556, 6,393,148, 6,404,900, 6,407,777,
`6.421,468, 6,438,264, 6,456,732, 6,459,436, 6,473,199,
`6,501,857, 6,502,107, 6,504,942, 6,504,951, 6,516,154,
`6,526,161, 6,564,225, and 6,567,983;
`0053 United States published patent applications no.
`2003/008.4065, 2003/0059121, 2003/0059107, 2003/
`0052991, 2003/004.8950, 2003/0025812, 2002/0172419,
`2002/0168108, 2002/0114535, 2002/0105662, and 2001/
`0031142:
`0054 Japanese patent application no.JP5260360A2;
`0055 British patent application no. GB0031423.7; and
`0056 Yang et al., IEEE Transactions on Pattern Analysis
`and Machine Intelligence, Vol. 24, no. 1, pp. 34-58 (January
`2002).
`
`Illustrative Definitions
`0057 “Face Detection' involves the art of isolating and
`detecting faces in an image; Face Detection includes a pro
`cess of determining whether a human face is present in an
`input image, and may include or is preferably used in com
`bination with determining a position and/or other features,
`properties, parameters or values of parameters of the face
`within the input image:
`0058 “Face Recognition' involves the art of matching an
`unknown facial region from an image with a set of "known
`facial regions.
`0059) “Image-enhancement' or “image correction'
`involves the art of modifying a digital image to improve its
`quality. Such modifications may be “global applied to the
`entire image, or “selective' when applied differently to dif
`ferent portions of the image. Some main categories non
`exhaustively include:
`0060 (i) Contrast Normalization and Image Sharpening.
`0061
`(ii) Image Crop, Zoom and Rotate.
`0062 (iii) Image Color Adjustment and Tone Scaling.
`0063 (iv) Exposure Adjustment and Digital Fill Flash
`applied to a Digital Image.
`0064 (v) Brightness Adjustment with ColorSpace Match
`ing; and Auto-Gamma determination with Image Enhance
`ment.
`0065 (vi) Input/Output device characterizations to deter
`mine Automatic/Batch Image Enhancements.
`0066 (vii) In-Camera Image Enhancement
`0067 (viii) Face Based Image Enhancement.
`0068 “Auto-focusing involves the ability to automati
`cally detect and bring a photographed object into the focus
`field.
`0069. A 'pixel’ is a picture element or a basic unit of the
`composition of an image or any of the Small discrete elements
`that together constitute an image.
`0070 “Digitally-Acquired Image' includes an image that
`is digitally located and held in a detector.
`0071 “Digitally-Captured Image' includes an image that
`is digitally recorded in a permanent file and/or preserved in a
`more or less permanent digital form.
`
`0072 “Digitally-Detected Image': an image comprising
`digitally detected electromagnetic waves.
`0073. A “face region' is a region of a main image which
`has been determined to contain a human face. In particular, it
`may contain a Substantially oval, skin-colored region which
`has physical features corresponding to eyes, nose and mouth,
`or some portion of a face or subset of these facial features.
`0074. A face region is preferably “normalized in accor
`dance with the invention. Prior to extracting face classifier
`parameters (see definition below) from a face region, it is
`preferably first transformed into a normalized form. This may
`involve any or all of three principle steps: (i) resizing to a
`standard "size', e.g., based on the separation of eyes, nose
`and/or mouth; (ii) “orientation” in an upright or other selected
`direction which may involve rotation of the face region; and
`(iii) orientation to compensate for up/down or left/right varia
`tions in the “pose of the face. Note that these normalizations
`may usually performed in reverse order in accordance with a
`preferred embodiment: first pose normalization is imple
`mented, followed by orientation normalization and finally the
`face region is normalized for size. A fourth form of normal
`ization that may be preferably performed is luminance nor
`malization (see below definition), but it is treated or charac
`terized separately from the above, which are referred to as
`spatial normalizations.
`0075 "Face classifier parameters' are a set of values of
`vector and/or scalar classifiers extracted from a normalized
`face region. Typical examples of such a set of classifiers could
`be: (i) principle component vectors, (ii) independent compo
`nent vectors, (iii) 2D fourier transform components, (iv)
`wavelet transform components, (v) gabor components, etc.
`Note that several face classifier techniques may be combined
`to provide a definitive faceprint.
`0076. The set of face classifier parameters associated with
`aparticular face region is known as the “faceprint of that face
`region. The faceprint is preferably a set of face classifier
`parameters and may be subdivided into two or more Subsets
`of face classifier parameters which may overlap.
`(0077. An “archived faceprint” is a set of face classifier
`parameters associated with a particular fate region ultimately
`extracted from a parent image and preferably normalized, and
`stored in the main recognition database, preferably along with
`links to the parent image and the face region.
`0078. A “known identity” is a set of (database) associa
`tions between a known person or other object and one or more
`face classes comprising one or more archived faceprints.
`007.9 The following process is referred to as “luminance
`normalization'. It is common for horizontal and/or vertical
`variations in luminance levels to occur across a face region
`due to the ambient lighting at the time an image was captured
`or other factors such as artificial sources or flashes. In this
`case, certain types of face classifiers may be distorted and it
`may be advantageous to normalize luminance levels across
`the face region prior to extracting face classifier parameters in
`accordance with a preferred embodiment. As typical varia
`tions are linear in form and as the variations manifest them
`selves principally in skin-colored pixels, it is relatively
`straightforward to adjust each image pixel of a face region to
`approximately compensate for Such luminance variations
`caused by ambient lighting.
`0080 When two or more faceprints lie within a certain
`geometric distance of each other in facespace, they may be
`preferably grouped into a single face class. If a newly deter
`mined faceprint lies within this geometric distance of the face
`
`Petitioner Apple Inc. - Ex. 1054, p. 33
`
`
`
`US 2010/0066822 A1
`
`Mar. 18, 2010
`
`class, then this face class may be expanded to include the new
`faceprint, or may be added to the face class without expansion
`if the all of its face classifier values lie within the existing face
`class parameter value ranges. This existing face class is
`referred to as a “prior face class