正在加载图片...
Challenges on Face Recognition Face Problem (1) Face HUman face composes of OIl complex elements expression, accessories, color, occlusions, and brightnes camera capture variation custom) Illumination 口 High ACCRUARY OPose variation problem OFacial expressions 口 Size variation OFacial Appearance 口 Processor intensive Hairstyles Occlusion aTime requirement 口 Need a large Database ORientation problem(Face The same person may have very different appearance for testing 口Gen ue-Low oVerlap faces Challenges: automatically locate and recognize a face from Uniqueness DDifferent moving directions rent illumination conditions, fa expressions, and aging effects Face Problem(2) ace Problem(3) FACES CAN LIE R Diferent persons may hae er similar apearanceFINGERPRINTS.NEVER Face Problem(4): Face Face Verification Performance Illumination Category Problem Illumination from Direct frontal Illumination from FAR right side Same day, same illumination 2 04 Same day, different from Illumination fro lumination amination Different days Facial Different days over 1.5 Expression apart Problem mouth wide to Evaluating Biometric syste Solution: Elastic 1111 Biometrics Research Centre (UGC/CRC) Lecture 8 - 61 Human face composes of complex elements Camera capture variation High ACCRUARY Requirement Size variation Age Processor intensive Time requirement Need a large Database for testing General issue – Low Uniqueness Challenges on Face Recognition Illumination problem (Uncontrolled background vs. uniform background in custom) Pose variation problem Facial expressions Facial Appearance z Glasses z Makeup z Hairstyles Orientation problem (Face angle) Overlap faces Different moving directions Biometrics Research Centre (UGC/CRC) Lecture 8 - 62 Face Problem (1) Face Problem (1) • Faces with intra-subject variations in pose, illumination, expression, accessories, color, occlusions, and brightness Occlusion Head pose Illumination ⇒ The same person may have very different appearance Challenges: automatically locate and recognize a face from a general view point under different illumination conditions, facial expressions, and aging effects Biometrics Research Centre (UGC/CRC) Lecture 8 - 63 ⇒ Different persons may have very similar appearance Twins www.marykateandashley.com Father and son news.bbc.co.uk/hi/english/in_depth/americas /2000/us_elections Face Problem (2) Face Problem (2) Biometrics Research Centre (UGC/CRC) Lecture 8 - 64 Face Problem (3) Face Problem (3) Biometrics Research Centre (UGC/CRC) Lecture 8 - 65 Face Problem (4): Face Verification Performance Face Problem (4): Face Verification Performance Different days over 1.5 years 2 43 apart Different days 2 11 Same day, different 2 9 illumination Same day, same illumination 2 0.4 FRR (%) FAR (%) Category P. Jonathan Phillips et al., An Introduction to Evaluating Biometric Systems, IEEE Computer, pp. 56-63, Feb 2000 Biometrics Research Centre (UGC/CRC) Lecture 8 - 66 Illumination Illumination Problem Illumination from right side Illumination from left side Direct frontal illumination Direct frontal illumination Illumination from top lighting Illumination from bottom lighting Solution: Illumination Cones Facial Expression Expression Subject with Problem smiling face Subject with no facial expression Subject with mouth wide open Solution: Elastic graph matching