Face recognition with local binary patterns. T, A. Ph.D. Thesis, 2004. abstract bibtex The purpose of automatic face recognition is to recognise a person from a
facial image. Face recognition is a significant area of research in the
field of machine vision since it is a challenging problem and has practical
applications in, e.g. user interfaces and automatic access control. The
face recognition methods that have been developed so far perform well under
controlled circumstances but changes in illumination or pose angle and
expression variation or aging of the subject cause problems to the
recognition systems.
This thesis presents a new method for automatic face recognition. The
method is based on dividing the facial image into small local regions, each
of which is then described with descriptors developed in texture analysis
research. The descriptors derived from each of the regions represent the
appearance of the corresponding region. The local descriptors are combined
into a feature vector describing the whole face and its geometry.
This work especially concentrates on applying the local binary pattern
descriptor. Additionally, grey-level difference histogram, texton histogram
and homogeneous texture descriptor are used as control methods in the
description of local regions. The results obtained with the proposed method
are compared to state-of-the-art methods in the face recognition research:
principal component analysis, the Bayesian intra/extrapersonal classifier
and elastic bunch graph matching. The comparison is carried out using the
FERET database which is commonly used in assessing the performance of face
recognition methods.
In the performed tests it was found that the local binary pattern method
produces better recognition results than other texture methods especially
on difficult test image sets that contain, e.g. illumination changes. The
proposed method allows for weighting different parts of the face in the
recognition phase based on the importance of the information they contain.
In the conducted study it was noticed that especially the eye area is
significant in terms of recognition. By weighting this area a notable
increase in the performance of the system was obtained. The system
utilising the weighting of facial regions reached better recognition
results than any of the control algorithms in all of the test image sets.
Based on the results produced it can be concluded that the presented method
is well suited for face recognition and it is advisable to continue
research on the subject. In this thesis also the weaknesses of the proposed
method are considered and possible solutions to them are presented.
@phdthesis{
title = {Face recognition with local binary patterns.},
type = {phdthesis},
year = {2004},
id = {8b3a3e64-1f2b-3880-84f9-85a121d3a8b2},
created = {2019-11-19T13:01:26.595Z},
file_attached = {false},
profile_id = {bddcf02d-403b-3b06-9def-6d15cc293e20},
group_id = {17585b85-df99-3a34-98c2-c73e593397d7},
last_modified = {2019-11-19T13:46:25.646Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
citation_key = {mvg:630},
source_type = {mastersthesis},
notes = {M.Sc. thesis, Department of Electrical and Information Engineering, University of Oulu, Finland, 55 p.},
private_publication = {false},
abstract = {The purpose of automatic face recognition is to recognise a person from a
facial image. Face recognition is a significant area of research in the
field of machine vision since it is a challenging problem and has practical
applications in, e.g. user interfaces and automatic access control. The
face recognition methods that have been developed so far perform well under
controlled circumstances but changes in illumination or pose angle and
expression variation or aging of the subject cause problems to the
recognition systems.
This thesis presents a new method for automatic face recognition. The
method is based on dividing the facial image into small local regions, each
of which is then described with descriptors developed in texture analysis
research. The descriptors derived from each of the regions represent the
appearance of the corresponding region. The local descriptors are combined
into a feature vector describing the whole face and its geometry.
This work especially concentrates on applying the local binary pattern
descriptor. Additionally, grey-level difference histogram, texton histogram
and homogeneous texture descriptor are used as control methods in the
description of local regions. The results obtained with the proposed method
are compared to state-of-the-art methods in the face recognition research:
principal component analysis, the Bayesian intra/extrapersonal classifier
and elastic bunch graph matching. The comparison is carried out using the
FERET database which is commonly used in assessing the performance of face
recognition methods.
In the performed tests it was found that the local binary pattern method
produces better recognition results than other texture methods especially
on difficult test image sets that contain, e.g. illumination changes. The
proposed method allows for weighting different parts of the face in the
recognition phase based on the importance of the information they contain.
In the conducted study it was noticed that especially the eye area is
significant in terms of recognition. By weighting this area a notable
increase in the performance of the system was obtained. The system
utilising the weighting of facial regions reached better recognition
results than any of the control algorithms in all of the test image sets.
Based on the results produced it can be concluded that the presented method
is well suited for face recognition and it is advisable to continue
research on the subject. In this thesis also the weaknesses of the proposed
method are considered and possible solutions to them are presented.},
bibtype = {phdthesis},
author = {T, Ahonen}
}
Downloads: 0
{"_id":"Qty5qr6m3554YWy2S","bibbaseid":"t-facerecognitionwithlocalbinarypatterns-2004","authorIDs":[],"author_short":["T, A."],"bibdata":{"title":"Face recognition with local binary patterns.","type":"phdthesis","year":"2004","id":"8b3a3e64-1f2b-3880-84f9-85a121d3a8b2","created":"2019-11-19T13:01:26.595Z","file_attached":false,"profile_id":"bddcf02d-403b-3b06-9def-6d15cc293e20","group_id":"17585b85-df99-3a34-98c2-c73e593397d7","last_modified":"2019-11-19T13:46:25.646Z","read":false,"starred":false,"authored":false,"confirmed":"true","hidden":false,"citation_key":"mvg:630","source_type":"mastersthesis","notes":"M.Sc. thesis, Department of Electrical and Information Engineering, University of Oulu, Finland, 55 p.","private_publication":false,"abstract":"The purpose of automatic face recognition is to recognise a person from a\nfacial image. Face recognition is a significant area of research in the\nfield of machine vision since it is a challenging problem and has practical\napplications in, e.g. user interfaces and automatic access control. The\nface recognition methods that have been developed so far perform well under\ncontrolled circumstances but changes in illumination or pose angle and\nexpression variation or aging of the subject cause problems to the\nrecognition systems.\n\nThis thesis presents a new method for automatic face recognition. The\nmethod is based on dividing the facial image into small local regions, each\nof which is then described with descriptors developed in texture analysis\nresearch. The descriptors derived from each of the regions represent the\nappearance of the corresponding region. The local descriptors are combined\ninto a feature vector describing the whole face and its geometry.\n\nThis work especially concentrates on applying the local binary pattern\ndescriptor. Additionally, grey-level difference histogram, texton histogram\nand homogeneous texture descriptor are used as control methods in the\ndescription of local regions. The results obtained with the proposed method\nare compared to state-of-the-art methods in the face recognition research:\nprincipal component analysis, the Bayesian intra/extrapersonal classifier\nand elastic bunch graph matching. The comparison is carried out using the\nFERET database which is commonly used in assessing the performance of face\nrecognition methods.\n\nIn the performed tests it was found that the local binary pattern method\nproduces better recognition results than other texture methods especially\non difficult test image sets that contain, e.g. illumination changes. The\nproposed method allows for weighting different parts of the face in the\nrecognition phase based on the importance of the information they contain.\nIn the conducted study it was noticed that especially the eye area is\nsignificant in terms of recognition. By weighting this area a notable\nincrease in the performance of the system was obtained. The system\nutilising the weighting of facial regions reached better recognition\nresults than any of the control algorithms in all of the test image sets.\n\nBased on the results produced it can be concluded that the presented method\nis well suited for face recognition and it is advisable to continue\nresearch on the subject. In this thesis also the weaknesses of the proposed\nmethod are considered and possible solutions to them are presented.","bibtype":"phdthesis","author":"T, Ahonen","bibtex":"@phdthesis{\n title = {Face recognition with local binary patterns.},\n type = {phdthesis},\n year = {2004},\n id = {8b3a3e64-1f2b-3880-84f9-85a121d3a8b2},\n created = {2019-11-19T13:01:26.595Z},\n file_attached = {false},\n profile_id = {bddcf02d-403b-3b06-9def-6d15cc293e20},\n group_id = {17585b85-df99-3a34-98c2-c73e593397d7},\n last_modified = {2019-11-19T13:46:25.646Z},\n read = {false},\n starred = {false},\n authored = {false},\n confirmed = {true},\n hidden = {false},\n citation_key = {mvg:630},\n source_type = {mastersthesis},\n notes = {M.Sc. thesis, Department of Electrical and Information Engineering, University of Oulu, Finland, 55 p.},\n private_publication = {false},\n abstract = {The purpose of automatic face recognition is to recognise a person from a\nfacial image. Face recognition is a significant area of research in the\nfield of machine vision since it is a challenging problem and has practical\napplications in, e.g. user interfaces and automatic access control. The\nface recognition methods that have been developed so far perform well under\ncontrolled circumstances but changes in illumination or pose angle and\nexpression variation or aging of the subject cause problems to the\nrecognition systems.\n\nThis thesis presents a new method for automatic face recognition. The\nmethod is based on dividing the facial image into small local regions, each\nof which is then described with descriptors developed in texture analysis\nresearch. The descriptors derived from each of the regions represent the\nappearance of the corresponding region. The local descriptors are combined\ninto a feature vector describing the whole face and its geometry.\n\nThis work especially concentrates on applying the local binary pattern\ndescriptor. Additionally, grey-level difference histogram, texton histogram\nand homogeneous texture descriptor are used as control methods in the\ndescription of local regions. The results obtained with the proposed method\nare compared to state-of-the-art methods in the face recognition research:\nprincipal component analysis, the Bayesian intra/extrapersonal classifier\nand elastic bunch graph matching. The comparison is carried out using the\nFERET database which is commonly used in assessing the performance of face\nrecognition methods.\n\nIn the performed tests it was found that the local binary pattern method\nproduces better recognition results than other texture methods especially\non difficult test image sets that contain, e.g. illumination changes. The\nproposed method allows for weighting different parts of the face in the\nrecognition phase based on the importance of the information they contain.\nIn the conducted study it was noticed that especially the eye area is\nsignificant in terms of recognition. By weighting this area a notable\nincrease in the performance of the system was obtained. The system\nutilising the weighting of facial regions reached better recognition\nresults than any of the control algorithms in all of the test image sets.\n\nBased on the results produced it can be concluded that the presented method\nis well suited for face recognition and it is advisable to continue\nresearch on the subject. In this thesis also the weaknesses of the proposed\nmethod are considered and possible solutions to them are presented.},\n bibtype = {phdthesis},\n author = {T, Ahonen}\n}","author_short":["T, A."],"bibbaseid":"t-facerecognitionwithlocalbinarypatterns-2004","role":"author","urls":{},"downloads":0},"bibtype":"phdthesis","creationDate":"2019-11-19T13:17:06.631Z","downloads":0,"keywords":[],"search_terms":["face","recognition","local","binary","patterns","t"],"title":"Face recognition with local binary patterns.","year":2004}