- |
-
New Texture descriptor Based on Improved Orthogonal Difference Local Binary Pattern
Poster Presentation
Authors
Department of Electrical Engineering, Faculty of Engineering, Yasouj University, Yasouj, Iran
Abstract
Local descriptor plays an important role in Content-
Based Image Retrieval (CBIR) and face recognition. Almost all
local patterns are based on the relationship between neighboring
pixels in a local area. The most famous local pattern is Local
Binary Pattern (LBP), in which the patterns are defined based
on the intensity difference between a central pixel and its
neighboring in a 3×3 local window. Orthogonal Difference Local
Binary Pattern (OLDBP) is an extended version of LBP which
is introduced recently. In this paper, ODLBP is improved. In
the proposed method each 3×3 local window is divided into two
groups and then local patterns of each group are extracted and
finally, the feature vector is provided by concatenating of groups
patterns. To evaluate the proposed method, three datasets Yale,
ORL and GT are used. Implementation results show the powerful
of the proposed method comparing to ODLBP. The proposed
method is more faster than the ODLBP while its precision and
recall are slightly higher than the ODLBP method.
Based Image Retrieval (CBIR) and face recognition. Almost all
local patterns are based on the relationship between neighboring
pixels in a local area. The most famous local pattern is Local
Binary Pattern (LBP), in which the patterns are defined based
on the intensity difference between a central pixel and its
neighboring in a 3×3 local window. Orthogonal Difference Local
Binary Pattern (OLDBP) is an extended version of LBP which
is introduced recently. In this paper, ODLBP is improved. In
the proposed method each 3×3 local window is divided into two
groups and then local patterns of each group are extracted and
finally, the feature vector is provided by concatenating of groups
patterns. To evaluate the proposed method, three datasets Yale,
ORL and GT are used. Implementation results show the powerful
of the proposed method comparing to ODLBP. The proposed
method is more faster than the ODLBP while its precision and
recall are slightly higher than the ODLBP method.
Keywords
content-based image retrieval; face recognition; local binary pattern; orthogonal difference local binary pattern
Proceeding Title [Persian]
New Texture descriptor Based on Improved Orthogonal Difference Local Binary Pattern
Authors [Persian]
Abstract [Persian]
Local descriptor plays an important role in Content-
Based Image Retrieval (CBIR) and face recognition. Almost all
local patterns are based on the relationship between neighboring
pixels in a local area. The most famous local pattern is Local
Binary Pattern (LBP), in which the patterns are defined based
on the intensity difference between a central pixel and its
neighboring in a 3×3 local window. Orthogonal Difference Local
Binary Pattern (OLDBP) is an extended version of LBP which
is introduced recently. In this paper, ODLBP is improved. In
the proposed method each 3×3 local window is divided into two
groups and then local patterns of each group are extracted and
finally, the feature vector is provided by concatenating of groups
patterns. To evaluate the proposed method, three datasets Yale,
ORL and GT are used. Implementation results show the powerful
of the proposed method comparing to ODLBP. The proposed
method is more faster than the ODLBP while its precision and
recall are slightly higher than the ODLBP method.
Based Image Retrieval (CBIR) and face recognition. Almost all
local patterns are based on the relationship between neighboring
pixels in a local area. The most famous local pattern is Local
Binary Pattern (LBP), in which the patterns are defined based
on the intensity difference between a central pixel and its
neighboring in a 3×3 local window. Orthogonal Difference Local
Binary Pattern (OLDBP) is an extended version of LBP which
is introduced recently. In this paper, ODLBP is improved. In
the proposed method each 3×3 local window is divided into two
groups and then local patterns of each group are extracted and
finally, the feature vector is provided by concatenating of groups
patterns. To evaluate the proposed method, three datasets Yale,
ORL and GT are used. Implementation results show the powerful
of the proposed method comparing to ODLBP. The proposed
method is more faster than the ODLBP while its precision and
recall are slightly higher than the ODLBP method.
Keywords [Persian]
content-based image retrieval، face recognition، local binary pattern، orthogonal difference local binary pattern