Ask Your Question

Local binary patterns

asked 2014-03-10 10:54:21 -0500

raqghu gravatar image

can lbp regonize face eventhough there is a change in position of eyes,nose etc i.e patterns of face and if yes how?i will thankful if u help me.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2014-03-10 11:45:02 -0500

Your question should be clearer. AFAIK, a face recognition system has serveral stages: face detecion, face cropping, illumination normalization, feature extraction, dimensionality reduction and classification. The main usage of LBP is for feature extraction although someone used it (very rarely) for illumination normalization, that means it is used after you have a good aligned face image to build a distinctive representation to distinguish face images. The working mechanism of LBP does not depend on the position of eyes, nose ..

edit flag offensive delete link more


(Thanks for the response.But my question is) we use LBP for feature extraction and it depends on intensities over a subregion.If position of nose or eye varies from one image to another,we get that subregion at different place and lbp value at position varies in both will it be able to recognise as it takes only lbp values into consideration.

raqghu gravatar imageraqghu ( 2014-03-12 09:37:39 -0500 )edit

If you apply LBP on that regions, which are detected and cropped correctly, then the answer is yes, in the performance limit of LBP.

tuannhtn gravatar imagetuannhtn ( 2014-03-12 10:03:33 -0500 )edit

thanks for the reply.i have cropped the faces properly.But while creating a template of a person's database,i used his all angles.Here i have a doubt that features of frontal face are entirely different from that of side angles etc and template is average of now if i give side angle pic as test pic,will my template will be strong enough to distinguish the person? If anyone feel that my question is inavalid,i request you to please help me with the concept.thank u.

raqghu gravatar imageraqghu ( 2014-03-24 10:45:41 -0500 )edit

As you noticed, LBP is not strong enough to distinguish pose variant face images from their frontal ones, especially when the pose is large (>45 degree). So I said "in the performance limit of LBP". You can use an advanced method which is dedicated to pose variant face recognition, such as this (

tuannhtn gravatar imagetuannhtn ( 2014-03-24 11:44:45 -0500 )edit

thank you for helping me

raqghu gravatar imageraqghu ( 2014-03-26 07:43:43 -0500 )edit
Login/Signup to Answer

Question Tools


Asked: 2014-03-10 10:54:21 -0500

Seen: 53 times

Last updated: Mar 10 '14