Local Terinary Pattern ( Feature extarction ) and DOG in opencv

asked 2013-09-19 07:52:59 -0500

UserOpenCV gravatar image

updated 2013-09-19 08:22:35 -0500

I am working on improving a "Face recognition" project in which there is a matlab code which uses "LTP ( Local Terinary Pattern )" and "DOG (difference of Gausiann Filter)". I am trying convert this Matlab part to Opencv.

Is there any good implementations of "LTP" and "DOG" in opencv? The current "LTP" is the implementation of this paper - Enhanced Local Texture Feature Sets for Face Recognition Under Difficult Lighting Conditions by Xiaoyang Tan and Bill Triggs.

I found a code for "LTP" here in opencv. But it just looks like pre-processing step? Can some one refer to some LTP - feature extraction C++/Opencv implementations?

edit retag flag offensive close merge delete



yes the code you found is (only ? atleast !) the preprocessing. but it covers the dog as well as the gamma correction and the contrast equalization already.

but the ltp pattern is easily derived from the lbp one, right ?

  • one pattern for: neighbour > center + t
  • one pattern for: neighbour > center - t && neighbour < center + t
  • one pattern for: neighbour < center - t

you actually don't need to calculate the middle one, since it's determined by the 2 others

look at the facerecognition module from the same site , or the one in opencv/contrib/facerec for hints how to do get the spacial histograms from the lbp patterns. openbr also has an ltp implementation, but i don't know it well enough.

berak gravatar imageberak ( 2013-09-19 08:17:42 -0500 )edit