CODEWITHSUNDEEP
matlabcomputer-visionbinaryfeature-extractionlbph-algorithm
keyserSoze
keyserSoze

Reputation: 135

Matlab LBP Real Valued

MATLAB's built in function to compute local binary patterns, from the computer vision toolbox, produces real valued numbers.
The function is named: extractLBPFeatures()

I understood the values from LBP to be binary - it's in the name! What is it doing to get real valued numbers rather than binary values?

Upvotes: 0

Views: 412

Answers (1)

Tonechas
Tonechas

Reputation: 13743

According to Matlab documentation, the call extractLBPFeatures(<your_image>) returns a 1-by-59 vector representing the L2-normalized histogram of the non rotation-invariant uniform LBPs, i.e. the relative frequencies of these 59 possible patterns throughout the image. The code of a non rotation-invariant uniform LBP formed by 8 neighbouring pixels is a whole number ranging from 0 to 58, whereas the 59 components of the feature vector yielded by extractLBPFeatures are floating point numbers in the range 0..1.

If you wish to obtain an image where the pixel intensity values are replaced by the LBP codes you could use the Matlab implementation developed by the creators of the LBP descriptor.

Upvotes: 0

Related Questions