Use FDtool (face detection tool) toolbox. You can download it from Matlab central.

thanks for reply Girish, but i need the explanation on my document image, now i am studying this method but at one stage i am struck, so i am asking....

you can find details in:

Article: Multimodal Biometric Recognition using Human Ear and Palmprint

thank you Nabil.

This is a very simple and straightforward implementation: https://github.com/carolinepacheco/lbplibrary/blob/master/package_lbp/olbp/OLBP.cpp

thank you Eanes Pereira. 

Satish, the main ideia is compare the intensity value of a central pixel with the intensities of those in neighborhood. Afterwards, creates a matrix changing values of intensities by the values of the bit chain converted to decimal after comparison. Then, you create a histogram of the values. If you want rotation invariance, you may label codes by the quantities of transitions from 1 to 0 and vice-versa. This way you have a more compact code and invariance to rotation and a little bit of invariance to scale.

thank you Eanes Pereira 

First, encode each pixel of your image into LBP code or LBPri code. This LBP encoded image will have same no. of pixels as your original image but each pixel will contain a LBP code value as its gray value. This LBP encoded image will show the edges of your original image. Now, build the histogram of this LBP encoded image. Then, extract the features like skewness, SD, energy etc from this histogram. 

Thanks PK Bhagat for your explanation.  

Thank you, Aseel O Ajlouni

Check the following:

http://www.cs.tut.fi/~mehta/drlbp

https://www.mathworks.com/help/vision/ref/extractlbpfeatures.html

Thank you,

LBP is a very well-established method, I believe you can find many tutorials and code online if you google. There are many LBP variations, the basic philosophy is kinda like convolution operations. The difference is in convolution we do straightforward summing/subtraction/dividation/multiplication of neighborhoods and put in the center point without any rule, but in LBP we have rules to select neighborhood.  The basic LBP code (both C++ and Matlab) can be downloaded form http://www.cse.oulu.fi/wsgi/CMV/Downloads/LBPSoftware.

Thank you, Saimunur.