The local binary pattern is easy, intuitive, and quite simple to compute. These are some very rare qualities of a feature descriptor! Let's first see what local binary pattern means and how it is computed (on paper, we will come to the implementation soon). After we are done explaining the LBP operator using code samples, we will also try to develop an intuition regarding the type of information that it captures from images. Such an understanding would be crucial when you want to decide whether to use LBP features for a particular problem or not.
At the most basic level, the LBP operator assigns a number between 0 to 255 (inclusive) to every pixel in the input image. After this assignment is made, we construct a histogram out of the values having 256 bins-one for each possible value.
First, let's get into the details of how a number is assigned to every pixel. For a given pixel in the input image, we consider a 3 x 3 neighborhood of that pixel location...