Local Binary Pattern Algorithm

Question

I study the LBP algorithm and read the Face Detection and Verification document using local binary patterns, Y Rodriguez, which is the theme of PHD. On page 21 (section 2.2.2 weak classifiers), the author noted that:

A weak classifier hp (x) consists of a look-up table of 2^9 − 1 = 511 bins 

I can’t understand why there are 2 ^ 9-1 bins. So far I can understand that the LBP code for a particular pixel is 8 numbers, either 0 or 1, calculated from 8 surrounding pixels. However, why does the lookup table contain 2 ^ 9-1 bins? Any help is appreciated!

Answer 1

I suggest you read the article, “Multiresistant Grayscale and Rotational Invariant Texture Classification with Local Binary Patterns,” written by Timo Oyala, Matti Pietika Inen, Senior Member, IEEE and Topi Maenpan. You will find the answer in chapter 2.2. For example, a pixel is surrounded by 8 pixels, the lbp bit will be 256 (-1) = 1111 1111, the same one surrounded by a 9-pixel bin 512 (-1) = 1111 1111 1 ... you can think of a pixel surrounded by a circular chain formed by 9 pixels.