In local

binary pattern the original operator of LBP labeling the pixels of the needed

image with decimal numbers, the produced decimal numbers called Local Binary

Patterns or LBP codes, this happens with encoding the local structure around

each pixel. Like it showed in the figure before. Each pixel is compared with

its eight neighbors in a 3×3 neighborhood by subtracting the center pixel

value; if the value of the center pixel is less than the pixel of the

neighborhood it will be encoded with 0 otherwise it will be encoded with 1; After

that A binary number is acquired by linking all the binary codes we get in a

clockwise direction. We starts from the top-left one, then to label we use their

corresponding decimal value. The binary numbers we get called Local Binary

Patterns or LBP codes.

The small

3×3 neighborhood of the basic LBP operator cannot catch or capture the dominant

features with large scale structures and this is one of the restrictions of the

basic LBP operator. To deal with the texture at different scales, the operator

was later generalized to use neighborhoods of different sizes 1. Though what

is a local neighborhood? Local neighborhood can be defined as

a set of sampling points evenly spaced on a circle which is centered at the

pixel to be labeled, and the sampling points that do not fall within the pixels

are interpolated using bilinear interpolation, thus allowing for any radius and

any number of sampling points in the neighborhood. Fig. 2 shows some examples

of the extended LBP operator. For neighborhoods we will use the notation

(P, R) which means P sampling points on a circle of radius of ROfficially, given a pixel at (), the resulting LBP can be

expressed in decimal form as:) = where and are notations of gray-level values of the

central pixel and P surrounding pixels in the circle neighborhood with a radius

R respectively, and function F(x) is defined as: From the definition above, the basic LBP operator is invariant to

monotonic gray-scale transformations preserving pixel intensity order in the

local neighborhoods. The histogram of LBP labels calculated over a region can

be inviting as a texture descriptor.The operator which produces different output values, corresponding to different binary patterns formed by P pixels

in the neighborhood. If the image is rotated, these surrounding pixels in each

neighborhood will move correspondingly along the perimeter of the circle, this

will result in a different LBP value, except patterns with only 1s and 0s. In

order to remove rotation effect, a rotation-invariant LBP is proposed in:=

min Where ROR(x,) referred

to the circular bit-wise right shift on the P-bit number x times. The

operator quantifies occurrence statistics of individual rotation invariant

patterns corresponding to certain micro-features in the image; for this reason,

in some articles they consider the patterns as a feature detector 1. Nevertheless,

in some articles it was shown that such a rotation-invariant LBP operator does

not necessarily provide discriminative information, since the occurrence

frequencies of the individual patterns incorporated in vary

greatly and the simple quantization of the angular spaces at 45° intervals.

From studies it has been shown that particular patterns contain

more information than others. It is possible only to use a subset of binary patterns for describing the texture of

the images. Ojala et al. refers to these patterns as uniform patterns and denoted

by.

A local binary pattern is called uniform if it contains at most two bitwise

transitions from 0 to 1 or vice versa when the corresponding bit string is considered

circular. For example, 00000000 (0 transitions) and 01110000 (2 transitions)

are both uniform whereas 11001001 (4 transitions) and 01010011 (6 transitions)

are not. From observations it is shown that the uniform patterns account for

around 90% of all the patterns in a (8, 1) neighborhood and around 70% in a

(16, 2) neighborhood in texture images 1. A similar experiment was conducted

on the FERET database, and it was found that 90.6% of the patterns in a (8, 1)

neighborhood and 85.2% in a (8, 2) neighborhood are uniform 20. More

recently, Shan and Gritti verified validity of uniform patterns for

representing faces from the viewpoint of machine learning. Specifically, they

applied AdaBoost to select the discriminative patterns for facial expression

recognition, and their experiments demonstrated that, using LBP (8, 2)

operator, 91.1% of these selected patterns are uniform. Accumulating the

non-uniform patterns into a single bin yields an LBP operator with less than labels. For instance, the number of labels

with the neighborhood of 8 pixels is 256 for the standard LBP but only 59 for.

Also it should be noted that, at about the same time that the original LBP

operator was proposed, a method by Zabih and Woodfill introduced which is a

Census Transform (CT) method 42. This method is very similar to LBP. CT also

maps the local neighborhood surrounding a pixel onto a binary string, and the

only difference between LBP and CT is the opposite order of bit string.