Description of the description, principle and feature vector extraction of LBP features

LBP (Local Binary Pattern) is an operator used to describe the local texture features of an image; it has significant advantages such as rotation invariance and gray invariance. It was first proposed by T. Ojala, M. Pietikäinen, and D. Harwood in 1994 for texture feature extraction. Moreover, the extracted features are local texture features of the image;

1. Description of LBP features

The original LBP operator is defined as a window within the 3*3 window, with the window center pixel as a threshold, and the gray value of the adjacent 8 pixels is compared with it. If the surrounding pixel value is greater than the central pixel value, the pixel point is The position is marked as 1, otherwise 0. In this way, 8 points in the 3*3 neighborhood can be compared to produce an 8-bit binary number (usually converted to a decimal number, ie, LBP code, a total of 256 types), that is, the LBP value of the center pixel of the window is obtained, and this value is used to reflect The texture information for this area. As shown below:

Description of the description, principle and feature vector extraction of LBP features

Improved version of LBP:

After the original LBP was proposed, the researchers continued to propose various improvements and optimizations.

(1) Circular LBP operator:

The biggest drawback of the basic LBP operator is that it covers only a small area within a fixed radius, which obviously does not meet the needs of different size and frequency textures. In order to adapt to the texture features of different scales and achieve the requirements of gray scale and rotation invariance, Ojala et al. improved the LBP operator, extended the 3×3 neighborhood to any neighborhood, and replaced the square neighbor with a circular neighborhood. Domain, the improved LBP operator allows for any number of pixels within a circular neighborhood of radius R. Thereby obtaining an LBP operator containing P sampling points in a circular area such as a radius R;

Description of the description, principle and feature vector extraction of LBP features

(2) LBP rotation invariant mode

As can be seen from the definition of LBP, the LBP operator is gray-scale invariant, but it is not rotation-invariant. The rotation of the image will result in different LBP values.

Maenpaa et al. extended the LBP operator and proposed an LBP operator with rotation invariance, that is, continuously rotating the circular neighborhood to obtain a series of initially defined LBP values, taking the minimum value as the LBP value of the neighborhood. .

Figure 2.5 shows the process of obtaining the LBP with invariant rotation. The number below the operator in the figure indicates the LBP value corresponding to the operator. The eight LBP modes shown in the figure are processed by rotation and invariably. The obtained LBP value of rotation invariance is 15. That is to say, the rotation-invariant LBP modes corresponding to the eight LBP modes in the figure are all 00001111.

Description of the description, principle and feature vector extraction of LBP features

(3) LBP equivalent mode

An LBP operator can generate different binary modes. For LBP operators with P samples in a circular region of radius R, P2 modes will be generated. Obviously, as the number of sample points in the neighborhood set increases, the type of binary pattern increases dramatically. For example: 20 sample points in the 5 × 5 neighborhood, there are 220 = 1, 048, 576 binary modes. So many binary modes are disadvantageous for texture extraction and for texture recognition, classification and access to information. At the same time, too many pattern types are detrimental to the expression of the texture. For example, when the LBP operator is used for texture classification or face recognition, the statistical histogram of the LBP mode is often used to express the information of the image, and more pattern types will make the data amount too large, and the histogram is too sparse. Therefore, it is necessary to perform dimensionality reduction on the original LBP mode, so that the information of the image can be best represented in the case where the amount of data is reduced.

In order to solve the problem of too many binary patterns and improve statistics, Ojala proposed to use a "Uniform Pattern" to reduce the pattern of LBP operators. Ojala et al. believe that in the actual image, most LBP modes only contain two transitions from 1 to 0 or from 0 to 1. Therefore, Ojala defines the "equivalent mode" as: when the cyclic binary number corresponding to an LBP has two hops from 0 to 1 or from 1 to 0, the binary corresponding to the LBP is called an equal. Price model class. Such as 00000000 (0 times jump), 00000111 (including only one jump from 0 to 1), 10001111 (first jump from 1 to 0, then jump from 0 to 1, a total of two jumps) are equivalent modes class. Patterns other than the equivalence mode class are classified as another class, called a mixed mode class, such as 10010111 (a total of four hops) (this is my personal understanding, I don't know if it is right).

With such improvements, the variety of binary patterns is greatly reduced without losing any information. The number of modes is reduced from the original 2P to P (P-1) + 2, where P represents the number of samples in the neighborhood set. For 8 sample points in the 3 × 3 neighborhood, the binary mode is reduced from the original 256 to 58, which makes the feature vector have fewer dimensions and can reduce the impact of high frequency noise.

2, the principle of LBP features for detection

Obviously, the above extracted LBP operator can get an LBP "encoding" at each pixel, then extract the original LBP operator for an image (recording the gray value of each pixel). After that, the original LBP feature obtained is still "a picture" (the LBP value of each pixel is recorded).

In LBP applications, such as texture classification and face analysis, LBP maps are generally not used as feature vectors for classification and recognition, but statistical histograms of LBP feature spectra are used as feature vectors for classification and recognition.

Because, from the above analysis we can see that this "feature" is closely related to the location information. Extracting such "features" directly from the two images and performing discriminant analysis will cause large errors due to "position misalignment". Later, the researchers found that a picture can be divided into several sub-areas, LBP features are extracted for each pixel in each sub-area, and then a statistical histogram of LBP features is established in each sub-area. In this way, each sub-area can be described by a statistical histogram; the entire picture is composed of several statistical histograms;

For example: a picture of 100*100 pixels, divided into 10*10=100 sub-areas (the area can be divided by various ways), each sub-area has a size of 10*10 pixels; each in each sub-area Pixels, extract their LBP features, and then create a statistical histogram; thus, this image has 10 * 10 sub-regions, and there are 10 * 10 statistical histograms, using these 10 * 10 statistical histograms , you can describe this picture. After that, we can use various similarity measure functions to judge the similarity between the two images.

3. Steps for extracting the LBP feature vector

(1) First, the detection window is divided into 16×16 small cells;

(2) For one pixel in each cell, the gray value of the adjacent 8 pixels is compared with it, and if the surrounding pixel value is greater than the central pixel value, the position of the pixel is marked as 1, otherwise 0 . In this way, the 8 points in the 3*3 neighborhood are compared to generate an 8-bit binary number, that is, the LBP value of the central pixel of the window is obtained;

(3) Then calculate the histogram of each cell, that is, the frequency at which each number (assumed to be a decimal number LBP value) appears; then normalize the histogram.

(4) Finally, the obtained statistical histogram of each cell is connected into a feature vector, that is, the LBP texture feature vector of the entire image; then it can be classified by SVM or other machine learning algorithms.

Three Phase AC Source System

APM offers 3 phase ac power source system. It`s single 3-phase output programmable AC Power Supply which provides with high power density. With high speed DSP+CPLD control, high frequency PWM technology, active PFC design, It is able to provide not only stable DC/AC output power,but also 3-phase / 1-phase output.

The 3 phase ac power supply is featured with high power density, high reliability and high precision, meanwhile it possesses operation interface of touch screen and keys manually. It is able to analog output normal or abnormal power input for electrical device to meet test requirements, which is applicable to electric, lighting, aviation sectors, etc. It could be applied to enterprise`s production test as well.

Some features as below:


  • 5.6" large touch color screen, possess complete functions and easy to operate.
  • Support for USB data import/export and screen snap from front panel.
  • AC+DC mixed or independent output mode for voltage DC offset simulation.
  • Capable of setting voltage and current output restriction, support for constant current output mode.
  • Capable of setting output slope of voltage and frequency.
  • Capable of setting ON/OFF phase angle.
  • With reverse current protection to avoid current flowing backward.
  • Built-in power meter, which is capable of measuring 5 electrical parameters per phase, including voltage, current, power, etc.
  • Support mA current measurement function.


Three Phase AC Source System,Programmable 3 Phase Ac Power Supply,Ac Source System,Three Phase Source

APM Technologies Ltd , https://www.apmpowersupply.com