WO2019104627A1 - Image gray scale classification method and device, and readable storage medium - Google Patents

Abstract

Disclosed in the present invention are an image gray scale classification method and device, and a readable storage medium. The method comprises: obtaining an image to be classified; pre-classifying at least part of pixels of the image according to gray scale values of the pixels to obtain at least two target classes and at least one intermediate class, each intermediate class being located between two target classes in the case of ranking according to the gray scale values of the pixels; and classifying the pixels in the intermediate class into one of two adjacent target classes according to the gray scale values and/or position information of the pixels in the intermediate class.

Description

Gray scale classification method and device for image and readable storage medium

[Technical Field]

The present invention relates to the field of image processing technologies, and in particular, to a grayscale classification method and apparatus for an image, and a readable storage medium.

 【Background technique】

In the image processing process, the target image is usually classified to achieve the purpose of scene segmentation and recognition. In the actual application process, for the more complicated scene, the image will be polished and blurred during the actual shooting process. The gray level classification cannot be correctly performed due to interference of series of problems such as noise.

At present, for the case where the target gray scale difference in the scene is large, the binarization method is usually adopted. For the case of uneven lighting or other interference factors, local dynamic classification and the like may be applied, for example, the maximum inter-class variance method. (OTSU); for scenarios where multiple classifications are required, clustering algorithms (such as kmeans) or improved binarization methods are often employed.

In the long-term research, the inventors found that for a pixel whose target gray scale is close to the segmentation threshold, the classification is inaccurate using the binarization method; and the clustering algorithm has a certain randomness, which is not suitable for accurate classification, and the improved binary value. The same method as the binarization method has the drawback of inaccurate classification.

 [Summary of the Invention]

The invention provides a method and a device for grading gray scale of an image, and a readable storage medium, so as to solve the problem that the gray scale is inaccurate due to interference of a series of problems such as lighting, blur, noise, etc. in the prior art. The problem of classification.

In order to solve the prior art problem, a technical solution adopted by the present invention is to provide a grayscale classification method for an image, the method comprising:

Obtain an image to be classified;

Pre-classifying the pixels according to a grayscale value of at least a portion of the pixels of the image to obtain at least two target classes and at least one intermediate class, each of the intermediate classes being located when sorted according to grayscale values of the pixels Between the target classes;

The pixels in the intermediate class are divided into one of two adjacent target classes according to grayscale values and/or positional information of the pixels in the intermediate class.

According to an embodiment of the invention, the pre-classifying the pixels according to the grayscale value of the pixels of at least part of the image comprises:

Using a clustering algorithm to divide the pixels into m classes according to grayscale values of at least a portion of pixels of the image, the m classes including n target classes and mn the intermediate classes, m and n are An integer and n < m ≤ 2n-1.

According to an embodiment of the present invention, the method includes: before using the clustering algorithm to divide the pixel into m classes according to grayscale values of pixels of at least part of the image, the method includes:

The initial cluster center in the clustering algorithm is determined by means of a gray histogram.

According to an embodiment of the present invention, the step of determining an initial clustering center in a clustering algorithm by using a gray histogram method includes:

Pixel points of the image to be classified are divided into m equal parts by using a gray histogram;

The average grayscale value of all of the pixels in all aliquots is calculated as the initial clustering center in the clustering algorithm.

According to an embodiment of the present invention, before the step of determining an initial cluster center in the clustering algorithm by using a gray histogram, the method further includes:

Smoothing all pixels of the image.

According to an embodiment of the present invention, the dividing the pixel in the intermediate class into one of two adjacent target classes according to a grayscale value of the pixel in the intermediate class comprises:

The pixel of the intermediate class is divided into one of a cluster center of two adjacent target classes and a grayscale value difference of the pixels of the intermediate class.

According to an embodiment of the present invention, the two adjacent target classes include a first target class and a second target class, and the grayscale value and location information of the pixel according to the intermediate class are Dividing the pixel in the intermediate class to one of two adjacent target classes includes:

If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the first target class is smaller than a difference from a cluster center of the second target class, and the middle And the number of the neighboring pixels of the pixel in the class belonging to the first target class is greater than the number of the second target class, and the pixel in the intermediate class is divided into the first target class ;

If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the second target class is smaller than a difference from a cluster center of the first target class, and the middle The number of the neighboring pixels of the pixel in the class that belong to the second target class is greater than the number of the first target class, and the pixels in the intermediate class are divided into the first target class ;

Otherwise it is determined that the pixels in the intermediate class are isolated pixel points.

According to an embodiment of the present invention, after determining that the pixel in the intermediate class is an isolated pixel, the method further includes:

The isolated pixel points are smoothed and then divided into one of the two adjacent target classes.

According to an embodiment of the invention, the neighborhood pixels of the pixel comprise pixels other than the pixels in the neighborhood of the pixel.

According to an embodiment of the present invention, the neighborhood of the pixel is an M*N-sized window centered on the pixel, where M and N are preset values, and are integers greater than 1.

According to an embodiment of the invention, the neighborhood of the pixel is divided according to edge information.

In order to solve the prior art problem, another technical solution adopted by the present invention is to provide an image gray scale classification device, including at least one processor, which operates separately or in cooperation, and the processor is configured to execute instructions to implement the foregoing Methods.

In order to solve the prior art problem, another technical solution adopted by the present invention is to provide a readable storage medium storing instructions which, when executed, implement the method described above.

The beneficial effects of the present invention are: different from the prior art, the gray-scale classification method of the image provided by the present invention is pre-classified, and then the pixels of the intermediate class are accurately classified into the target class according to the result of the pre-classification, thereby being able to accurately classify the pixels of the intermediate class into the target class. Effectively improve the accuracy of the classification, and the uneven lighting, noise and other problems can also solve the problem by using the classification method locally.

 [Description of the Drawings]

1 is a schematic flow chart of an embodiment of a gray scale classification method for an image of the present invention;

2 is a schematic flow chart of another embodiment of a gray scale classification method for an image of the present invention;

3 is a schematic structural diagram of an embodiment of an image gray scale classification device according to the present invention;

4 is a schematic block diagram of an embodiment of a readable storage medium of the present invention.

【Detailed ways】

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

Referring to FIG. 1 , FIG. 1 is a schematic flow chart of steps of an image gray scale classification method according to an embodiment of the present invention. The gray-scale classification method of the image provided by the embodiment of the present invention can be applied to a relatively complicated scene, so that gray-scale classification is correctly performed when the gray-scale classification is performed without interference of a series of problems such as lighting, blur, and noise. Methods include:

Step S101: Acquire an image to be classified.

It can be understood that the image to be classified may be a complete image, or may be a partial image taken from a complete image, for example, dividing a complete image to be processed into a plurality of windows, and selecting a partial image of at least one of the windows. As an image to be classified.

Step S102, pre-classifying the pixels according to the grayscale value of at least part of the pixels of the image to obtain at least two target classes and at least one intermediate class. When sorting according to the grayscale value of the pixels, each intermediate class is located between the two target classes. .

In an embodiment, pre-classifying the pixels according to the grayscale values of the pixels of at least part of the image comprises:

The clustering algorithm is used to classify the pixels into m classes according to the grayscale values of the pixels of at least part of the image, and m classes include n target classes and m-n intermediate classes, m and n are integers and n<m≤2n-1. For example, when m is 5, it may include 3 target classes and 2 intermediate classes, and may also include 4 target classes and 1 intermediate class.

It can be understood that the number of target classes and intermediate classes can be adjusted according to requirements, and an intermediate class can be set in every two target classes, or an intermediate class can be set in any two target classes, wherein the intermediate class The pixels in the picture will be divided into one of the two target classes adjacent to it, and the detailed division method is described below.

In this embodiment, the clustering algorithm refers to k-means. Algorithm, of course, in other embodiments, the clustering algorithm may be other algorithms, such as the k-medoids algorithm.

Step S103, the pixels in the intermediate class are divided into one of the adjacent two target classes according to the grayscale value and/or the position information of the pixels in the intermediate class.

For example, the value of m is 3, and the value of n is 2, that is, the pixels of the image to be classified are classified into three categories according to the grayscale value of the pixel, which are the first class, the second class, and the third class, respectively. Wherein the first class and the third class are target classes, and the second class is an intermediate class. In this case, according to the grayscale value of the pixel in the second class or/and the position information of the pixel point, the second class is Pixels are classified into the first or third class.

For another example, when m is 5, and n is 4, the pixels of the image to be classified are classified into five categories according to the grayscale value of the pixel, which are first, second, and third. Class, fourth class, and fifth class, wherein the first class, the second class, the third class, and the fifth class are target classes, and when the fourth class is an intermediate class, the grayscales of pixels according to the fourth class are The value or / and the position information of the pixel class classify the pixels in the fourth class into the third class or the fifth class, while the first class and the second class remain as the original pixels.

It can be understood that the selection of the intermediate class can be selected according to the corresponding situation.

It can be understood that the grayscale values of the pixels of the image are pre-classified by the clustering algorithm, so that the grayscale values of the pixels in the same class have a small difference in magnitude, and the grayscale values of the pixels in the same class are different. Large, and then by the precise classification of the pixels of the intermediate class, so that the grayscale value classification of the pixel is more accurate.

In an embodiment of the invention, dividing the pixels in the intermediate class into one of the adjacent two target classes according to the grayscale value of the pixels in the intermediate class comprises: dividing the pixels of the intermediate class into two adjacent pixels The difference between the grayscale value of the cluster center and the pixel of the intermediate class in the target class is smaller.

The clustering algorithm divides all pixels of the image into multiple clusters, and the grayscale values of the pixels in the same cluster are relatively close, while the grayscale values of the pixels in different clusters are relatively small. In this embodiment, the cluster center may be an average value of gray scale values of pixels in the cluster.

In another embodiment of the present invention, the two adjacent target classes include a first target class and a second target class, and the pixels in the intermediate class are divided into adjacent ones according to grayscale values and position information of pixels in the intermediate class. One of the two target classes includes:

If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the first target class is smaller than the difference between the cluster center of the second target class, and the neighborhood pixel of the pixel in the intermediate class belongs to the first If the number of target classes is greater than the number of the second target class, the pixels in the intermediate class are divided into the first target class;

If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the second target class is smaller than the difference between the cluster center of the first target class, and the neighborhood pixel of the pixel in the intermediate class belongs to the second If the number of target classes is greater than the number of the first target class, the pixels in the intermediate class are divided into the first target class;

Otherwise, the pixels in the intermediate class are determined to be isolated pixels.

The neighboring pixels of the pixel include pixels other than the pixel in the neighborhood of the pixel, and the neighborhood of the pixel is a window of M*N size centered on the pixel, where M and N are preset values, and An integer greater than one. The neighborhood of the pixel can also be divided according to the edge information.

In this embodiment, the location information specifically refers to the classification information of the neighboring pixels, that is, the number of the neighboring pixels belonging to the first target class and the number belonging to the second target class.

It can be understood that combining the grayscale values of the pixels and the position information for classification can ensure the edge information contained in all directions around the pixels to be classified, thereby further improving the accuracy of the classification.

In an embodiment of the present invention, after determining that the pixels in the intermediate class are isolated pixel points, the method further includes:

Smoothing isolated pixels and then dividing them into one of the two adjacent target classes.

The smoothing process may be one of mean filtering, median filtering, Gaussian filtering, or bilateral filtering.

It can be understood that after the isolated pixel is smoothed, it may be divided according to the grayscale value of the isolated pixel, or may be divided according to the grayscale value of the isolated pixel and the position information of the isolated pixel.

It can be understood that smoothing the isolated pixel points can effectively solve the problem of uneven lighting, noise, and the like of the pixel to be classified, so that the isolated pixel points are more accurate when classified again.

In an embodiment of the present invention, the clustering algorithm is used to divide the pixels into m classes according to the grayscale values of the pixels of at least part of the image, and the method includes:

The initial clustering center in the clustering algorithm is determined by using a gray histogram.

Referring to FIG. 2, the steps of determining the initial cluster center in the clustering algorithm by using a gray histogram method include:

Step S201, dividing the pixel points of the image to be classified into m equal parts by using a gray histogram;

Step S202, calculating an average grayscale value of all pixels in all the equal parts as an initial clustering center in the clustering algorithm.

It can be understood that the gray histogram represents the number of pixels having each gray scale value in the image, reflecting the frequency of occurrence of each gray scale in the image, and the abscissa indicates the gray scale value of each pixel in the image, and the ordinate. Indicates the frequency at which each pixel of the image appears on each grayscale value. Generally, in the clustering algorithm, the initial clustering center is arbitrarily selected, and in this embodiment, the pixels are divided into m equal parts by using a gray histogram, and the average grayscale value in each aliquot can be used as In the initial clustering center in the clustering algorithm, compared with the arbitrarily selected initial clustering center, determining the initial clustering center in the clustering algorithm by using the gray histogram can effectively improve the classification accuracy.

In an embodiment of the present invention, before the step of determining the initial cluster center in the clustering algorithm by using a gray histogram, the method further includes:

Smooth all pixels of the image.

It can be understood that before the pixels of the image are pre-classified, all the pixels are smoothed, and when the classification is performed later, the problem of uneven illumination and noise of all the pixels in the image can be solved, thereby effectively improving the accuracy of the classification. .

Referring to FIG. 3, an embodiment of a grayscale classification device for an image of the present invention includes a processor 110. Only one processor 110 is shown in the figure, and the actual number can be more. The processor 110 can work alone or in concert.

The processor 110 controls the operation of grayscale classification of images, and the processor 110 may also be referred to as a CPU (Central Processing). Unit, central processing unit). Processor 110 may be an integrated circuit chip with the processing capabilities of a signal sequence. The processor 110 can also be a general purpose processor, a digital signal sequence processor (DSP), an application specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware. Component. The general purpose processor may be a microprocessor or the processor or any conventional processor or the like.

The processor 110 is operative to execute instructions to implement a grayscale classification method of the image of the present invention.

Referring to FIG. 4, an embodiment of the readable storage medium of the present invention includes a memory 310 that stores instructions that, when executed, implement a grayscale classification method of an image of the present invention.

The memory 310 can include a read only memory (ROM, Read-Only) Memory), Random Access Memory (RAM), Flash Memory, hard disk, optical disk, etc.

Different from the prior art, the gray-scale classification method provided by the embodiment of the present invention first pre-classifies, and then accurately classifies the pixels of the intermediate class into the target class according to the result of the pre-classification, thereby effectively improving the accuracy of the classification, and playing Problems such as uneven light, noise, etc. can also solve the problem by using the classification method locally.

The above is only the embodiment of the present invention, and is not intended to limit the scope of the invention, and the equivalent structure or equivalent process transformation made by the specification and the drawings of the present invention may be directly or indirectly applied to other related technical fields. The same is included in the scope of patent protection of the present invention.

Claims (13)

1. A grayscale classification method for an image, the method comprising:

   Obtain an image to be classified;

   Pre-classifying the pixels according to grayscale values of at least part of the pixels of the image to obtain at least two target classes and at least one intermediate class. When the grayscale values of the pixels are sorted, each of the intermediate classes is located at two Between the target classes;

   The pixels in the intermediate class are divided into one of two adjacent target classes according to grayscale values and/or positional information of the pixels in the intermediate class.
2. The method of claim 1 wherein

   The pre-sorting the pixels according to a grayscale value of at least a portion of the pixels of the image includes:

   Using a clustering algorithm to divide the pixels into m classes according to grayscale values of at least a portion of pixels of the image, the m classes including n target classes and mn the intermediate classes, m and n are An integer and n < m ≤ 2n-1.
3. The method according to claim 2, wherein said using a clustering algorithm to divide said pixels into m classes according to grayscale values of at least a portion of said pixels of said image, said method comprising:

   The initial cluster center in the clustering algorithm is determined by means of a gray histogram.
4. The method according to claim 3, wherein the step of determining an initial cluster center in the clustering algorithm by using a gray histogram comprises:

   Pixel points of the image to be classified are divided into m equal parts by using a gray histogram;

   The average grayscale value of all of the pixels in all aliquots is calculated as the initial clustering center in the clustering algorithm.
5. The method according to claim 3, wherein before the step of determining the initial cluster center in the clustering algorithm by means of a gray histogram, the method further comprises:

   Smoothing all pixels of the image.
6. The method according to claim 1, wherein said dividing said pixels in said intermediate class into two adjacent said target classes according to a grayscale value of said pixels in said intermediate class One of them includes:

   The pixel of the intermediate class is divided into one of a cluster center of two adjacent target classes and a grayscale value difference of the pixels of the intermediate class.
7. The method according to claim 1, wherein said two adjacent target classes comprise a first target class and a second target class, said gray scale according to said pixels in said intermediate class The value and location information divides the pixels in the intermediate class into one of two adjacent target classes including:

   If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the first target class is smaller than a difference from a cluster center of the second target class, and the middle And the number of the neighboring pixels of the pixel in the class belonging to the first target class is greater than the number of the second target class, and the pixel in the intermediate class is divided into the first target class ;

   If the difference between the grayscale value of the pixel in the intermediate class and the cluster center of the second target class is smaller than a difference from a cluster center of the first target class, and the middle The number of the neighboring pixels of the pixel in the class that belong to the second target class is greater than the number of the first target class, and the pixels in the intermediate class are divided into the first target class ;

   Otherwise it is determined that the pixels in the intermediate class are isolated pixel points.
8. The method according to claim 7, wherein the determining that the pixel in the intermediate class is an isolated pixel further comprises:

   The isolated pixel points are smoothed and then divided into one of the two adjacent target classes.
9. The method of claim 7 wherein:

   The neighborhood pixels of the pixel include pixels other than the pixel in the neighborhood of the pixel.
10. The method of claim 9 wherein:

    The neighborhood of the pixel is an M*N-sized window centered on the pixel, where M and N are preset values and are integers greater than one.
11. The method of claim 9 wherein:

    The neighborhood of the pixel is divided according to edge information.
12. A grayscale classification device for an image, comprising at least one processor, operating alone or in cooperation, for executing instructions to implement the method of any of claims 1-11.
13. A readable storage medium storing instructions, wherein the instructions are executed to implement the method of any of claims 1-11.