CN110751662A - Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means - Google Patents

Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means Download PDF

Info

Publication number
CN110751662A
CN110751662A CN201910998269.6A CN201910998269A CN110751662A CN 110751662 A CN110751662 A CN 110751662A CN 201910998269 A CN201910998269 A CN 201910998269A CN 110751662 A CN110751662 A CN 110751662A
Authority
CN
China
Prior art keywords
image
particle
segmentation
fuzzy
quantum
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910998269.6A
Other languages
Chinese (zh)
Other versions
CN110751662B (en
Inventor
赵晶
王晓莉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qilu University of Technology
Original Assignee
Qilu University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Qilu University of Technology filed Critical Qilu University of Technology
Priority to CN201910998269.6A priority Critical patent/CN110751662B/en
Publication of CN110751662A publication Critical patent/CN110751662A/en
Application granted granted Critical
Publication of CN110751662B publication Critical patent/CN110751662B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20172—Image enhancement details

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses an image segmentation method and system for a fuzzy C mean value optimized by quantum-behaved particle swarm, comprising the following steps: an acquisition step: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image; an image preprocessing step: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result; and (3) information entropy calculation: calculating the element information entropy of the image set I for the result after image enhancement; an image segmentation step: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing step.

Description

Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means
Technical Field
The disclosure relates to the technical field of image segmentation, in particular to an image segmentation method and system for optimizing a fuzzy C-means by quantum-behaved particle swarm.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
Images carry vivid and rich information and play an extremely important role in the age of multimedia information. The image may directly mimic or actually describe the objective presence of the thing. Image segmentation is an important pre-process for image recognition and computer vision. Without a correct segmentation, no correct identification is possible. This is a key step from image processing to image analysis. Image segmentation has an increasing demand in military, remote sensing, meteorology, communications, traffic, and medical image applications.
In the course of implementing the present disclosure, the inventors found that the following technical problems exist in the prior art:
generally, image segmentation decomposes a complete image into several regions with the same or different features and extracts objects of interest and technical processes from these regions. Various difficulties are encountered when computers automatically process partitions. Segmentation errors often occur, for example, due to uneven lighting, the effects of noise, the presence of unclear portions in the image, and shadows. Therefore, image segmentation is a technique that needs further investigation.
Disclosure of Invention
In order to solve the defects of the prior art, the disclosure provides an image segmentation method and system for optimizing a fuzzy C mean value by using quantum behavior particle swarm;
in a first aspect, the present disclosure provides an image segmentation method for a fuzzy C-means by quantum-behaved particle swarm optimization;
the image segmentation method for the fuzzy C mean value by quantum-behaved particle swarm optimization comprises the following steps:
an acquisition step: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image preprocessing step: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
and (3) information entropy calculation: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation step: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing step.
In a second aspect, the disclosure further provides an image segmentation system for quantum-behaved particle swarm optimization fuzzy C-means;
the image segmentation system of the fuzzy C mean value of quantum behavior particle swarm optimization comprises:
an acquisition module configured to: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image pre-processing module configured to: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
an information entropy calculation module configured to: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation module configured to: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing module.
In a third aspect, the present disclosure also provides an electronic device comprising a memory and a processor, and computer instructions stored on the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the steps of the method of the first aspect.
In a fourth aspect, the present disclosure also provides a computer-readable storage medium for storing computer instructions which, when executed by a processor, perform the steps of the method of the first aspect.
Compared with the prior art, the beneficial effect of this disclosure is:
the invention provides a middle-intelligence image segmentation method based on a quantum-behavior particle swarm optimization fuzzy C-means algorithm. The fuzzy control algorithm is optimized through the simple effectiveness of the fuzzy control algorithm and the global optimization capability of QPSO. The method is feasible and has good anti-noise capability through experimental verification, the overall segmentation effect of the method is better than that of FCM, the segmentation boundary is clearer, and the optimal result is selected without multiple experiments. The algorithm effectively solves the problems that the FCM has strong dependency on the initial value and is easy to fall into local optimization. The algorithm has strong global search capability and good image segmentation effect. Although good segmentation can be accurately achieved, the single time of program operation is not shortened significantly.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
FIG. 1 is a flow chart of a method according to a first embodiment of the present disclosure;
fig. 2(a) is a images (240 × 320) of a vessel according to a first embodiment of the present disclosure;
fig. 2(b) is a diagram of the effect of a mesology image with α mean operation and image enhancement operation according to the first embodiment of the disclosure;
fig. 2(C) is a graph of the fuzzy C-means segmentation effect based on fig. 2(b) according to the first embodiment of the present disclosure;
fig. 2(d) is a graph of the effect of the improved particle swarm optimization fuzzy C-means segmentation based on fig. 2(b) in the first embodiment of the disclosure.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiment provides an image segmentation method for a fuzzy C mean value by quantum-behaved particle swarm optimization;
as shown in fig. 1, the image segmentation method for fuzzy C-means by quantum-behaved particle swarm optimization includes:
an acquisition step: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image preprocessing step: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
and (3) information entropy calculation: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation step: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing step.
As one or more embodiments, the fuzzy C-means algorithm optimized by quantum-behaved particle swarm is used for segmenting the mesology image to obtain an image segmentation result; the method comprises the following specific steps:
s41: the method comprises the steps of firstly, obtaining an initial clustering class number C, an ambiguity parameter m, a particle swarm size N and a maximum iteration number MaxIt; the number of cluster centers is the dimension of each particle;
s42: carrying out initialization coding on the N cluster centers to form N first-generation particles; the number of the clustering centers is equivalent to the dimension of the particles; pbest for each particle is its current location, and gbest is the best location for all particles in the current population;
s43: calculating a center vector U (k) of each clustering center C (k) and membership degree;
s44: calculating the fitness of each particle; and if the fitness of the particle is better than the fitness of the current optimal position of the particle, updating the optimal position of the single particle. If the fitness of the current global optimal position is better than the fitness of the optimal positions in all the particles, updating the global optimal position;
s45: updating the location of each particle to generate a new population of particles;
s46: stopping iteration if the current iteration times reach the maximum times set previously; the best solution is found in the last generation, otherwise S43 is repeated.
Further, the step S45, updating the position of each particle by equations (30) - (33) to generate a new particle group:
Figure BDA0002240441220000051
Xi,j(t+1)＝pi,j(t)±α·|Cj(t)-Xi,j(t)|·ln[1/ui,j(t)]；ui,j(t)～U(0,1) (31)
the parameter C in equation (31), called the mean optimal position, can be denoted as mbest, is the center point of the individual optimal positions of all particles:
Figure BDA0002240441220000052
wherein p isi,j(t) is the potential well (local attraction point) of the jth dimension of the ith particle at the tth iteration, whose location is actually at the individual optimal location pbestj(t) and the population-optimal position gbest (t) are in the hyper-rectangle of vertices and vary as pbest and gbest vary. Phi is aj(t) and ui,j(t) are all t iterations with j dimension at [0,1 ]]Random numbers uniformly distributed, Xi,j(t +1) is the position of the ith particle in the jth dimension, C, at the tth iterationj(t) is a vector of C (t), α is the coefficient of contraction and expansion of QPSO, α is determined by equation (33):
α＝(α1-α2)*(MaxIt-t)/MaxIt+α2(33)
wherein, α1And α2Initial and final values of the parameter α, respectively, t is the current iteration number, MaxIt is the maximum number of iterations allowed, by changing the value of α, from 1.0 at the beginning of the search to 0.5 at the end of the search.
As one or more embodiments, the wisdom images include wisdom subset image T, wisdom subset image I, and wisdom subset image F:
the middle wisdom subset image T is expressed as the reality expression of the original image;
the image I of the Chinese wisdom set is expressed as uncertainty expression of the original image;
the wisdom subset image F is represented as a non-reality representation of the original image.
As one or more embodiments, the converting the image to be processed into a mesology image; the method comprises the following specific steps:
calculating the Zhongzhi subset image T according to the pixel value area mean value of the original image;
calculating the image I of the Zhongzhi subset according to the absolute value of the difference between the pixel value of the original image and the average value of the area pixels;
and calculating the wisdom subset image F according to the wisdom subset image T.
As one or more embodiments, the converting the image to be processed into a mesology image; the method comprises the following specific steps:
PNS＝{T,I,F} (1)
Figure BDA0002240441220000061
Figure BDA0002240441220000071
Figure BDA0002240441220000072
Figure BDA0002240441220000073
F(i,j)＝1-T(i,j) (6)
wherein, PNSIs a pixel point of the image in the NS domain;
t (i, j) is the value of the point (i, j) of the Zhongzhi subset image T;
i (I, j) is the value of the point (I, j) of the Zhongzhi subset image I;
f (i, j) is the value of the point (i, j) of the Zhongzhi subset image F;
Figure BDA0002240441220000074
is the mean of g (i, j) over the w × w region;
Figure BDA0002240441220000075
to represent
Figure BDA0002240441220000076
The minimum value of (a) is determined,
Figure BDA0002240441220000077
to representMaximum value of (d);
g (m, n) is the value of the point (m, n) of the Zhongzhi subset image T;
δ (i, j) is the mean value of pixel points g (i, j) and g (i, j) in the region w × w
Figure BDA0002240441220000079
The absolute value of the difference.
As one or more embodiments, the denoising processing is performed on the mesopic image; the method comprises the following specific steps:
Figure BDA00022404412200000710
Figure BDA00022404412200000714
Figure BDA00022404412200000715
Figure BDA0002240441220000081
wherein,
Figure BDA0002240441220000082
the pixels of the image in the NS domain pass through a set of points α -means,
Figure BDA0002240441220000083
α -mean subset of the noon subset images T,
Figure BDA0002240441220000084
for the α -mean subset of the wisdom subset image I,
Figure BDA0002240441220000085
α -mean subset for the mesoscopic subset image F, T is the set of true values of the original image,
Figure BDA0002240441220000086
for α -mean value set, I is the set of the uncertain values of the original image, α takes the value of 0.85, w is the size of the limited area, (I, j) is the pixel point of the original image, (m, n) is the pixel information point of the area in w, T (m, n) is the value of the point (m, n) of the image T of the Zhongzhi subset,
Figure BDA0002240441220000087
α -mean gray-scale mean intensity values are made for the wisdom subset image T,
Figure BDA0002240441220000088
α -mean set is needed, F (m, n) is the value of the point (m, n) of the image F of the Zhongzhi subset,
Figure BDA0002240441220000089
α -mean gray-scale mean intensity values are made for the noon-child set image F.
Figure BDA00022404412200000810
For the average intensity value of the image I of the wisdom subset,
Figure BDA00022404412200000811
t is obtained by performing α mean value operation on the pixel point (i, j)α(i, j) and
Figure BDA00022404412200000812
the absolute value of the difference between the two values,
Figure BDA00022404412200000813
and
Figure BDA00022404412200000814
is that
Figure BDA00022404412200000815
Minimum and maximum values of.
It should be understood that the purpose of the denoising step is to reduce noise points and high-uncertainty pixel points in the image, so that the distribution of image pixel information is more reasonable and beneficial to the subsequent processing of the image.
As one or more embodiments, the performing an image enhancement operation on the denoised result; the method comprises the following specific steps:
Figure BDA00022404412200000816
Figure BDA00022404412200000817
Figure BDA00022404412200000818
Figure BDA00022404412200000819
Figure BDA00022404412200000820
Figure BDA0002240441220000091
wherein:
Figure BDA0002240441220000093
the set of realisms for the wisdom set undergoing β enhancement operations,
Figure BDA0002240441220000094
is as follows
Figure BDA0002240441220000095
The wisdom set is subjected to β augmented reality subsets,
Figure BDA0002240441220000096
a subset of uncertainty values, representing the set of mesogens, β is a value of 0.85,
Figure BDA0002240441220000097
is composed of
Figure BDA0002240441220000098
The true value of the enhancement operation is performed β,
Figure BDA0002240441220000099
is the value of the set T after the mean operation,
Figure BDA00022404412200000910
the non-authenticity set subjected to β enhancement operations for the wisdom set,
Figure BDA00022404412200000911
is as follows
Figure BDA00022404412200000912
The medium wisdom set is subjected to β a non-authenticity subset of enhancement operations,
Figure BDA00022404412200000913
is composed of
Figure BDA00022404412200000914
A non-true value of the enhancement operation is performed β,
Figure BDA00022404412200000915
the non-real pixel points after the mean value operation,
Figure BDA00022404412200000916
is that
Figure BDA00022404412200000917
δ' (i, j) is the value of the pixel (i, j) after image enhancement operationAnd mean value
Figure BDA00022404412200000919
Absolute value of the difference, δ'minValue after image enhancement operation for pixel point (i, j)
Figure BDA00022404412200000920
And mean valueMinimum value of absolute value of the difference, δ'maxImage enhancement operation for pixel points (i, j)Post value
Figure BDA00022404412200000922
And mean value
Figure BDA00022404412200000923
The maximum value of the absolute value of the difference,as a subset
Figure BDA00022404412200000925
The average intensity in the w x w region,
Figure BDA00022404412200000926
as a subsetThe truth set after β enhancement operation in the w x w region delta' (i, j) is obtained after the image enhancement operation is carried out on the pixel points (i, j)
Figure BDA00022404412200000928
And
Figure BDA00022404412200000929
the absolute value of the difference.
It will be appreciated that the purpose of the enhancement processing step is to enhance the visual effect of the image to enlarge the feature differences between different objects in the image, such as objects and backgrounds, to improve the recognition of the image.
As one or more embodiments, calculating the element information entropy of the image set I as a result of the image enhancement; the method comprises the following specific steps:
Figure BDA00022404412200000930
wherein, EnIIs the entropy, p, of the image of the Zhongzhi subset II(i) Is the probability of element I in the wisdom subset image I.
In order to overcome the limitation of a general image segmentation method and improve the expression and processing capacity of image uncertainty information, an extension fuzzy theory of an interval value fuzzy set, an intuitive fuzzy set and an interval value intuitive fuzzy set is provided. The fuzzy theory uses a membership function to describe the degree of the elements belonging to a certain class, so that the ambiguity and the uncertainty in the segmentation process can be well expressed and processed. The theory of central intelligence is a new extended fuzzy theory, which summarizes the classical fuzzy theory and the related extended fuzzy theory.
The research of the middle wisdom is the relationship between a real thing, a theory, a proposition, a concept or an entity 'A' and the opposite 'Anti-A' thereof, the negation 'Non-A' thereof and the 'Netu-A' (being neither 'A' nor 'Anti-A'). The middle wisdom is the basis for the middle wisdom logic, the middle wisdom probability, the middle wisdom aggregate and the middle wisdom statistics.
The basic idea of neutron physics is that any view has a true value of t%, an uncertainty of i%, and a false value of f%.
The fuzzy segmentation technology generated based on the fuzzy theory can effectively describe the fuzziness of the image, and is very suitable for processing the uncertainty problem of accurate segmentation in the medical image. With the continuous improvement of the fuzzy theory, the application of the fuzzy theory in image segmentation is increasingly active, and the fuzzy theory becomes a research hotspot.
The main basis of image segmentation is the similarity and discontinuity of the gray levels. Image segmentation algorithms can be divided into region-based segmentation algorithms and boundary-based segmentation algorithms and segmentation methods incorporating certain theoretical tools. Region-based image segmentation algorithms separate objects from the background in the image by selecting an appropriate threshold. The method is simple and intuitive, but only considers the gray information of the image, does not consider the spatial information of the image, and is very sensitive to noise and gray nonuniformity. Edge-based image segmentation algorithms complete the segmentation of regions by detecting edges between different homogeneous regions. The difficulty of the edge detection method is the contradiction between noise immunity and detection accuracy. At present, a few improved multi-scale edge detection methods have been proposed, but the effect is still not ideal. The segmentation method combined with a specific theoretical tool is to use many theoretical methods at present for image segmentation, such as FCM method based on cluster analysis. The fuzzy C-means algorithm (FCM) based on the cluster analysis is more and more concerned by people due to the characteristics of simple design, wide solving range, easy realization of a computer and the like, and is widely applied in various fields. In addition, the conventional FCM algorithm does not consider spatial information during image segmentation, and is sensitive to noise and gray level non-uniformity.
The fuzzy C-means algorithm is widely applied due to simple realization; the fuzzy C-means algorithm is very sensitive to initial parameters and sometimes even to initial parameters. Manual intervention parameters are needed to approximate the global optimal solution, and the segmentation speed is improved. But it is limited by the initial cluster center selection, is easy to fall into local optimum and is more sensitive to noise. The classical improvement of FCM has two main aspects: one is an improvement to the objective function in FCM; and the other is to introduce other intelligent algorithms in the iterative solution process. In recent years, many scholars have tried various methods to solve the problem that the fuzzy C-means algorithm falls into local extrema due to improper initial value selection. If the FCM is combined with the particle swarm optimization algorithm, the image segmentation quality is improved to a certain extent.
Particle Swarm Optimization (PSO) algorithms were proposed by Kennedy and Eberhar in 1995 in the United states. PSO calculation is simple and easy to realize, but the flight speed of the particles cannot be effectively controlled in the later evolution stage, so that the algorithm is easy to fly and jump the optimal solution, the convergence speed of the algorithm is further caused, and the accuracy is reduced. In QPSO algorithm, each individual is regarded as a particle without weight and volume in a D-dimensional search space by QPSO, and can randomly appear at any position in the space with a certain probability, thereby realizing the global convergence of the algorithm. The algorithm has the advantages of small number of parameters and better searching capability than the developed PSO algorithm.
In order to solve the problems of the FCM algorithm, a quantum-behaved particle swarm optimization (QPSO) algorithm is applied to fuzzy C-means clustering.
The FCM obtains the membership degree of each data sample point to all class centers by optimizing an objective function, so that the attribution of the sample points is determined to achieve the purpose of automatically classifying the data samples. Since the FCM algorithm essentially finds the optimal solution by the gradient descent method, there is a problem of falling into the local optimal solution. Aiming at the situations, the particle swarm algorithm capable of ensuring global convergence is introduced, namely the quantum behavior particle swarm algorithm forms a fuzzy clustering algorithm based on evolutionary computation.
(1) Image for middle intelligence
Due to the fact that the classical fuzzy segmentation algorithm is difficult to obtain good results when the problems of complexity and high uncertainty are processed. Therefore, the intelligent theory is introduced into the image to generate the intelligent image so as to enhance the expression capability of uncertain information in the image.
The image is expressed in the field of middle intelligence, and the basic method is to convert the original digital image and the relevant features of the image into a middle intelligence area by utilizing the middle intelligence theory so as to obtain the middle intelligence image. The intelligent image can better express and describe the fuzziness and uncertainty of the image, and not only can utilize gray information, but also can utilize edge and space information to solve the problem which cannot be solved by a fuzzy segmentation algorithm.
Definition 1 (mesopic images): given an image as P, assume the full domain of discourse is U, the full set of image pixels is W, and W is a non-empty subset of U. PNSThe image has three membership sets T, I and F.
PNS＝{T,I,F} (1)
The subset image T is expressed as an original image authenticity expression, the subset image I is expressed as an original image uncertainty expression, and the subset image F is expressed as an original image non-authenticity expression.
The characteristic information used by the middle-intelligence image is the gray value of the image pixel point. A pixel P in the image is described as P (t, i, f), belonging to W for the element P (t, i, f) in the following way: t% true, I% uncertain, and F% false, where T ∈ T, I ∈ I, and F ∈ F. The pixel P (i, j) in the image domain is converted into an intellectual region.
In general, we represent the uncertainty of the data by the standard deviation. Since the uncertainty of the target region and the background region is much smaller than the uncertainty of the edge, this section defines I according to the standard deviation and discontinuity of the image gray scale. The standard deviation represents the difference in the local area of the image and the discontinuity represents an abrupt change in the gray level. The standard deviation reflects the difference of the local area of the image. The background and object regions in the image are uniform, while the blurred edges are gradual from object to background, and then we describe T, I, F using the mean, standard deviation, and range of pixel values.
In the process of image conversion, the pixel point (i, j) is limited in a small area, and the size of the area is w × w. The transformation process of the middle wisdom image is the acquisition process of the three sets T, I and F. The process is calculated by the following formula:
Figure BDA0002240441220000131
Figure BDA0002240441220000132
Figure BDA0002240441220000134
F(i,j)＝1-T(i,j) (6)
wherein:
Figure BDA0002240441220000135
is the local average of the image, and δ (i, j) is the pixel g (i, j) and its local averageThe absolute value of the difference.
(2) Information entropy of Zhongzhi image
The information entropy of the image describes the aggregation characteristics of the gray level distribution of the image pixel points. In the image of the middle intelligence, the information entropy of the image is used for describing the aggregation characteristics of the gray level distribution of image pixels.
Definition 2 (entropy of information of mesopic image) the entropy of information of a mesopic image is defined as the sum of the entropies of the three sets T, I and F, for evaluating the distribution of elements in a mesopic area:
EnNS＝EnT+EnI+EnF(7)
Figure BDA0002240441220000142
wherein: enT，EnI，EnFThe entropy of sets T, I and F, respectively, and the probability of element i in the three sets T, I and F, respectively. And EnTAnd EnFUsed to describe the distribution of elements in the mesology collection; enITo describe the distribution of the uncertainty of the elements.
The larger the entropy of the set T is, the smaller the entropy of the set I is, the more uniform the gray level distribution of the image is, and the more reasonable the distribution of the gray level values is, in order to increase the entropy of the set T and reduce the entropy of the set I, a new operation, α -mean value operation, is proposed in the noise removal of the mesopic image, and the operation utilizes the values of the pixel points in the uncertainty set to judge and update the current pixel value in the field of mesopic.
In order to solve the problem, the α mean filtering method is adopted by the present disclosure to eliminate the image noise, so as to facilitate the subsequent image segmentation.
Definition 3(α -mean operation) for PNSα -average value calculation,
Figure BDA0002240441220000144
Is defined as:
Figure BDA0002240441220000145
Figure BDA0002240441220000146
Figure BDA0002240441220000147
Figure BDA0002240441220000151
Figure BDA0002240441220000152
Figure BDA0002240441220000153
Figure BDA0002240441220000154
wherein:and
Figure BDA0002240441220000156
the mean values of T (i, j) and F (i, j) in the region w x w, respectively; deltaα(i, j) is T of the pixel point (i, j) after α mean operationα(i, j) and
Figure BDA0002240441220000157
the absolute value of the difference.
After α -mean operation, the entropy of the uncertain subset I is reduced, the distribution of elements in I becomes more uneven, and the uncertainty of the intelligent set PNS is reduced due to the unevenness.
(4) Image enhancement operations
In the process of the mesopic image segmentation, an image is converted into the mesopic image, the α mean value operation is carried out, the image outline can be blurred, and therefore the problem is solved by utilizing image enhancement in the processing processNSβ -enhancement operation of the T image to obtain
Figure BDA0002240441220000158
The blurred image contour becomes clearer, and the enhancement formula is
Figure BDA0002240441220000159
Figure BDA0002240441220000161
Figure BDA0002240441220000162
Figure BDA0002240441220000163
Figure BDA0002240441220000164
Figure BDA0002240441220000165
Wherein: delta' (i, j) is the pixel (i, j) after the image enhancement operation
Figure BDA0002240441220000166
And
Figure BDA0002240441220000167
the absolute value of the difference.
(5) Fuzzy C-means algorithm
The key to image segmentation based on the FCM clustering algorithm is how to unify the mathematical form of the image with that of the FCM clustering algorithm. For this purpose, the image may be regarded as a sample set, each pixel in the image may be regarded as a clustering sample, the feature of the pixel may be regarded as a feature vector of the sample, and the pixels may be clustered in the feature space. Dividing pixels with the same or similar characteristics into a class as much as possible; then, the class of each pixel is marked, so that the image segmentation is completed.
Aiming at the noon subset, a new clustering method and process are defined
Figure BDA0002240441220000168
α mean and noon subset after image enhancement operation.
Considering the influence of uncertainty, we combine the two sets T and I into a new cluster value.
Figure BDA00022404412200001610
An improved fuzzy c-means algorithm for the middle intelligent set is provided. The objective function is defined as:
Figure BDA0002240441220000171
Figure BDA0002240441220000172
Figure BDA0002240441220000173
the FCM algorithm obtains fuzzy classification of the sample set by performing iterative optimization on the objective function.
The algorithm is sensitive to the initial value and depends largely on the choice of initial cluster center. When the initial cluster center deviates significantly from the global optimal cluster center, the FCM tends to fall into a local minimum. The disadvantage is more pronounced when the number of clusters is larger.
(6) Improved PSO-FCM
The FCM algorithm based on gradient descent is essentially a local search algorithm, is prone to fall into local minimum values, and is sensitive to initial values, i.e., different initial values may result in different clustering results. The QPSO algorithm has global searching capability and is not easy to sink into a local area, so the present disclosure provides a new intermediate intelligent image segmentation hybrid clustering algorithm (QPSO-FCM), which combines the QPSO algorithm and the FCM to be applied to the intermediate intelligent image for segmentation.
QPSO operates on the fitness of individuals (particles) using the concepts of "population" and "evolution". Assuming that m is the number of particles in a multi-target search space, the position of the first particle is represented as vector Xi ═ (Xi1, Xi2, …. xiD), in each iteration, the particle updates itself by tracking the two best positions. One is the optimal solution found by the particle itself, called the single optimal position pi (pi1, pi2, …. pid); the other is the optimal solution found by the whole particle swarm optimization at present, and is called as the global optimal position pg. After finding the two extreme values, the average optimal position (mbest) or c (t) is taken as the average of the optimal positions of all particles. Move to search for the best solution to the problem according to the following formula.
Figure BDA0002240441220000181
Xi,j(t+1)＝pi,j(t)±α·|Cj(t)-Xi,j(t)|·ln[1/ui,j(t)]ui,j(t)～U(0,1) (31)
pi,j(t) is the individual optimum position of the ith particle at the tth iteration (1)<＝j<D) α is the contraction expansion coefficient of QPSO, an important parameter for QPSO convergence, and the value α is determined by the following equation:
α＝(α1-α2)*(MaxIt-t)/MaxIt+α2(33)
in the formula, α1And α2The initial and final values of the parameter, respectively, t is the current iteration number, MaxIt is the maximum number of iterations allowed, by varying value, from 1.0 at the beginning of the search to 0.5 at the end of the search.
In the application of the QPSO-FCM algorithm, the fitness function of each individual in the QPSO is defined as follows:
Figure BDA0002240441220000183
wherein, Jm(U, C) is the fitness function of the image. k is the gray scale, k is 0-C, and C is the maximum value of the gray scale.
The QPSO-based FCM algorithm randomly generates a next generation solution, and therefore, it is easy to search for a global optimum value. Moreover, each generation of solution has the dual advantages of self-promotion and learning to others; therefore, it has a fast convergence speed.
(7) Intelligent image segmentation based on QPSO-FCM algorithm
A new intermediate intelligent image segmentation algorithm-QPSO-FCM algorithm is provided, which comprises the following steps:
step 1: inputting an image;
step 2: transforming the image into the NS domain using equations (1) - (6);
step 3, α mean value operation is executed by using equations (11) - (17);
and 4, step 4: performing an image enhancement operation using equations (18) - (24);
and 5: calculating the entropy of the indefinite subset I using equation (9);
step 6: if it is not
Figure BDA0002240441220000191
Entering step 7; otherwise, entering step 3;
and 7: applying the improved FCM algorithm to the set of nooses:
Figure BDA0002240441220000192
the initial clustering class number C, the ambiguity parameter m, the particle swarm size N and the maximum iteration number MaxIt. The number of cluster centers is the dimension of each particle.
Figure BDA0002240441220000193
And carrying out initialization coding on the N cluster centers to form N first-generation particles. The number of cluster centers corresponds to the dimension of the particle. Pbest for each particle is its current location, and gbest is the best location for all particles in the current population.
Figure BDA0002240441220000194
Calculating a center vector U (k) of each cluster center C (k) and the degree of membership in k steps using equations (28) - (29);
Figure BDA0002240441220000195
calculating the fitness of each particle according to equation (34)Degree of reaction. And if the fitness of the particle is better than the fitness of the current optimal position of the particle, updating the optimal position of the single particle. If the fitness of the current global optimal position is better than the fitness of the optimal positions in all the particles, the global optimal position is updated.
Figure BDA0002240441220000196
The position of each particle is updated with equations (30) - (33) to generate a new population of particles.
Figure BDA0002240441220000197
If the current number of iterations reaches the previously set maximum number, the iteration is stopped. Finding the best solution in the last generation, otherwise repeating step (3).
And 8: and obtaining an image segmentation result.
The QPSO-FCM segmentation algorithm is applied to the mesopic image subjected to denoising of various real images, and is compared with the mesopic denoising image segmentation method based on the fuzzy C mean value.
In the FCM image segmentation method and the QPSO-FCM image segmentation method, the clustering centers set by the invention are both 5, so that the clustering centers are compared with each other.
As can be seen from fig. 2(a) -2 (d), the warship boundary is clearer in fig. 2(d) compared with fig. 2 (c). Especially at the front of the ship, the boundaries of the section of fig. 2(d) are more clear. The warship boundary in fig. 2(c) is blurred.
From the experimental data of fig. 2(a) -2 (d), it can be seen that the QPSO and FCM are applied to the mesology image more clearly than the FCM is applied to the mesology image.
The ultimate goal of image segmentation is to accurately acquire the target image in one run. Therefore, we need to constantly find new methods to test whether this method is good or bad. As can be seen from the data obtained in a large number of experiments, FCM is greatly affected by the initial value, so we can only test whether a global optimum can be found. In many experiments we can find the best performing image (the border is clearer). The QPSO-FCM method is applied to the intelligent image, the global optimum can be found through one-time operation, the segmentation boundary is clearer, the processing of the fuzzy area is greatly improved, and the global optimum is the same even if the experiment is repeated.
However, the running time of the program is not shortened significantly because the data of the method needs to be calculated repeatedly. According to the free lunch theorem, the good and accurate segmentation effect is obtained at the cost of spending more running time.
The second embodiment further provides an image segmentation system for quantum-behaved particle swarm optimization fuzzy C-means;
the image segmentation system of the fuzzy C mean value of quantum behavior particle swarm optimization comprises:
an acquisition module configured to: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image pre-processing module configured to: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
an information entropy calculation module configured to: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation module configured to: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing module.
In a third embodiment, the present embodiment further provides an electronic device, which includes a memory, a processor, and computer instructions stored in the memory and executed on the processor, where the computer instructions, when executed by the processor, implement the steps of the method in the first embodiment.
In a fourth embodiment, the present embodiment further provides a computer-readable storage medium for storing computer instructions, and the computer instructions, when executed by a processor, perform the steps of the method in the first embodiment.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims (6)

1. The image segmentation method for the fuzzy C mean value by quantum-behaved particle swarm optimization is characterized by comprising the following steps of:
an acquisition step: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image preprocessing step: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
and (3) information entropy calculation: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation step: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing step.
2. The method as claimed in claim 1, wherein the fuzzy C-means algorithm using quantum-behaved particle swarm optimization is used for segmenting the mesopic image to obtain an image segmentation result; the method comprises the following specific steps:
s41: the method comprises the steps of firstly, obtaining an initial clustering class number C, an ambiguity parameter m, a particle swarm size N and a maximum iteration number MaxIt; the number of cluster centers is the dimension of each particle;
s42: carrying out initialization coding on the N cluster centers to form N first-generation particles; the number of the clustering centers is equivalent to the dimension of the particles; pbest for each particle is its current location, and gbest is the best location for all particles in the current population;
s43: calculating a center vector U (k) of each clustering center C (k) and membership degree;
s44: calculating the fitness of each particle; if the fitness of the particle is better than the fitness of the current optimal position of the particle, updating the optimal position of a single particle; if the fitness of the current global optimal position is better than the fitness of the optimal positions in all the particles, updating the global optimal position;
s45: updating the location of each particle to generate a new population of particles;
s46: stopping iteration if the current iteration times reach the maximum times set previously; the best solution is found in the last generation, otherwise S43 is repeated.
3. The method of claim 2, wherein said step S45, updating the position of each particle using equations (30) - (33) to generate a new population of particles:
Figure FDA0002240441210000021
Xi,j(t+1)＝pi,j(t)±α·|Cj(t)-Xi,j(t)|·ln[1/ui,j(t)]；ui,j(t)～U(0,1) (31)
wherein p isi,j(t) is the potential well in the jth dimension of the ith particle at the tth iteration, whose location is actually at the individual optimal location pbestj(t) and the population-optimal position gbest (t) are in the hyper-rectangle of vertices and vary with pbest and gbest; phi is aj(t) and ui,j(t) are all t iterations with j dimension at [0,1 ]]Random numbers uniformly distributed, Xi,j(t +1) is the position of the ith particle in the jth dimension, C, at the tth iterationj(t) is a vector of C (t), α is the coefficient of contraction and expansion of QPSO, α is determined by equation (33):
α＝(α1-α2)*(MaxIt-t)/MaxIt+α2(33)
wherein, α1And α2An initial value and a final value of the parameter α, respectively, t being the current iteration numberNumber, MaxIt, is the maximum number of iterations allowed, from 1.0 at the beginning of the search to 0.5 at the end of the search, by changing the value of α.
4. The image segmentation system of fuzzy C mean value of behavior particle swarm optimization of quantum, characterized by, include:
an acquisition module configured to: acquiring an image to be processed, and converting the image to be processed into a Zhongzhi image;
an image pre-processing module configured to: denoising the intermediate intelligent image, and then performing image enhancement operation on a denoised result;
an information entropy calculation module configured to: calculating the element information entropy of the image set I for the result after image enhancement;
an image segmentation module configured to: if the ratio of the information entropies of the adjacent elements is smaller than a set threshold, carrying out segmentation on the mesology image by using a fuzzy C-means algorithm optimized by quantum-behaved particle swarm to obtain an image segmentation result; otherwise, returning to the image preprocessing module.
5. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executable on the processor, the computer instructions when executed by the processor performing the steps of the method of any of claims 1 to 3.
6. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the steps of the method of any one of claims 1 to 3.
CN201910998269.6A 2019-10-17 2019-10-17 Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means Active CN110751662B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910998269.6A CN110751662B (en) 2019-10-17 2019-10-17 Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910998269.6A CN110751662B (en) 2019-10-17 2019-10-17 Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means

Publications (2)

Publication Number Publication Date
CN110751662A true CN110751662A (en) 2020-02-04
CN110751662B CN110751662B (en) 2022-10-25

Family

ID=69278997

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910998269.6A Active CN110751662B (en) 2019-10-17 2019-10-17 Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means

Country Status (1)

Country Link
CN (1) CN110751662B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114943701A (en) * 2022-05-20 2022-08-26 南通鼎彩新材料科技有限公司 Intelligent control system for granulation equipment of heat-shrinkable tube

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108898602A (en) * 2018-06-27 2018-11-27 南京邮电大学 A kind of FCM medical image cutting method based on improvement QPSO
US20190163950A1 (en) * 2017-11-30 2019-05-30 Metal Industries Research & Development Centre Large scale cell image analysis method and system
CN110111343A (en) * 2019-05-07 2019-08-09 齐鲁工业大学 A kind of middle intelligence image partition method and device based on improvement fuzzy C-mean algorithm

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190163950A1 (en) * 2017-11-30 2019-05-30 Metal Industries Research & Development Centre Large scale cell image analysis method and system
CN108898602A (en) * 2018-06-27 2018-11-27 南京邮电大学 A kind of FCM medical image cutting method based on improvement QPSO
CN110111343A (en) * 2019-05-07 2019-08-09 齐鲁工业大学 A kind of middle intelligence image partition method and device based on improvement fuzzy C-mean algorithm

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李引: "聚类算法的研究与应用", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑（月刊）》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114943701A (en) * 2022-05-20 2022-08-26 南通鼎彩新材料科技有限公司 Intelligent control system for granulation equipment of heat-shrinkable tube
CN114943701B (en) * 2022-05-20 2024-06-14 南通鼎彩新材料科技有限公司 Intelligent control system of granulating equipment for heat-shrinkable tube

Also Published As

Publication number Publication date
CN110751662B (en) 2022-10-25

Similar Documents

Publication Publication Date Title
Vu et al. Softgroup for 3d instance segmentation on point clouds
Sathya et al. PSO-based Tsallis thresholding selection procedure for image segmentation
Ducournau et al. Random walks in directed hypergraphs and application to semi-supervised image segmentation
CN108389211B (en) Image segmentation method based on improved whale optimized fuzzy clustering
CN107633226B (en) Human body motion tracking feature processing method
Syu et al. Hierarchical image segmentation based on iterative contraction and merging
Mirghasemi et al. A target-based color space for sea target detection
CN110322445A (en) A kind of semantic segmentation method based on maximization prediction and impairment correlations function between label
CN110111343A (en) A kind of middle intelligence image partition method and device based on improvement fuzzy C-mean algorithm
Wang et al. Pixel classification based color image segmentation using quaternion exponent moments
CN106570874A (en) Image marking method combining local image constraint and overall target constraint
Wang et al. Energy based competitive learning
CN109165698A (en) A kind of image classification recognition methods and its storage medium towards wisdom traffic
Yang et al. Color texture segmentation based on image pixel classification
Majeed et al. Uncertain fuzzy self-organization based clustering: interval type-2 fuzzy approach to adaptive resonance theory
Kumar et al. A novel approach for multi-cue feature fusion for robust object tracking
CN115984223A (en) Image oil spill detection method based on PCANet and multi-classifier fusion
Sun et al. Visual tracking via joint discriminative appearance learning
CN105160666B (en) SAR image change detection based on Non-Stationary Analysis and condition random field
CN110751662B (en) Image segmentation method and system for quantum-behaved particle swarm optimization fuzzy C-means
Firouznia et al. Adaptive chaotic sampling particle filter to handle occlusion and fast motion in visual object tracking
Tang et al. General scheme of region competition based on scale space
CN116452940A (en) Three-dimensional instance segmentation method and system based on dense and sparse convolution fusion
WO2023193026A1 (en) Systems and methods for graph-based active learning using semi-supervised classification of sar data
Chacon-Murguia et al. Adaptive fuzzy weighted color histogram and HOG appearance model for object tracking with a dynamic trained neural network prediction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CP03 Change of name, title or address

Address after: 250353 University Road, Changqing District, Ji'nan, Shandong Province, No. 3501

Patentee after: Qilu University of Technology (Shandong Academy of Sciences)

Country or region after: China

Address before: 250353 University Road, Changqing District, Ji'nan, Shandong Province, No. 3501

Patentee before: Qilu University of Technology

Country or region before: China