CN114359200A - Image definition evaluation method based on pulse coupling neural network and terminal equipment - Google Patents

Abstract

The invention relates to an image processing method, in particular to an image definition evaluation method based on a pulse coupled neural network and a terminal device, which apply the pulse coupled neural network to solve the technical problem of low image definition, by setting relevant parameters of a feedback input domain, a coupling connection domain and a pulse generation domain of the pulse coupling neural network in advance, circularly iterating, calculating an image definition evaluation function value corresponding to an image sequence obtained in the focusing process, and drawing a curve chart to represent the definition transformation trend corresponding to the image sequence, can be suitable for different scene types, and only the gray information of the image is concerned, so that the operation accuracy can be improved, the interference of other factors is avoided, the real-time performance is better, and the calculation efficiency is high to meet the engineering requirement when the terminal equipment runs.

Description

Image clarity evaluation method and terminal equipment based on pulse coupled neural network

technical field

The invention relates to an image processing method, in particular to an image definition evaluation method and terminal equipment based on a pulse coupled neural network.

Background technique

During the focusing process, the mechanical structure of the optical system may not be able to locate the focus due to various reasons, and the focus position of the system that has completed focusing may also drift, resulting in poor image quality due to defocusing. Therefore, it is necessary to realize the automatic focusing function. Autofocus technology is widely used in many fields. The current research focus is mainly on the autofocus method based on image processing. Control; using the image sharpness evaluation function to characterize the focusing degree of the optical system, and using the image sharpness evaluation function to quickly and accurately calculate the image sharpness is the key to image processing; at the same time, limited by the inherent defects of the mechanical structure, the lens cannot Accurately reach the best position of the image focus, or the image quality is still poor when it is already at the focus position, so an image sharpness evaluation method is needed to improve the final image quality.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the technical problem of low image clarity for the pulse coupled neural network, and to provide an image clarity evaluation method and terminal device based on the pulse coupled neural network, which utilizes the pulse coupled neural network to calculate the image clarity combined with the control strategy Enables the lens to focus while enabling the Image Enhancement feature to output higher quality clear images.

In order to solve the above-mentioned technical problems, the technical scheme adopted in the present invention is:

An image sharpness evaluation method based on pulse coupled neural network, which is special in that it includes the following steps:

Step 1: Build a pulse coupled neural network;

The impulse coupled neural network includes feedback input domain, coupling connection domain and impulse generating domain;

The feedback input domain includes the input matrix F, the coupling connection domain includes the coupling connection input matrix L, and the impulse generation domain includes the dynamic threshold matrix E, the neuron internal activity term matrix U and the impulse output matrix Y;

Step 2: Preprocess the image and set the initialization parameters;

2.1) Find the grayscale matrix of the image, normalize the grayscale matrix, and use it as the stimulus input matrix S;

The size i*j of the stimulus input matrix S is determined according to the resolution of the image, i is the ith row of the stimulus input matrix S, and j is the jth column of the stimulus input matrix S;

2.2) Assign the value of the stimulus input matrix S to the input matrix F;

2.3) The coupling connection input matrix L, the activity term matrix U inside the neuron, the pulse output matrix Y and the global feature matrix Q in the pulse coupled neural network are initialized to a zero matrix;

2.4) Set the feedback input domain coefficient matrix M and the coupling connection domain coefficient matrix W in the impulse coupled neural network;

2.5) Calculate the dynamic threshold matrix E;

Solve the convolution of the Laplacian operator of the input matrix F, and return a matrix of the same size as the input matrix F; subtract the two to obtain the initial dynamic threshold matrix E;

2.6) Screen the maximum value T in the input matrix F;

2.7) Set the amplification factor and decay time constant;

Set the amplification factor vF of the feedback input domain, the amplification factor vL of the coupling connection domain, the amplification factor vE of the pulse generation domain, the decay time constant αF of the input matrix F, the decay time constant αL of the coupling connection input matrix L, and the dynamic threshold matrix E. Decay time constant αE; the value ranges of vF, vL and vE are natural numbers greater than or equal to 1, and the value ranges of αF, αL and αE are 0<αF<1, 0<αL<1, 0<αE<1 ;

2.8) Set the connection coefficient β of the coupling connection input matrix L to the activity item matrix U inside the neuron, where β is the proportional relationship between the activity item matrix U inside the neuron and the coupling connection input matrix L, and the value range of β is 0< β<1;

2.9) Set the number of cycles to n, N≥n≥1, where N is the upper limit of the number of cycles; set the initial value of the number of cycles to n=1;

Step 3: Based on the pulse coupled neural network, calculate the dynamic threshold matrix E(n), the internal activity term matrix U(n) of the neuron and the pulse output matrix Y(n);

Step 4: Judge the loop result

4.1) When U(n)≤E(n), then Y(n) is a 0 matrix, the loop ends, and step 5 is executed;

When U(n)>E(n), judge whether n is equal to N;

If n≠N, then let n=n+1, go back to step 3;

If n=N, the loop ends, and step 5 is executed.

Step 5: Output the result

5.1) Calculate the global feature matrix Q

5.2) Standardize the global feature matrix Q obtained in step 5.1, and output the standardized global feature matrix Q;

5.3) Calculate the sum of all elements in the standardized global feature matrix Q obtained in step 5.2, that is, the grayscale sum of the image corresponding to the matrix, as the value of the image sharpness evaluation function.

Further, in step 3, based on the pulse coupled neural network, the dynamic threshold matrix E(n), the internal activity term matrix U(n) of the neuron and the pulse output matrix Y(n) are calculated as follows:

3.1) Calculate the convolution of the pulse output matrix Y(n-1) and the feedback input domain coefficient matrix M, and save it as the intermediate matrix J(n); when n=1, Y(n-1) means Y(0) represents the pulse The initial value of the output matrix Y;

3.2) Calculate the convolution of the pulse output matrix Y(n-1) and the coupling domain coefficient matrix W, and save it as an intermediate matrix K(n);

3.3) Calculate the input matrix F(n)=exp(-αF)*F(n-1)+vF*J(n-1)

where:

F(n-1) is the value of F(n) in the previous cycle. When n=1, F(n-1), that is, F(0), represents the initial value of the input matrix F;

J(n-1) is the value of J(n) in the previous cycle. When n=1, J(n-1), that is, J(0), represents the initial value of the intermediate matrix J, and the initial value of the intermediate matrix J is zero matrix;

3.4) Calculate the coupled connection input matrix L(n)=exp(-αL)*L(n-1)+vL*K(n-1)

where:

L(n-1) is the value of L(n) in the previous cycle. When n=1, L(n-1), that is, L(0), represents the initial value of the coupling connection input matrix L;

K(n-1) is the value of K(n) in the previous cycle. When n=1, K(n-1) means K(0) represents the initial value of the intermediate matrix K, and the initial value of the intermediate matrix K is zero matrix;

3.5) Calculate the dynamic threshold matrix E(n)=exp(-αE)*E(n-1)+vE*Y(n-1)

where:

E(n-1) is the value of E(n) in the previous cycle. When n=1, E(n-1), that is, E(0), represents the initial value of the dynamic threshold matrix E;

Y(n-1) is the value of Y(n) in the previous cycle;

3.6) Calculate the component item U _ij (n) of the activity item matrix U(n) inside the neuron, the formula is:

U _ij (n)=F _ij (n)*(1+β*L _ij (n))

where:

U _ij (n) is the item in the i-th row and the j-th column in U(n);

F _ij (n) is the item in the ith row and the jth column of F(n);

L _ij (n) is the item in the i-th row and the j-th column in L(n);

3.7) Calculate the pulse output matrix Y(n)=(lnT-(n-1)*αE)*(U(n)-E(n)).

Further, it also includes step 6: repeating steps 1-5.3 to obtain the image sharpness evaluation function value corresponding to the image sequence, and drawing a curve graph to represent the sharpness transformation trend corresponding to the image sequence.

Further, in step 2.3), the size of the feedback input domain coefficient matrix M and the coupling connection domain coefficient matrix W are both 3*3.

Further, the input of the feedback input field is the pulse output of the previous loop, and its output is the feedback input of the current loop; the input of the coupling connection field is the pulse output of the neurons around the previous loop, and its output is the coupling of the current loop. Connection input; the input of the pulse generation domain is the internal activity item intensity determined by the current feedback input and the coupling connection input, and the output is the current pulse output and the dynamic threshold that determines the output intensity.

The present invention also provides a terminal device including a memory, a processor and a computer program stored in the memory and running on the processor, the special feature of which is that the processor implements the steps of the above method when executing the computer program.

Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

The present invention is based on the image definition evaluation method of the pulse coupled neural network, and applies the pulse coupled neural network in the field of image definition evaluation and automatic focusing; the network parameters of the pulse coupled neural network are set in advance, and it does not need to go through multiple samples for a long time. It can be directly applied to engineering and save a lot of engineering time; compared with traditional image processing methods, it is more diversified and realizes multiple functions: pulse coupled neural network can be used for image enhancement alone , to improve the image quality, it can also be used for image clarity evaluation alone, or the two can be combined; in addition, by modifying the network parameters of the pulse coupled neural network, it can be applied to different scene types, and only focus on the grayscale information of the image, It can improve the calculation accuracy, avoid the interference of other factors, has better real-time performance, high computing efficiency when running on hardware, and the processing speed can meet engineering needs.

Description of drawings

FIG. 1 is a schematic diagram of the model loop iteration structure of the pulse coupled neural network in the image sharpness evaluation method based on the pulse coupled neural network of the present invention.

The reference numbers in the figure are:

1-Feedback input domain, 2-Coupling connection domain, 3-Pulse generating domain.

Detailed ways

The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention and the accompanying drawings. Obviously, the described embodiments do not limit the present invention.

Pulse-coupled neural network (PCNN) is a quantitative description of the signal transduction characteristics of mammalian visual cortex neurons. Edge extraction and other image processing projects; changing the network structure and parameters can also achieve image enhancement, filtering, fusion and other functions.

The pulse coupled neural network includes three parts: feedback input domain 1, coupling connection domain 2 and pulse generation domain 3. The input of feedback input domain 1 is the pulse output of the previous cycle, and its output is the feedback input of the current cycle; coupling connection domain 2 The input is the pulse output of the neurons around the previous cycle, and its output is the coupled connection input of the current cycle; the input of the pulse generation domain 3 is the internal activity item intensity determined by the current feedback input and the coupled connection input, and its output is the current pulse. Output and a dynamic threshold that determines the output strength.

The image processing model of the pulse-coupled neural network consists of a two-dimensional single-layer neuron array composed of pulse-coupled neurons. The number of neurons is the same as the number of pixels. The center of the 3*3 feedback input domain coefficient matrix M and the coupled connection domain coefficient matrix W (the feedback input domain coefficient matrix M and the coupled connection domain coefficient matrix W are the weight connection matrix), and the adjacent pixels are the feedback input domain coefficient matrix M, For the neurons corresponding to the coupling connection domain coefficient matrix W, there are various connection weights between each neuron and adjacent neurons, which are reflected by the value of the weight connection matrix.

The model of the pulse coupled neural network is evolved from the biological visual neuron model, and it is more sensitive to the darker area, that is, the area with low gray value; before image processing, in order to better highlight the edge features of the dark part of the image, often It will artificially expand the grayscale difference between image pixels, enhance the edge of the image to highlight the image feature information; for this, it is necessary to modify the initial value of the dynamic threshold matrix E, and perform edge enhancement on the stimulus input matrix S, and use Laplace calculation. Then the image is inverted; at this time, the position of the original dark area of the image becomes brighter, and the gray value becomes larger, and its value is assigned to the dynamic threshold matrix E, so that the threshold value of the dark area increases, and the decay time is longer. More processing can be done during loop iterations.

After the stimulus input matrix S is generated, the neuron is activated; each neuron receives the signal related to the input matrix F and the coupled connection input matrix L, and generates the internal activity matrix U of the neuron. Only when the internal activity item matrix U of the neuron is greater than the dynamic threshold matrix E, the neuron will be activated, and the pulse output matrix Y will be generated. The intensity of the pulse output matrix Y is equal to the difference between the neuron internal activity item matrix U and the dynamic threshold matrix E. The integral formed by the continuous accumulation of the pulse output matrix Y is the global characteristic matrix Q. Each neuron is related to the surrounding neurons in the previous cycle. The pulse output matrix Y of the previous cycle will act on the adjacent neurons by feeding back the input domain coefficient matrix M and the coupled connection domain coefficient matrix W to generate a new input matrix F and The coupling is connected to the input matrix L, and the loop iteration process is similar to the impulse function. After several pulse cycles, the internal activity matrix U of the neuron is no longer larger than the dynamic threshold matrix E, and the pulse output matrix Y is no longer excited, that is, the cycle ends. The global feature matrix Q at this time is the enhanced image, or The analogy is the image seen by the human eye after being processed by a spiking neural network.

During the lens focusing process of the same target image (or video), the image at the focus position has richer dark details, and the gray value is larger when the image is inverted, which requires more cycles to activate the neurons around the image to generate pulses , so the accumulated global feature matrix Q will get a larger gray value. By calculating the gray sum of the global feature matrix Q of the enhanced image, this feature information can be extracted as a representation of image clarity, and output according to the final result The enhanced image and the sharpness evaluation function value are used for the next stage of operation, and the automatic focusing is realized in combination with the control strategy. The specific method is as follows:

A method for evaluating image clarity based on pulse coupled neural network, comprising the following steps:

Step 1: Build a pulse coupled neural network;

The impulse coupled neural network includes feedback input domain 1, coupling connection domain 2 and impulse generating domain 3;

Feedback input domain 1 includes input matrix F, coupling connection domain 2 includes coupling connection input matrix L, and impulse generating domain 3 includes dynamic threshold matrix E, neuron internal activity term matrix U and impulse output matrix Y;

Step 2: Preprocess the image and set the initialization parameters;

2.1) Find the grayscale matrix of the image, normalize the grayscale matrix, and use it as the stimulus input matrix S;

The size i*j of the stimulus input matrix S is determined according to the resolution of the image, i is the ith row of the stimulus input matrix S, and j is the jth column of the stimulus input matrix S;

2.2) Assign the value of the stimulus input matrix S to the input matrix F;

2.3) Initialize the coupling connection input matrix L, the activity term matrix U inside the neuron, the pulse output matrix Y and the global feature matrix Q in the impulse coupled neural network to zero matrix; among them, the feedback input domain coefficient matrix M and the coupling connection domain The size of the coefficient matrix W is 3*3;

2.4) Set the feedback input domain coefficient matrix M and the coupling connection domain coefficient matrix W in the impulse coupled neural network;

2.5) Calculate the dynamic threshold matrix E;

Solve the convolution of the Laplacian operator of the input matrix F, and return a matrix of the same size as the input matrix F; subtract the two to obtain the initial dynamic threshold matrix E;

2.6) Screen the maximum value T in the input matrix F;

2.7) Set the amplification factor and decay time constant;

Set the amplification factor vF of the feedback input domain 1, the amplification factor vL of the coupling connection domain 2, the amplification factor vE of the pulse generation domain 3, the decay time constant αF of the input matrix F, the decay time constant αL of the coupling connection input matrix L, and the dynamic threshold Decay time constant αE of matrix E; among them, the value ranges of vF, vL and vE are natural numbers greater than or equal to 1, and the value ranges of αF, αL and αE are 0<αF<1, 0<αL<1, 0< αE<1;

2.8) Set the connection coefficient β of the coupling connection input matrix L to the activity item matrix U inside the neuron, where β is the proportional relationship between the activity item matrix U inside the neuron and the coupling connection input matrix L, and the value range of β is 0< β<1;

2.9) Set the number of cycles to n, N≥n≥1, where N is the upper limit of the number of cycles; set the initial value of the number of cycles to n=1;

Step 3: Based on the pulse coupled neural network, calculate the dynamic threshold matrix E(n), the internal activity term matrix U(n) of the neuron and the pulse output matrix Y(n);

3.1) Calculate the convolution of the pulse output matrix Y(n-1) and the feedback input domain coefficient matrix M, and save it as the intermediate matrix J(n); when n=1, Y(n-1) means Y(0) represents the pulse The initial value of the output matrix Y;

3.2) Calculate the convolution of the pulse output matrix Y(n-1) and the coupling domain coefficient matrix W, and save it as an intermediate matrix K(n);

3.3) Calculate the input matrix F(n)=exp(-αF)*F(n-1)+vF*J(n-1)

where:

F(n-1) is the value of F(n) in the previous cycle. When n=1, F(n-1), that is, F(0), represents the initial value of the input matrix F;

J(n-1) is the value of J(n) in the previous cycle. When n=1, J(n-1), that is, J(0), represents the initial value of the intermediate matrix J, and the initial value of the intermediate matrix J is zero matrix;

3.4) Calculate the coupled connection input matrix L(n)=exp(-αL)*L(n-1)+vL*K(n-1)

where:

L(n-1) is the value of L(n) in the previous cycle. When n=1, L(n-1), that is, L(0), represents the initial value of the coupling connection input matrix L;

K(n-1) is the value of K(n) in the previous cycle. When n=1, K(n-1) means K(0) represents the initial value of the intermediate matrix K, and the initial value of the intermediate matrix K is zero matrix;

3.5) Calculate the dynamic threshold matrix E(n)=exp(-αE)*E(n-1)+vE*Y(n-1)

where:

E(n-1) is the value of E(n) in the previous cycle. When n=1, E(n-1), that is, E(0), represents the initial value of the dynamic threshold matrix E;

Y(n-1) is the value of Y(n) in the previous cycle;

3.6) Calculate the component item U _ij (n) of the activity item matrix U(n) inside the neuron, the formula is:

U _ij (n)=F _ij (n)*(1+β*L _ij (n))

where:

U _ij (n) is the item in the i-th row and the j-th column in U(n);

F _ij (n) is the item in the ith row and the jth column of F(n);

L _ij (n) is the item in the i-th row and the j-th column in L(n);

3.7) Calculate the pulse output matrix Y(n)=(lnT-(n-1)*αE)*(U(n)-E(n)).

Step 4: Judge the loop result

4.1) When U(n)≤E(n), then Y(n) is a 0 matrix, the loop ends, and step 5 is executed;

When U(n)>E(n), judge whether n is equal to N;

If n≠N, then let n=n+1, go back to step 3;

If n=N, the loop ends, and step 5 is executed.

Step 5: Output the result

5.1) Calculate the global feature matrix Q

5.2) Standardize the global feature matrix Q obtained in step 5.1, and output the standardized global feature matrix Q;

5.3) Calculate the sum of all elements in the standardized global feature matrix Q obtained in step 5.2, that is, the grayscale sum of the image corresponding to the matrix, as the value of the image sharpness evaluation function.

Step 6: Repeat steps 1-5.3 to obtain the image sharpness evaluation function value corresponding to the image sequence, and draw a curve graph to represent the sharpness transformation trend corresponding to the image sequence.

In addition, the image sharpness evaluation method based on the pulse coupled neural network of the present invention can also be applied to terminal equipment. The terminal equipment includes a memory, a processor, and a computer program stored in the memory and running on the processor. The processor executes the computer In the program, the steps of the image sharpness evaluation method of the present invention are realized. The terminal device here can be computing devices such as a computer, a notebook, a palmtop computer, and various cloud servers, and the processor can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, or other programmable logic devices.

In this embodiment, the model loop iteration of the pulse coupled neural network is as follows:

The feedback input domain 1 has the feedback input domain coefficient matrix M, the intermediate matrix J, the input matrix F, the amplification coefficient vF of the feedback input domain 1 and the decay time constant αF of the input matrix F;

In the feedback input domain 1, the input matrix F consists of two parts: one part comes from the pulse output matrix Y of the previous cycle (ie Y(n-1)), the pulse output matrix Y of the previous cycle and the feedback input domain coefficient matrix M volume The product obtains the intermediate matrix J, which is multiplied by the amplification factor vF of the feedback input domain 1; the other part of the input matrix F (ie F(n-1)) from the previous cycle is multiplied by its decay time constant αF; The partial operation results are added to obtain the input matrix F (ie, F(n)) of the current loop.

The coupling connection field 2 has a coupling connection field coefficient matrix W, an intermediate matrix K, a coupling connection input matrix L, an amplification coefficient vL of the coupling connection field 2 and a decay time constant αL of the coupling connection input matrix L;

In the coupling-connection domain 2, the coupling-connection input matrix L consists of two parts: a part of the pulse output matrix Y from the surrounding neurons in the previous cycle is convolved with the coupling-connection domain coefficient matrix W to obtain an intermediate matrix K, which is connected with the coupling-connection domain. The amplification factor vL of 2 is multiplied; the other part of the coupling connection input matrix L (ie L(n-1)) from the previous cycle is multiplied by its decay time constant αL; the two parts of the operation results are added to obtain the coupling connection of the current cycle Input matrix L (ie L(n)).

The pulse generation domain 3 has the dynamic threshold matrix E, the amplification factor vE of the pulse generation domain 3, the decay time constant αE of the dynamic threshold matrix E, the pulse output matrix Y, the maximum value T of the image gray level and the neuron internal activity term matrix U;

In the pulse generation domain 3, the dynamic threshold matrix E consists of two parts: one part comes from the pulse output matrix Y of the previous cycle and the amplification factor vE of the pulse generation domain 3 is multiplied; the other part comes from the dynamic threshold matrix E of the previous cycle (ie E(n-1)) is multiplied by its decay time constant αE; the results of the two operations are added to obtain the dynamic threshold matrix E (ie, E(n)) of the current cycle.

At the same time, the activity matrix U of the current neuron is also composed of two parts: one part comes from the influence of the central neuron, that is, the input matrix F of the current cycle; the other part comes from the influence of the pulse output matrix Y of the surrounding neurons in the previous cycle. The current cyclic coupling connection input matrix L is multiplied by the connection coefficient β, plus the basic weight 1; the result of the multiplication of these two parts is the neuron's internal activity term matrix U.

When the internal activity matrix U of the neuron is greater than the dynamic threshold matrix E, pulse excitation is generated, and the pulse generator generates a pulse output matrix Y whose intensity is positively correlated with the maximum value T, and the next judgment or operation is performed; the pulse output generated in each cycle is output. The matrix Y is accumulated, and the obtained global feature matrix Q will be the final output result.

The above description is only an embodiment of the present invention, and does not limit the protection scope of the present invention. Any equivalent structural transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are included in the within the scope of patent protection of the present invention.

Claims (6)

1. an image clarity assessment method based on pulse coupled neural network, is characterized in that, comprises the following steps: Step 1: Build a pulse coupled neural network; The pulse coupling neural network includes a feedback input domain (1), a coupling connection domain (2) and an impulse generating domain (3); The feedback input domain (1) includes an input matrix F, the coupling connection domain (2) includes a coupling connection input matrix L, and the impulse generation domain (3) includes a dynamic threshold matrix E, a neuron internal activity term matrix U and pulse output matrix Y; Step 2: Preprocess the image and set the initialization parameters; 2.1) Find the grayscale matrix of the image, normalize the grayscale matrix, and use it as the stimulus input matrix S; The size i*j of the stimulus input matrix S is determined according to the resolution of the image, i is the ith row of the stimulus input matrix S, and j is the jth column of the stimulus input matrix S; 2.2) Assign the value of the stimulus input matrix S to the input matrix F; 2.3) The coupling connection input matrix L, the activity term matrix U inside the neuron, the pulse output matrix Y and the global feature matrix Q in the pulse coupled neural network are initialized to a zero matrix; 2.4) Set the feedback input domain coefficient matrix M and the coupling connection domain coefficient matrix W in the impulse coupled neural network; 2.5) Calculate the dynamic threshold matrix E; Solve the convolution of the Laplacian operator of the input matrix F, and return a matrix of the same size as the input matrix F; subtract the two to obtain the initial dynamic threshold matrix E; 2.6) Screen the maximum value T in the input matrix F; 2.7) Set the amplification factor and decay time constant; Set the amplification factor vF of the feedback input domain (1), the amplification factor vL of the coupling connection domain (2), the amplification factor vE of the pulse generation domain (3), the decay time constant αF of the input matrix F, and the attenuation of the coupling connection input matrix L The time constant αL and the decay time constant αE of the dynamic threshold matrix E; wherein, the value ranges of vF, vL and vE are natural numbers greater than or equal to 1, and the value ranges of αF, αL and αE are 0<αF<1, 0< αL<1, 0<αE<1; 2.8) Set the connection coefficient β of the coupling connection input matrix L to the neuron's internal activity item matrix U, where β is the proportional relationship between the neuron's internal activity item matrix U and the coupling connection input matrix L, and the value range of β is 0 < β<1; 2.9) Set the number of cycles to n, N≥n≥1, where N is the upper limit of the number of cycles; set the initial value of the number of cycles to n=1; Step 3: Based on the pulse coupled neural network, calculate the dynamic threshold matrix E(n), the internal activity term matrix U(n) of the neuron and the pulse output matrix Y(n); Step 4: Judge the loop result 4.1) When U(n)≤E(n), then Y(n) is a 0 matrix, the loop ends, and step 5 is executed; When U(n)>E(n), judge whether n is equal to N; If n≠N, then let n=n+1, go back to step 3; If n=N, the cycle ends, and step 5 is executed; Step 5: Output the result 5.1) Calculate the global feature matrix Q

5.2) Standardize the global feature matrix Q obtained in step 5.1, and output the standardized global feature matrix Q; 5.3) Calculate the sum of all elements in the standardized global feature matrix Q obtained in step 5.2, that is, the grayscale sum of the image corresponding to the matrix, as the value of the image sharpness evaluation function. 2. The image definition evaluation method based on pulse coupled neural network according to claim 1, is characterized in that: in step 3, described based on pulse coupled neural network, calculates dynamic threshold matrix E(n), neuron internal The activity item matrix U(n) and the pulse output matrix Y(n) are specifically: 3.1) Calculate the convolution of the pulse output matrix Y(n-1) and the feedback input domain coefficient matrix M, and save it as an intermediate matrix J(n); when n=1, Y(n-1) means Y(0) represents the pulse The initial value of the output matrix Y; 3.2) Calculate the convolution of the pulse output matrix Y(n-1) and the coupling domain coefficient matrix W, and save it as an intermediate matrix K(n); 3.3) Calculate the input matrix F(n)=exp(-αF)*F(n-1)+vF*J(n-1) where: F(n-1) is the value of F(n) in the previous cycle. When n=1, F(n-1), that is, F(0), represents the initial value of the input matrix F; J(n-1) is the value of J(n) in the previous cycle. When n=1, J(n-1), that is, J(0), represents the initial value of the intermediate matrix J, and the initial value of the intermediate matrix J is zero matrix; 3.4) Calculate the coupled connection input matrix L(n)=exp(-αL)*L(n-1)+vL*K(n-1) where: L(n-1) is the value of L(n) in the previous cycle. When n=1, L(n-1), that is, L(0), represents the initial value of the coupling connection input matrix L; K(n-1) is the value of K(n) in the previous cycle. When n=1, K(n-1) means K(0) represents the initial value of the intermediate matrix K, and the initial value of the intermediate matrix K is zero matrix; 3.5) Calculate the dynamic threshold matrix E(n)=exp(-αE)*E(n-1)+vE*Y(n-1) where: E(n-1) is the value of E(n) in the previous cycle. When n=1, E(n-1), that is, E(0), represents the initial value of the dynamic threshold matrix E; Y(n-1) is the value of Y(n) in the previous cycle; 3.6) Calculate the component item U _ij (n) of the activity item matrix U(n) inside the neuron, the formula is: U _ij (n)=F _ij (n)*(1+β*L _ij (n)) where: U _ij (n) is the item in the i-th row and the j-th column in U(n); F _ij (n) is the item in the ith row and the jth column of F(n); L _ij (n) is the item in the i-th row and the j-th column in L(n); 3.7) Calculate the pulse output matrix Y(n)=(lnT-(n-1)*αE)*(U(n)-E(n)). 3. The image sharpness evaluation method based on pulse coupling neural network according to claim 2, is characterized in that, also comprises step 6: repeating step 1-step 5.3, obtains the image sharpness evaluation function value corresponding to the image sequence, and A curve graph is drawn to represent the corresponding sharpness transformation trend of the image sequence. 4. The image sharpness evaluation method based on impulse coupling neural network according to claim 1 or 2, is characterized in that: in step 2.3), described feedback input domain (1) coefficient matrix M and coupling connection domain (2) The size of the coefficient matrix W is all 3*3. 5. The image sharpness assessment method based on pulse coupled neural network according to claim 1 or 2, characterized in that: the input of the feedback input field (1) is the pulse output of the previous cycle, and its output is the current cycle The input of the coupled connection domain (2) is the pulse output of the neurons around the previous cycle, and its output is the coupled connection input of the current cycle; the input of the pulse generation domain (3) is determined by the current feedback input and the coupled connection input. The intensity of the internal activity term, its output is the current pulse output and the dynamic threshold that determines the output intensity. 6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: when the processor executes the computer program, the implementation as claimed in the claims Steps of any one of 1 to 5 of the method.

Images