Multi-image encryption method combining phase contrast filtering and convolutional neural network auxiliary decryption

By combining phase contrast filtering and convolutional neural network multi-image encryption methods, the problem of difficulty in ensuring the security of multi-image information is solved, and a high security and robust encryption and decryption process is realized, simplifying the decryption process and enhancing algorithm security is enhanced. sex.

CN118984357APending Publication Date: 2024-11-19FUJIAN NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411041964.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The prior art is difficult to effectively ensure the security of multi-image information, especially in the process of information sharing and communication, image encryption technology has the risk of being tampered with, forged and leaked, and traditional encryption methods are difficult to achieve high security when facing multi-images. Sexual and efficient decryption.

Method used

Using a multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption, phase templates are generated through phase truncation nonlinear encryption, artificial neural grid model training, Fresnel inverse diffraction and phase recovery algorithms, and phase contrast is used to generate an encrypted and decrypted phase template. Filtering technology and convolutional neural networks realize secure encryption and decryption of multiple images.

Benefits of technology

Improve the security and robustness of multi-image information. Any decryption phase template missing or mismatched parameters will lead to decryption failure, simplifying the decryption process, and recovering the original image through ANN-assisted decryption without a traditional private key, enhancing the algorithm Security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984357A_ABST
    Figure CN118984357A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-image encryption method combining phase contrast filtering and convolutional neural network auxiliary decryption. The method comprises the following steps: encrypting a plaintext image to obtain a ciphertext image; training an artificial neural network by the plaintext and ciphertext image pair to obtain a trained neural network model; taking the ciphertext image as a phase to construct a virtual index signal; multiplying the imaginary index signals of the ciphertext images to obtain a product signal, multiplying the product signal by a random phase template to obtain a product signal in a noise form, and calculating Fresnel inverse diffraction to obtain a complex signal; the amplitude of the complex signal and the unit amplitude template are used as constraints to calculate and obtain a public encryption phase template; a first decryption phase template is calculated based on the encryption phase template; and generating a second decryption phase template by using each ciphertext image and the first decryption phase template. Obtaining a ciphertext image by using the public encryption phase template, the two decryption phase templates of each ciphertext image and a phase contrast filtering technology; and sending the ciphertext image into a trained artificial neural network to obtain a plaintext image. According to the invention, the security of image encryption and decryption is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of multi-image encryption and decryption, and in particular to a multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption. Background Art

[0002] With the rapid development of the Internet and multimedia technology, all kinds of digital information can be easily and quickly transmitted online, but this has also caused a series of security issues. As the most widely used information carrier, digital images are easy to copy and spread. How to ensure the security of image information is of great significance to promoting information sharing and communication and promoting the development of an information society. Image encryption can effectively protect the security and integrity of image information and prevent images from being tampered with, forged, or leaked by confusing and diffusing the original information through technical means. Virtual optical encryption is a technology that uses optical principles to achieve image encryption. Its basic principle is to use the interference and diffraction characteristics of light to convert image information into optical signals, and modulate and convert them in space to achieve encryption and decryption. Virtual optical encryption has the advantages of high security, high resolution, and high speed. Compared with traditional encryption technology, virtual optical encryption technology uses the special properties of light to make the encryption process more difficult to crack and attack. In an information society, with the continuous improvement of information security requirements, virtual optical encryption technology will have a wider application prospect. Multi-image encryption achieves the purpose of encryption and decryption by confusing and diffusing the information of multiple images. Multi-image encryption is widely used in digital watermarking, copyright protection, information hiding and other fields to protect the information of multiple images from being illegally obtained or tampered with. Summary of the invention

[0003] The purpose of the present invention is to provide a multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption, so as to improve the information security of multiple images.

[0004] The technical solution adopted by the present invention is:

[0005] The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption specifically includes the following steps:

[0006] Encryption steps:

[0007] Step S11, firstly, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image;

[0008] Step S12, building an artificial neural network model (ANN), using the plaintext and ciphertext image pairs to train the artificial neural network (ANN) to obtain a neural network model that has been trained for decryption;

[0009] Step S13, using the ciphertext image as the phase to construct an imaginary exponential signal;

[0010] Step S14, multiplying the imaginary exponential signals from the ciphertext images to obtain a product signal containing multiple image information, and then multiplying the product signal by a random phase template to obtain a product signal containing multiple image information in the form of noise;

[0011] Step S15, calculating the Fresnel inverse diffraction of the product signal to obtain a complex signal;

[0012] Step S16, using the amplitude of the complex signal and the unit amplitude template as constraints, using a phase recovery algorithm between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template.

[0013] Step S17, using the amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located as constraints, and using a phase recovery algorithm between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located to generate a first decrypted phase template;

[0014] Step S18, generating a second decryption phase template using each ciphertext image and the obtained first decryption phase template.

[0015] Decryption steps:

[0016] Step S21, using a common encryption phase template and two decryption phase templates in a cascade structure corresponding to each ciphertext image and a phase contrast filtering technique to obtain a ciphertext image;

[0017] Step S22, sending the ciphertext image to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext.

[0018] Furthermore, the asymmetric encryption process in step S11 is expressed by the formula:

[0019]

[0020] Among them, X represents the plaintext image, F represents the Fourier transform, and F -1 represents inverse Fourier transform, abs() represents amplitude; e jR1 and e jR2 Represents two random phase templates, and R1 and R2 represent two random phases.

[0021] Furthermore, the product signal in the form of noise containing multiple images in step S14 is expressed as:

[0022]

[0023] Among them, C1, C2, ...C N Respectively represent the plaintext images X1, X2, ...X NThe ciphertext image generated after encryption by the phase-truncated asymmetric linear system;

[0024] Furthermore, in step S15, the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images is calculated, and the specific expression is as follows:

[0025]

[0026] Where A represents the amplitude, represents phase; IFrT represents inverse Fresnel diffraction transform.

[0027] Furthermore, the specific steps of generating the encrypted phase template in step S16 are as follows:

[0028] Step S16-1, the phase signal of the observation plane P1 where the encrypted phase template is located is initially set to a random phase signal;

[0029] Step S16-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d1, and obtaining a complex signal of the diffracted light wave at the P2 observation surface;

[0030] Step S16-3, the amplitude of the diffracted light wave signal of the P2 observation surface is replaced by the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images, and the phase is retained to obtain the diffracted light wave signal after amplitude constraint;

[0031] Step S16-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance d1 to obtain a diffracted light wave signal on the P1 plane;

[0032] Step S16-5, the amplitude of the diffraction light wave signal of the P1 plane is constrained to 1, the phase is retained, and the object plane complex signal after being constrained to unit amplitude is obtained;

[0033] Step S16-6, the object plane complex signal is used as the next initial object wave function for re-iteration until the algorithm converges or reaches the number of iterations; the obtained P1 plane diffraction light wave signal is the encrypted phase template E; the encrypted phase template e generated after the iteration is completed jE Satisfies the following relationship:

[0034]

[0035] Wherein, the symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d1 represents the diffraction distance.

[0036] Furthermore, step S17 specifically includes the following steps:

[0037] Step S17-1, the phase signal of the observation plane P2 where the decrypted phase template is located is initially set to a random phase signal, and the amplitude constraint is the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images;

[0038] Step S17-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d2, and obtaining a complex signal of the diffracted light wave at the P3 observation surface;

[0039] Step S17-3, constraining the amplitude of the diffracted light wave signal of the P3 observation surface to be unit amplitude;

[0040] Step S17-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance d2 to obtain a diffracted light wave signal on the P2 plane;

[0041] Step S17-5, retain the phase of the diffracted light wave signal of the P2 plane, and constrain the amplitude to the amplitude A obtained by the inverse Fresnel diffraction transform of the product signal Z in the form of noise containing multiple images, that is, the amplitude A generated in formula (3); the signal after amplitude constraint is used as the next initial object wave function for re-iteration until the algorithm converges or the number of iterations is reached; the obtained light wave signal of the P2 plane is first decrypted using the phase template D1.

[0042] The iteration ends, the first decryption phase e jD1 Satisfies the relationship:

[0043]

[0044] The symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d2 represents the diffraction distance.

[0045] Furthermore, in step S18, the second decrypted phase template e jD2 The specific calculation formula is as follows:

[0046]

[0047] Among them, C i Represents the plaintext image X i The corresponding ciphertext, ang() represents the phase operation.

[0048] Furthermore, in step S21, a virtual optical path is first constructed, and an encrypted phase template E, a first decrypted phase template D1 corresponding to the decrypted i-th image, a second decrypted phase template D2 corresponding to the decrypted i-th image, and a phase contrast filter circuit are sequentially placed on a specific plane of the virtual optical path, and a lens is provided between the decrypted phase template D2 and the phase contrast filter circuit, and between the phase contrast filter circuit and the output surface; then, a ciphertext image C corresponding to the i-th plaintext image is reconstructed on the output surface. iBy changing the decryption phase template corresponding to each image respectively, the ciphertext image corresponding to each plaintext image can be reconstructed.

[0049] Furthermore, the specific steps of step S21 are:

[0050] Step S21-1, calculate and obtain the plane signal where the decrypted phase template D2 is located, and the specific expression is as follows:

[0051]

[0052] Wherein, E represents the encrypted phase template, D1 and D2 represent the first and second decrypted phase templates respectively,

[0053] Step S21-2: the second decryption phase template D2 is located on the plane signal e jCi After the phase contrast filter circuit, the output surface will obtain the amplitude signal corresponding to the plaintext image X i The ciphertext image C i , the specific calculation formula is as follows:

[0054]

[0055] Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, W1 represents the zero-order image window, e jCi Indicates phase C i The imaginary exponential function of .

[0056] Specifically, the information processing process of phase contrast filtering is as follows: the diffracted light wave signal e on the D2 plane is jCi Calculate the Fourier transform, and the phase contrast filter located on the spectrum plane realizes the attenuation of the zero-order image and phase shifts it by 90 degrees. Then, after the inverse Fourier transform, the ciphertext C is obtained on the output plane. i , which can be expressed by formula (8).

[0057]

[0058] Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, and W1 represents the zero-order image window. Indicates phase C i The imaginary exponential function of .

[0059] Furthermore, the quality of the plaintext image reconstructed in step S22 is evaluated by the correlation coefficient, and the calculation formula of the normalized correlation coefficient NC is as follows:

[0060]

[0061] Among them, m and n are the horizontal and vertical coordinates of the pixel points of the image respectively; M and N are the image width and height respectively; f1 and f2 are two signals respectively.

[0062] The present invention adopts the above technical scheme. First, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image; an artificial neural network model (ANN) is built, and the plaintext and ciphertext images are used to train the artificial neural network (ANN) to obtain a trained neural network model that can be used for decryption; the ciphertext image is used as the phase to construct an imaginary exponential signal; the imaginary exponential signals from each ciphertext image are multiplied to obtain a product signal containing multiple image information, and the product signal is multiplied by a random phase template to obtain a product signal containing multiple image information in the form of noise; the Fresnel inverse diffraction of the product signal is calculated to obtain a complex signal; the amplitude and unit amplitude template of the complex signal are used as constraints, and a phase recovery algorithm is used between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template. The amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located are used as constraints, and a first decrypted phase template is generated between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located using a phase recovery algorithm; and then the second decrypted phase template is generated using each ciphertext image and the obtained first decrypted phase template. The ciphertext image is obtained by using a public encryption phase template and two decryption phase templates in the form of a cascade structure corresponding to each ciphertext image and phase contrast filtering technology; the ciphertext image is sent to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext. The test results show that the proposed method has good security and robustness. Any missing decryption phase template or mismatched reconstruction parameters will lead to the failure of reconstruction of the original plaintext image. ANN-assisted decryption allows the original image to be restored without the private key in the traditional phase truncation nonlinear encryption system, simplifying the decryption process. Encoding the ciphertext into phase and combining it with phase contrast filtering decryption further improves the security of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments;

[0064] Figure 1 It is a flow chart of the encryption algorithm of the present invention;

[0065] Figure 2 It is a flow chart of the decryption algorithm of the present invention;

[0066] Figure 3 It is the principle diagram of the phase truncation nonlinear encryption system of the present invention;

[0067] Figure 4 A virtual light path diagram for generating an encrypted phase template according to the present invention;

[0068] Figure 5 It is a principle block diagram of the present invention for generating an encrypted phase template;

[0069] Figure 6 Generate virtual optical paths of first and second decrypted phase templates D1 and D2 for the present invention;

[0070] Figure 7 It is a principle block diagram of generating the first decryption phase template D1 of the present invention;

[0071] Figure 8 A virtual optical path for decrypting each ciphertext image of the present invention;

[0072] Fig. 9 It is a schematic diagram of the ANN decryption process of the present invention;

[0073] Fig.10 The ANN network structure of the present invention;

[0074] FIG. 11( a ) shows six original images of the present invention;

[0075] FIG11( b ) is a common encrypted phase template E generated based on six original images;

[0076] Fig.12 Decrypting the phase template and decryption results for the present invention;

[0077] Fig.13 The decryption results of the decryption phase template of the present invention when it is damaged to different degrees;

[0078] Fig.14 is the robustness test result of the encrypted phase template E of the present invention;

[0079] Fig.15 is the reconstruction result of the reconstruction parameter d1 of the present invention at different deviation rates;

[0080] Fig.16 d2 are reconstruction results of the present invention under different deviation rates. Implementation

[0081] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0082] like Figures 1 to 16 As shown in FIG. 1 , the present invention discloses a multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption, including an encryption part and a decryption part, specifically including the following steps:

[0083] Encryption steps:

[0084] Step S11, firstly, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image;

[0085] Step S12, building an artificial neural network model (ANN), using the plaintext and ciphertext image pairs to train the artificial neural network (ANN) to obtain a neural network model that has been trained for decryption;

[0086] Step S13, using the ciphertext image as the phase to construct an imaginary exponential signal;

[0087] Step S14, multiplying the imaginary exponential signals from the ciphertext images to obtain a product signal containing multiple image information, and then multiplying the product signal by a random phase template to obtain a product signal containing multiple image information in the form of noise;

[0088] Step S15, calculating the Fresnel inverse diffraction of the product signal to obtain a complex signal;

[0089] Step S16, using the amplitude of the complex signal and the unit amplitude template as constraints, using a phase recovery algorithm between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template.

[0090] Step S17, using the amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located as constraints, and using a phase recovery algorithm between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located to generate a first decrypted phase template;

[0091] Step S18, generating a second decryption phase template using each ciphertext image and the obtained first decryption phase template.

[0092] Decryption steps:

[0093] Step S21, using a common encryption phase template and two decryption phase templates in a cascade structure corresponding to each ciphertext image and a phase contrast filtering technique to obtain a ciphertext image;

[0094] Step S22, sending the ciphertext image to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext.

[0095] Furthermore, the asymmetric encryption process in step S11 is expressed by the formula:

[0096]

[0097] Among them, X represents the plaintext image, F represents the Fourier transform, and F -1 represents inverse Fourier transform, abs() represents amplitude; e jR1 and e jR2 Represents two random phase templates, and R1 and R2 represent two random phases.

[0098] Furthermore, the product signal in the form of noise containing multiple images in step S14 is expressed as:

[0099]

[0100] Among them, C1, C2, ...C N Respectively represent the plaintext images X1, X2, ...X N The ciphertext image generated after encryption by the phase-truncated asymmetric linear system;

[0101] Furthermore, in step S15, the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images is calculated, and the specific expression is as follows:

[0102]

[0103] Where A represents the amplitude, represents phase; IFrT represents inverse Fresnel diffraction transform.

[0104] Furthermore, the specific steps of generating the encrypted phase template in step S16 are as follows:

[0105] Step S16-1, the phase signal of the observation plane P1 where the encrypted phase template is located is initially set to a random phase signal;

[0106] Step S16-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d1, and obtaining a complex signal of the diffracted light wave at the P2 observation surface;

[0107] Step S16-3, the amplitude of the diffracted light wave signal of the P2 observation surface is replaced by the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images, and the phase is retained to obtain the diffracted light wave signal after amplitude constraint;

[0108] Step S16-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance D1 to obtain a diffracted light wave signal on the P1 plane;

[0109] Step S16-5, the amplitude of the diffraction light wave signal of the P1 plane is constrained to 1, the phase is retained, and the object plane complex signal after being constrained to unit amplitude is obtained;

[0110] Step S16-6, the object plane complex signal is used as the next initial object wave function for re-iteration until the algorithm converges or reaches the number of iterations; the obtained P1 plane diffraction light wave signal is the encrypted phase template E; the encrypted phase template e generated after the iteration is completed jE Satisfies the following relationship:

[0111]

[0112] Wherein, the symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d1 represents the diffraction distance.

[0113] Furthermore, step S17 specifically includes the following steps:

[0114] Step S17-1, the phase signal of the observation plane P2 where the decrypted phase template is located is initially set to a random phase signal, and the amplitude constraint is the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images;

[0115] Step S17-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d2, and obtaining a complex signal of the diffracted light wave at the P3 observation surface;

[0116] Step S17-3, constraining the amplitude of the diffracted light wave signal of the P3 observation surface to be unit amplitude;

[0117] Step S17-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance d2 to obtain a diffracted light wave signal on the P2 plane;

[0118] Step S17-5, retain the phase of the diffracted light wave signal of the P2 plane, and constrain the amplitude to the amplitude A obtained by the inverse Fresnel diffraction transform of the product signal Z in the form of noise containing multiple images, that is, the amplitude A generated in formula (3); the signal after amplitude constraint is used as the next initial object wave function for re-iteration until the algorithm converges or the number of iterations is reached; the obtained light wave signal of the P2 plane is first decrypted using the phase template D1.

[0119] The iteration ends, the first decryption phase e jD1 Satisfies the relationship:

[0120]

[0121] The symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d2 represents the diffraction distance.

[0122] Furthermore, in step S18, the second decrypted phase template The specific calculation formula is as follows:

[0123]

[0124] Among them, C i Representation and plaintext image X i The corresponding ciphertext, ang() represents the phase operation.

[0125] Furthermore, in step S21, a virtual optical path is first constructed, and an encrypted phase template E, a first decrypted phase template d1 corresponding to the decrypted i-th image, a second decrypted phase template D2 corresponding to the decrypted i-th image, and a phase contrast filter circuit are sequentially placed on a specific plane of the virtual optical path, and a lens is provided between the decrypted phase template D2 and the phase contrast filter circuit, and between the phase contrast filter circuit and the output surface; then, a ciphertext image C corresponding to the i-th plaintext image is reconstructed on the output surface. i By changing the decryption phase template corresponding to each image respectively, the ciphertext image corresponding to each plaintext image can be reconstructed.

[0126] Furthermore, the specific steps of step S21 are:

[0127] Step S21-1, calculate and obtain the plane signal where the decrypted phase template D2 is located, and the specific expression is as follows:

[0128]

[0129] Wherein, E represents the encrypted phase template, D1 and D2 represent the first and second decrypted phase templates respectively,

[0130] Step S21-2: the second decryption phase template D2 is located on the plane signal After the phase contrast filter circuit, the output surface will obtain the amplitude signal corresponding to the plaintext image X i The ciphertext image C i , the specific calculation formula is as follows:

[0131]

[0132] Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, and W1 represents the zero-order image window. Indicates phase C i The imaginary exponential function of .

[0133] Specifically, the information processing process of phase contrast filtering is as follows: the diffracted light wave signal on the D2 plane Calculate the Fourier transform, and the phase contrast filter located on the spectrum plane realizes the attenuation of the zero-order image and phase shifts it by 90 degrees. Then, after the inverse Fourier transform, the ciphertext C is obtained on the output plane. i , which can be expressed by formula (8).

[0134]

[0135] Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, and W1 represents the zero-order image window. Indicates phase C i The imaginary exponential function of .

[0136] Furthermore, the quality of the plaintext image reconstructed in step S22 is evaluated by the correlation coefficient, and the calculation formula of the normalized correlation coefficient NC is as follows:

[0137]

[0138] Among them, m and n are the horizontal and vertical coordinates of the pixel points of the image respectively; M and N are the image width and height respectively; f1 and f2 are two signals respectively.

[0139] The specific principle of the present invention is described in detail below:

[0140] See also Figure 1 , a multi-image encryption and decryption method using phase contrast filtering and artificial neural network assisted decryption, comprising:

[0141] Step S11, firstly, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image;

[0142] Step S12, building an artificial neural network model (ANN), using the plaintext and ciphertext image pairs to train the artificial neural network (ANN) to obtain a trained neural network model that can be used for decryption;

[0143] Step S13, using the ciphertext image as the phase to construct an imaginary exponential signal;

[0144] Step S14, multiplying the imaginary exponential signals from the ciphertext images to obtain a product signal containing multiple image information, and then multiplying the product signal by a random phase template to obtain a product signal containing multiple image information in the form of noise;

[0145] Step S15, calculating the Fresnel inverse diffraction of the product signal to obtain a complex signal;

[0146] Step S16, using the amplitude of the complex signal and the unit amplitude template as constraints, using a phase recovery algorithm between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template.

[0147] Step S17, using the amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located as constraints, and using a phase recovery algorithm between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located to generate a first decrypted phase template;

[0148] Step S18, generating a second decryption phase template using each ciphertext image and the obtained first decryption phase template.

[0149] See also Figure 2 , the decryption steps include:

[0150] Step S21, using a common encryption phase template and two decryption phase templates in a cascade structure corresponding to each ciphertext image and a phase contrast filtering technique to obtain a ciphertext image;

[0151] Step S22, sending the ciphertext image to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext.

[0152] See also Figure 3 , Figure 3 This is the principle diagram of the phase truncation nonlinear encryption system. The input of the asymmetric encryption system is the plaintext image X, which is multiplied by the random phase template e jR1 Then the Fourier transform is calculated, and the amplitude of the calculated Fourier transform is multiplied by the random phase template e jR2 The inverse Fourier transform is then calculated and the obtained amplitude is used as the ciphertext C.

[0153] The asymmetric encryption process is expressed as:

[0154]

[0155] Where X represents the plaintext image, F represents the Fourier transform, and F -1 represents inverse Fourier transform, abs() represents amplitude; and Represents two random phase templates, R1 and R2 represent random phases.

[0156] The obtained ciphertext image C is used as the phase to construct the imaginary exponential signal e jC ;

[0157] Multiplying imaginary exponential signals from each image to obtain a product signal containing multiple images, and then multiplying the product signal by a random phase template to obtain a product signal containing multiple images in a noise form;

[0158] The product signal containing multiple images in the form of noise can be expressed as:

[0159]

[0160] Among them, C1, C2, ... C N Respectively represent the plaintext images X1, X2, ...X N The ciphertext image generated after encryption by the phase-truncated asymmetric linear system;

[0161] Calculate the inverse Fresnel diffraction transform of the product signal Z containing the noise form of multiple images:

[0162]

[0163] Where A represents the amplitude, Indicates phase.

[0164] The amplitude A of the complex signal and the unit amplitude are used as constraints, and a phase recovery algorithm is used to construct a public encrypted phase template.

[0165] See also Figure 4 , Figure 4 Here, P1 and P2 are the planes where the encrypted phase template E and the first decrypted phase template D1 are located respectively.

[0166] See also Figure 5 , Figure 5 It is a principle block diagram for generating an encrypted phase template E.

[0167] See also Figure 4 and Figure 5 , generating an encrypted phase template includes the following steps:

[0168] (1) The phase signal of the observation plane P1 where the encrypted phase template is located is initially set to a random phase signal;

[0169] (2) Calculate the Fresnel diffraction forward transform with wavelength λ and distance d1 to obtain the complex signal of the diffracted light wave at the P2 observation surface;

[0170] (3) The amplitude of the diffracted light wave signal at the P2 observation surface is replaced by the amplitude A generated in formula (3), and the phase is retained to obtain the diffracted light wave signal with amplitude constraint;

[0171] (4) calculating the inverse Fresnel diffraction transform of the amplitude-constrained signal at a distance of d1 to obtain the diffracted light wave signal on the P1 plane;

[0172] (5) The amplitude of the diffracted light wave signal in the P1 plane is constrained to 1, and the phase is retained, so that the complex signal of the object plane is obtained after being constrained to unit amplitude;

[0173] (6) The complex signal is used as the next initial object wave function for reiteration until the algorithm converges or the number of iterations is reached; the obtained P1 plane diffraction light wave signal is the encrypted phase template E.

[0174] At the end of the iteration, the generated encrypted phase template e jE Satisfies the relationship:

[0175]

[0176] The symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d1 represents the diffraction distance.

[0177] To restore the imaginary exponential signal e corresponding to the ciphertext image C1 jC1 For example, the process of generating the first and second decryption phase templates D1 and D2 using the phase recovery algorithm is as follows:

[0178] See also Figure 6 and Figure 7 , using the phase recovery algorithm to generate the first and second decrypted phase templates D1 and D2, P2 and P3 represent the planes where the first decrypted phase template and the second decrypted phase template are located respectively. Generating the first and second decrypted phase templates includes the following steps:

[0179] (1) The phase signal of the observation plane P2 where the decrypted phase template is located is initially set to a random phase signal, and the amplitude constraint is the amplitude A generated in formula (3);

[0180] (2) Calculate the Fresnel diffraction forward transform with wavelength λ and distance d2 to obtain the complex signal of the diffracted light wave at the P3 observation surface;

[0181] (3) The amplitude of the diffracted light wave signal at the P3 observation surface is constrained to be unit amplitude;

[0182] (4) calculating the inverse Fresnel diffraction transform of the amplitude-constrained signal at a distance d2 to obtain the diffracted light wave signal on the P2 plane;

[0183] (5) The phase of the diffracted light wave signal of the P2 plane is retained, and the amplitude is constrained to be the amplitude A generated in formula (3). The signal after the amplitude constraint is used as the next initial object wave function for re-iteration until the algorithm converges or the number of iterations is reached; the light wave signal of the P2 plane is first decrypted using the phase template D1.

[0184] The iteration ends, the first decryption phase e jD1 Satisfies the relationship:

[0185]

[0186] The symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d2 represents the diffraction distance.

[0187] The second decrypted phase template e jD2 The generation of is shown in formula (6).

[0188]

[0189] Among them, C1 represents the ciphertext corresponding to the plaintext image X1, and ang() represents the phase operation.

[0190] See also Figure 8, build a virtual optical path, place the encrypted phase template E, the first decrypted phase template D1 and the second decrypted phase template D2 corresponding to the decrypted i-th image on a specific plane of the virtual optical path, and combine with the phase contrast filter circuit to reconstruct the ciphertext image Ci corresponding to the i-th plaintext image on the output plane. By changing the decrypted phase template, the ciphertext image corresponding to each plaintext image can be reconstructed.

[0191] Figure 8 In the above equation, E represents the encrypted phase template, D1 and D2 represent the first and second decrypted phase templates respectively, and the plane signal where the decrypted phase template D2 is located can be expressed as:

[0192]

[0193] Decrypt the signal on the plane where the phase template D2 is located After the phase contrast filter circuit, the output surface will obtain the ciphertext image C whose amplitude signal corresponds to the plaintext image Xi. i The information processing process of phase contrast filtering is: the diffracted light wave signal of the D2 plane Calculate the Fourier transform, and the phase contrast filter located on the spectrum plane realizes the attenuation of the zero-order image and phase shifts it by 90 degrees. Then, after the inverse Fourier transform, the ciphertext C is obtained on the output plane. i , which can be expressed by formula (8).

[0194]

[0195] Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, W1 represents the zero-order image window, e jCi Indicates phase C i The imaginary exponential function of .

[0196] The ciphertext C i Send it to the trained ANN neural network to obtain the plaintext image X i The ANN network structure is as follows Fig.10 shown.

[0197] The quality of the reconstructed image can be evaluated using the correlation coefficient, which is defined as shown in formula (8).

[0198] For signals f1 and f2, the normalized cross-correlation coefficient NC can be calculated using formula (8).

[0199]

[0200] Where M and N are the image width and height.

[0201] Please refer to FIG. 11 , which shows six original images and the generated public encrypted phase template E.

[0202] Fig.12The first column is the original plaintext image, the second column is the first decrypted phase template corresponding to each image generated by the phase recovery algorithm, in which the number of iterations is 1000, z1=500mm, and the wavelength is λ=532nm. The third column is the second decrypted phase template iteratively generated between the planes where the first and second decrypted phase templates are located by the phase recovery algorithm, with the number of iterations being 1000, z2=300mm, and the wavelength being λ=532nm. The fourth column is the image reconstructed by the ANN neural network.

[0203] Fig.13 Taking the image of the number '2' as an example, the decryption results when the decryption phase template is damaged to different degrees are given. From the experimental results, it can be seen that if the decryption phase template 1 or the decryption phase template 2 is missing, the ANN neural network will decrypt an incorrect image; when the error rate of the decryption phase template 1 does not exceed 2%, the correct original plaintext image can still be decrypted to a certain extent. If the error rate of the decryption phase template 2 exceeds 5%, the correct original plaintext image cannot be decrypted.

[0204] Fig.14 Taking the digital '2' picture as an example, the robustness test results of the encrypted phase template E are given. From the simulation experimental results, it can be seen that the encrypted phase template E has a certain robustness to Gaussian noise, Poisson noise, salt and pepper noise, and multiplicative interference.

[0205] Fig.14 Robustness test results of encrypted phase template E

[0206] Fig.15 The reconstruction results of the reconstruction parameter d1 under different deviation rates are shown, and reconstruction is performed using d1=d1*(1+deviation rate). The experimental results show that when the absolute value of the deviation rate exceeds 0.01, an erroneous image will be decrypted.

[0207] Fig.16 The reconstruction results of reconstruction parameter d2 under different deviation rates are shown, and reconstruction is performed using d2=d2*(1+deviation rate). The experimental results show that when the absolute value of the deviation rate exceeds 0.01, an erroneous image will be decrypted.

[0208] The present invention adopts the above technical scheme. First, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image; an artificial neural network model (ANN) is built, and the plaintext and ciphertext images are used to train the artificial neural network (ANN) to obtain a trained neural network model that can be used for decryption; the ciphertext image is used as the phase to construct an imaginary exponential signal; the imaginary exponential signals from each ciphertext image are multiplied to obtain a product signal containing multiple image information, and the product signal is multiplied by a random phase template to obtain a product signal containing multiple image information in the form of noise; the Fresnel inverse diffraction of the product signal is calculated to obtain a complex signal; the amplitude and unit amplitude template of the complex signal are used as constraints, and a phase recovery algorithm is used between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template. The amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located are used as constraints, and a first decrypted phase template is generated between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located using a phase recovery algorithm; and then the second decrypted phase template is generated using each ciphertext image and the obtained first decrypted phase template. The ciphertext image is obtained by using a public encryption phase template and two decryption phase templates in the form of a cascade structure corresponding to each ciphertext image and phase contrast filtering technology; the ciphertext image is sent to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext. The test results show that the proposed method has good security and robustness. Any missing decryption phase template or mismatched reconstruction parameters will lead to the failure of reconstruction of the original plaintext image. ANN-assisted decryption allows the original image to be restored without the private key in the traditional phase truncation nonlinear encryption system, simplifying the decryption process. Encoding the ciphertext into phase and combining it with phase contrast filtering decryption further improves the security of the algorithm.

[0209] Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. In the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians of the art without making creative work are within the scope of protection of the present application.

Claims

1. A multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption, characterized in that: It includes an encryption step and a decryption step, specifically including the following steps: Encryption steps: Step S11, firstly, the original plaintext image is encrypted by phase truncation nonlinear encryption to obtain a ciphertext image; Step S12, building an artificial neural grid model, and training the artificial neural grid model with plaintext and ciphertext image pairs to obtain a neural network model that has been trained for decryption; Step S13, using the ciphertext image as the phase to construct an imaginary exponential signal; Step S14, multiplying the imaginary exponential signals from the ciphertext images to obtain a product signal containing multiple image information, and then multiplying the product signal by a random phase template to obtain a product signal containing multiple image information in the form of noise; Step S15, calculating the Fresnel inverse diffraction of the product signal to obtain a complex signal; Step S16, using the amplitude of the complex signal and the unit amplitude template as constraints, using a phase recovery algorithm between the plane where the encrypted phase template is located and the plane where the first decrypted phase template is located to obtain a common encrypted phase template; Step S17, using the amplitude and unit amplitude generated by the encrypted phase template on the plane where the first decrypted phase template is located as constraints, and using a phase recovery algorithm between the plane where the first decrypted phase template is located and the plane where the second decrypted phase template is located to generate a first decrypted phase template; Step S18, generating a second decryption phase template using each ciphertext image and the obtained first decryption phase template; Decryption steps: Step S21, using a common encryption phase template and two decryption phase templates in a cascade structure corresponding to each ciphertext image and a phase contrast filtering technique to obtain a ciphertext image; Step S22, sending the ciphertext image to the trained artificial neural network ANN to obtain the plaintext image corresponding to the ciphertext.

2. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The asymmetric encryption process in step S11 is expressed by the formula: Among them, X represents the plaintext image, F represents the Fourier transform, and F -1 represents inverse Fourier transform, abs() represents amplitude; and Represents two random phase templates, and R1 and R2 represent two random phases.

3. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The product signal of the noise form including the noise form of multiple images in step S14 is expressed as: Among them, C1, C2, ...C N Respectively represent the plaintext images X1, X2, ...X N The ciphertext image generated after encryption by the phase-truncated asymmetric linear system.

4. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: In step S15, the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images is calculated, and the specific expression is as follows: Where A represents the amplitude, represents phase; IFrT represents inverse Fresnel diffraction transform.

5. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The specific steps of generating the encrypted phase template in step S16 are as follows: Step S16-1, the phase signal of the observation plane P1 where the encrypted phase template is located is initially set to a random phase signal; Step S16-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d1, and obtaining a complex signal of the diffracted light wave of the P2 observation surface; Step S16-3, replacing the amplitude of the diffracted light wave signal of the P2 observation surface with the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images, and retaining the phase to obtain the diffracted light wave signal after amplitude constraint; Step S16-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance d1 to obtain a diffracted light wave signal on the P1 plane; Step S16-5, the amplitude of the diffraction light wave signal of the P1 plane is constrained to 1, the phase is retained, and the object plane complex signal after being constrained to unit amplitude is obtained; Step S16-6, the object plane complex signal is used as the next initial object wave function for re-iteration until the algorithm converges or reaches the number of iterations; The obtained P1 plane diffraction light wave signal is the encrypted phase template E. The encrypted phase template e generated after the iteration is completed jE Satisfies the following relationship: Wherein, the symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d1 represents the diffraction distance.

6. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: Step S17 specifically includes the following steps: Step S17-1, the phase signal of the observation plane P2 where the decrypted phase template is located is initially set to a random phase signal, and the amplitude constraint is the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images; Step S17-2, calculating the Fresnel diffraction forward transform with a wavelength of λ and a distance of d2, and obtaining a complex signal of the diffracted light wave of the P3 observation surface; Step S17-3, constraining the amplitude of the diffracted light wave signal of the P3 observation surface to be a unit amplitude; Step S17-4, calculating the Fresnel diffraction inverse transform of the amplitude-constrained signal at a distance d2 to obtain a diffracted light wave signal on the P2 plane; Step S17-5, retain the phase of the diffracted light wave signal of the P2 plane, and constrain the amplitude to the amplitude A obtained by the Fresnel diffraction inverse transform of the product signal Z in the form of noise containing multiple images; the signal after amplitude constraint is used as the next initial object wave function for re-iteration until the algorithm converges or the number of iterations is reached; the obtained light wave signal of the P2 plane is first decrypted using the phase template D1; the iteration ends, and the first decrypted phase Satisfies the relationship: The symbol abs represents the modulo operation, FrT represents the Fresnel diffraction operation, λ is the wavelength, and d2 represents the diffraction distance.

7. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 6, characterized in that: The second decrypted phase template in step S18 The specific calculation formula is as follows: Among them, C i Representation and plaintext image X i The corresponding ciphertext, ang() represents the phase operation.

8. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The specific method of step S21 is: build a virtual optical path, and place the encrypted phase template E, the first decrypted phase template D1 corresponding to the decrypted i-th image, the second decrypted phase template D2 corresponding to the decrypted i-th image, and the phase contrast filter circuit in sequence on a specific plane of the virtual optical path, and a lens is provided between the decrypted phase template D2 and the phase contrast filter circuit, and between the phase contrast filter circuit and the output surface; and then reconstruct the ciphertext image C corresponding to the i-th plaintext image on the output surface. i ; By changing the decryption phase template corresponding to each image respectively, the ciphertext image corresponding to each plaintext image is reconstructed.

9. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The specific steps of step S21 are: Step S21-1, calculate and obtain the plane signal where the decrypted phase template D2 is located, and the specific expression is as follows: Wherein, E represents the encrypted phase template, D1 and D2 represent the first and second decrypted phase templates respectively, Step S21-2: the second decryption phase template D2 is located on the plane signal After the phase contrast filter circuit, the output surface will obtain the amplitude signal corresponding to the plaintext image X i The ciphertext image C i , the specific calculation formula is as follows: Where abs() represents the modulo function, the symbol F represents the calculation of Fourier transform, and W1 represents the zero-order image window. Indicates phase C i The imaginary exponential function of .

10. The multi-image encryption method combining phase contrast filtering and convolutional neural network assisted decryption according to claim 1, characterized in that: The quality of the plaintext image reconstructed in step S22 is evaluated by the correlation coefficient. The calculation formula of the normalized correlation coefficient NC is as follows: Among them, m and n are the horizontal and vertical coordinates of the pixel points of the image respectively; M and N are the image width and height respectively; f1 and f2 are two signals respectively.