A method for restoring information-hidden images based on symbiotic biological search

Abstract

The present invention discloses a method for restoring information hidden images based on symbiotic biological search, which is characterized by comprising the following steps: 1) ecosystem initialization for hidden image restoration; 2) mutualistic symbiosis stage for hidden image restoration; 3) partial symbiosis stage for hidden image restoration; 4) parasitic stage for hidden image restoration. The hidden image restored by this method not only has good visual quality but also has a higher peak signal-to-noise ratio.

Description

Technical Field

[0001] The invention relates to image processing technology, in particular to an information hiding image restoration method based on symbiotic organism search. Background Art

[0002] With the advent of the 5G era, the transmission rate of the network has been greatly improved. People can more conveniently and quickly obtain a lot of information in different forms from the Internet, including text, pictures, videos, etc. Although this has greatly enriched people's lives, it will also cause a lot of information leakage.

[0003] As one of the most common carriers of information, hiding information in images can greatly increase the amount of information transmission while protecting the information well. However, since images with hidden information need to be uploaded, transmitted, downloaded, etc. on the Internet, they will be interfered by interference sources of different degrees and types. Therefore, after the secret information is extracted from the image, it will also be damaged to varying degrees. If the secret information is an image, that is, the secret image is hidden in another image, after being transmitted on the Internet, the secret image may be greatly different from the original secret image when it is extracted.

[0004] If both the secret information and the carrier are images, after the secret image is damaged by noise, the extracted hidden image will also be damaged. Mark the damaged bits of the hidden image. If the high 5 bits of the hidden image pixel are not damaged, the pixel is less different from the original pixel. Such pixels are called credible pixels, otherwise they are untrustworthy pixels. Determine the distortion range of untrustworthy pixels, use the swarm intelligence symbiotic biological search algorithm to search for repair values, and restore the hidden image. Summary of the invention

[0005] The purpose of the present invention is to provide a method for restoring information hidden images based on symbiotic biological search in view of the shortcomings of the prior art. The hidden images restored by this method not only have good visual quality but also have a higher peak signal-to-noise ratio.

[0006] The technical solution for achieving the purpose of the present invention is:

[0007] An information hiding image restoration method based on symbiotic biological search comprises the following steps:

[0008] 1) Initialization of the ecosystem for hidden image recovery: To recover the hidden image, we need to repair the untrusted pixels. Assume that m bits of the untrusted pixel p are damaged. Since the damaged bits are either 0 or 1, there are 2 m possible values, let k = 2 m , k is the number of hidden image symbiotic organisms search ecosystem populations, and the individual ecosystem population established by the k possible values ​​of the untrustworthy pixel p is X = {p1, p2, p3, …, pk}, the maximum value of population X is ub=max{p1,p2,p3,…,p k}, the minimum value is lb=min{p1,p2,p3,…,p k}, the range of ecosystem search is: [lb, ub], and the maximum number of iterations of the search is 2 m , use variance to represent fitness. The smaller the variance, the better the fitness, and the larger the variance, the worse the fitness. Find the credible pixels in the neighborhood of p, calculate the variance of each individual in the population X and the credible pixels in the neighborhood, and select the individual with the best fitness as the current optimal solution Pbest;

[0009] 2) Mutual benefit phase of hidden image recovery: From the population X of hidden image pixels p, randomly select two possible values ​​p i and p j And i≠j, calculate p i and p j The relationship feature vector between:

[0010] MV=(p i +p j ) / 2,

[0011] Possible values ​​of hidden image pixel mutualism:

[0012] p inew =p i +α*(Pbest-MV*BF1),

[0013] p jnew =p j +β*(Pbest-MV*BF2),

[0014] Among them, BF1 and BF2 are benefit factors, the benefit factor value is 0 or 1, α, β are random numbers in the interval [0,1],

[0015] Calculate p inew 、p jnew With the variance of the neighborhood credible pixels: If the variance var(p inew )<variancevar(p i ), then use p inew Replace p i Update the population X of hidden image pixel p, otherwise, p i remains unchanged. Similarly, if the variance var(p jnew )<variancevar(p j ), then use p jnew Replace the hidden image pixel p in the population X j , otherwise, p j remain unchanged;

[0016] 3) The partial benefit symbiosis stage of hidden image recovery: From the population X of hidden image pixel point p, randomly select another pixel possible value p l (i≠l), calculate p i From p l The profit situation in the symbiotic relationship, while p l Keeping the same, the calculation method is as follows:

[0017] p inew =p i +δ*(Pbest-p l ),

[0018] Among them, δ is a random number between [-1,1], calculate p inew and the variance of the credible pixels in the neighborhood, if the variance var(p inew )<variancevar(p i ), then use p inew Replace the hidden image pixel p of the population X i , otherwise, p i remain unchanged;

[0019] 4) Parasitic stage of hidden image recovery: randomly generate a pixel estimate in the interval [lb, ub]:

[0020] p inew =lb+φ*(ub-lb),

[0021] Where φ is a random number in the interval [0,1], calculate p inew and the variance of the credible points in the neighborhood, if the variance var(p inew )<variancevar(p i ), then use p inew Replace the hidden image pixel p of the population X i , otherwise, p i Keep it unchanged, calculate the variance of each individual in population X and the credible pixels in the neighborhood of pixel p, select the individual with the best fitness as the current optimal solution Pbest, and repeat steps 2), 3), and 4). m Second, find the value with the smallest variance in the population X of the restored image pixel p, and use it as the repair value of the restored pixel p to restore the hidden image.

[0022] The hidden image restored by this method not only looks good but also has a higher peak signal-to-noise ratio. DETAILED DESCRIPTION

[0023] The present invention is further described below with reference to the embodiments, but is not intended to be limiting of the present invention.

[0024] Example:

[0025] An information hiding image restoration method based on symbiotic biological search comprises the following steps:

[0026] 1) Initialization of the ecosystem for hidden image recovery: To recover the hidden image, we need to repair the untrusted pixels. Assume that m bits of the untrusted pixel p are damaged. Since the damaged bits are either 0 or 1, there are 2 m possible values, let k = 2 m , k is the number of hidden image symbiotic organisms search ecosystem populations, and the individual ecosystem population established by the k possible values ​​of the untrustworthy pixel p is X = {p1, p2, p3, …, p k}, the maximum value of population X is ub=max{p1,p2,p3,…,p k}, the minimum value is lb=min{p1,p2,p3,…,p k}, the range of ecosystem search is: [lb, ub], and the maximum number of iterations of the search is 2 m , use variance to represent fitness, the smaller the variance, the better the fitness, the larger the variance, the worse the fitness, find the credible pixels in the neighborhood of p, calculate the variance var(p) between each individual in population X and the credible pixels in the neighborhood i ), select the individual with the best fitness as the current optimal solution Pbest. In this example, assume that the central pixel p=132 of the image block is an untrustworthy pixel, that is, p is the distorted pixel to be repaired. There are two credible points {157, 152} in the 3×3 neighborhood centered on p. Since the 4th and 5th bits of the pixel are damaged, the number of damaged bits m=2, and the damaged bits are randomly assigned 0 or 1, forming the initial population X={132, 140, 148, 156} of the pixel p ecosystem, with a total of 2 2 =4 individuals, the maximum value is ub=156, the minimum value is lb=132, that is, the maximum number of iterations of the search is 4, the range of the ecosystem search is: [132,156], the variance of the initial pixel population X, calculate the variance of the pixel value of each population individual and the credible point, and get {175,76.3333,20.3333,7}, select the individual with the smallest variance, that is, the best fitness as the current optimal solution Pbest=156;

[0027] 2) Mutual benefit phase of hidden image recovery: From the population X of hidden image pixels p, randomly select two possible values ​​p i and p j And i≠j, calculate p i and p j The relationship feature vector between:

[0028] MV=(p i +p j ) / 2,

[0029] Possible values of the mutualistic symbiosis relationship of hidden image pixels:

[0030] p inew = p i + α * (Pbest - MV * BF1),

[0031] p jnew = p j + β * (Pbest - MV * BF2),

[0032] where BF1 and BF2 are benefit factors, and the values of the benefit factors are 0 or 1, and α, β are random numbers in the interval [0, 1], and calculate p inew p jnew and the fitness of the neighboring trusted pixels: If variance var(p inew ) < variance var(p i ), then replace p inew with p i and update the population X of the hidden image pixel p, otherwise, p i remains unchanged. Similarly, p jnew is processed in the same way. In this example, from the population X of the hidden image pixel p, randomly select a possible value p1 = 132, and then randomly select another possible value p2 = 140, and calculate the relationship feature vector between p1 and p2:

[0033] MV = (132 + 140) / 2 = 136,

[0034] Assume that the benefit factors BF1 = 1 and BF2 = 0, the random numbers α = 0.5, β = 0.5, and the possible values of the mutualistic symbiosis relationship of the hidden image pixels are as follows:

[0035] p 1new = 132 + 0.5 * (156 - 136 * 1) = 142,

[0036] p 2new = 140 + 0.5 * (156 - 136 * 1) = 150,

[0037] Calculate and compare the variances of the new individuals and the original individuals, var(142) = 58.3333 < var(132) = 175 and var(150) = 13 < var(140) = 76.3333, replace 132 with 142, replace 140 with 150, and obtain the new population X of the hidden image pixel p = {142, 150, 148, 156};

[0038] 3) Commensalism stage of hidden image restoration: Randomly select another pixel possible value p j from the population X of the hidden image pixel p, and calculate pi The profit obtained from the symbiotic relationship with p j , while p j remains unchanged, and the calculation method is as follows:

[0039] p inew = p i + δ * (Pbest - p j ),

[0040] where δ is a random number between [-1, 1], calculate the variance of p inew and the reliable pixels in the neighborhood. If var(p inew ) < var(p i ), then replace p inew with p i and update the population X of the hidden image pixel p. Otherwise, p i remains unchanged. In this example, from the population X of the hidden image pixel p, randomly select another possible pixel value p3, calculate the profit obtained from the symbiotic relationship between 142 and 148, while 148 remains unchanged. Assume the random value δ = 0.2, and the calculation method is:

[0041] p 1new = 142 + 0.2 * (156 - 148) = 143.6,

[0042] Since var(143.6) = 45.8533 < var(142) = 58.3333, therefore, replace 142 with 143.6 and update the population X of the hidden image pixel p = {143.6, 150, 148, 156};

[0043] 4) Parasitic stage of hidden image recovery: Randomly generate a pixel estimate value in the interval [lb, ub]:

[0044] p inew = lb + φ * (ub - lb),

[0045] where φ is a random number in the interval [0, 1], calculate the variance of p inew and the reliable points in the neighborhood. If var(p inew ) < var(p i ), then replace p inew with p i , and update the population X of the hidden image pixel p. Otherwise, p i remains unchanged. Repeat steps 2), 3), and 4) until the number of iterations reaches 2 m, find the value with the smallest variance in the population X of the restored image pixel p, and use it as the repair value of the restored pixel p to restore the hidden image. In this example, a pixel estimation value is randomly generated in the interval [132,156], assuming φ = 0.3:

[0046] p 1new =132+0.3*(156-143.6)=135.72,

[0047] Since var(135.72)=123.8128>var(143.6)=45.8533, 143.6 remains unchanged. The new population pixel X={143.6,150,148,156} of pixel p is calculated again. The variance corresponding to each individual is {45.8533,13,20.3333,7}. Since 156 has the smallest variance, 156 is selected as the current optimal solution. Repeat steps 2), 3), and 4) for 4 times. Finally, the value with the smallest variance in population X is selected as the repair value of pixel p.

Claims

1. A method for restoring an image with information hiding based on symbiotic biological search, characterized in that: The steps include: 1) Ecosystem initialization for hidden image recovery: Assume that the untrusted pixel p has m bits destroyed, and the destroyed bits are either 0 or 1, then there are 2 m possible values, let k = 2 m , k is the number of hidden image symbiotic organisms search ecosystem populations, and the individual ecosystem population established by the k possible values ​​of the untrustworthy pixel p is X = {p1, p2, p3, …, p k }, the maximum value of population X is ub=max{p1,p2,p3,…,p k }, the minimum value is lb=min{p1,p2,p3,…,p k }, the range of ecosystem search is: [lb, ub], and the maximum number of iterations of the search is 2 m , use variance to represent fitness, the smaller the variance, the better the fitness, the larger the variance, the worse the fitness, find the credible pixels in the neighborhood of p, calculate the variance var(p) between each individual in population X and the credible pixels in the neighborhood i ), select the individual with the best fitness as the current optimal solution Pbest; 2) Mutual benefit phase of hidden image recovery: From the population X of hidden image pixels p, randomly select two possible values ​​p i and p j And i≠j, calculate p i and p j The relationship feature vector between: MV=(p i +p j ) / 2, Possible values ​​of hidden image pixel mutualism: p inew =p i +α*(Pbest-MV*BF1), p jnew =p j +β*(Pbest-MV*BF2), Among them, BF1 and BF2 are benefit factors, the benefit factor value is 0 or 1, α, β are random numbers in the interval [0,1], calculate p inew 、p jnew Fitness with neighborhood credible pixels: If the variance var(p inew )<variancevar(p i ), then use p inew Replace p i , and update the population X of hidden image pixel p, otherwise, p i Keep unchanged, similarly, p jnew Same treatment; 3) The partial benefit symbiosis stage of hidden image recovery: From the population X of hidden image pixel point p, randomly select another pixel possible value p j , calculate p i From p j The profit situation in the symbiotic relationship, while p j Keeping the same, the calculation method is as follows: p inew =p i +δ*(Pbest-p j ), Among them, δ is a random number between [-1,1], calculate p inew and the variance of the credible pixels in the neighborhood, if var(p inew ) <var(p i ), then use p inew Replace p i , and update the population X of hidden image pixel p, otherwise, p i remain unchanged; 4) Parasitic stage of hidden image recovery: randomly generate a pixel estimate in the interval [lb, ub]: p inew =lb+φ*(ub-lb), Where φ is a random number in the interval [0,1], calculate p inew and the variance of the credible points in the neighborhood, if var(p inew ) <var(p i ), then use p inew Replace p i , and update the population X of hidden image pixel p, otherwise, p i Keep unchanged and repeat steps 2), 3), and 4) until the number of iterations reaches 2 m , find the value with the smallest variance in the population X of the restored image pixel p, and use it as the repair value of the restored pixel p to restore the hidden image.