A robust image steganography method based on low-frequency embedding with no background noise

Abstract

This invention discloses an error-resistant robust image steganography method based on low-frequency embedding without background noise. The method includes: preprocessing and hybrid encoding the original secret image to generate a final secret embedding stream; performing a 5 / 3 integer wavelet transform on the carrier image to decompose the frequency domain, extracting the first-level low-frequency sub-band, calculating texture activity in blocks, and generating an embedding mask; setting the quantization step size and jitter bias based on quantization index modulation, adjusting the block mean based on the bits to be embedded, and superimposing the offset to complete the embedding; performing integer wavelet transform and block statistics on the received image, obtaining the initial bit stream using the minimum distance method of quantization index modulation, and reconstructing the output secret image through deinterleaving, error correction decoding, inverse whitening, and inverse scrambling. This invention solves the problems of poor security and weak resistance to burst errors in traditional steganography, achieving a balance between robustness and concealment, and improving the reliability of covert communication in complex channels.

Description

Technical Field

[0001] This invention relates to the field of multimedia security and covert communication technology, specifically to an error-resistant robust image steganography method based on low-frequency embedding without background noise. Background Technology

[0002] Digital steganography, as a core technology of covert communication, aims to hide secret information using digital media such as images and videos, achieving imperceptible transmission. With the widespread adoption of social media and cloud computing, information shielding on social networks has become a research hotspot. Unlike traditional solutions that focus on embedding capacity, real-world image carriers require lossy compression, format conversion, and size scaling, placing stringent demands on steganography robustness. Furthermore, specialized fields such as medical imaging, military mapping, and forensic identification require pixel-level lossless reconstruction of the secret information after transmission of the steganographically transmitted images.

[0003] Traditional robust steganography often relies on frequency domain algorithms such as Discrete Wavelet Transform (DWT) or Discrete Cosine Transform (DCT) to embed the payload by modifying the transform domain coefficients. However, these methods face several challenges in practical applications. First, frequency domain transformations involve a large number of floating-point operations, and numerical truncation and rounding during spatial pixel reconstruction can introduce rounding errors, interfering with the determination of secret information, increasing the bit error rate, and compromising the reversibility of steganography. Second, when facing geometric attacks such as pruning and partial overlay, the spatial distribution characteristics of the secret bits make it easy to completely lose local information, making it difficult to guarantee communication reliability.

[0004] To improve survivability in harsh channel environments, researchers have attempted to introduce error-correcting coding techniques such as Reed-Solomon (RS) codes or convolutional codes. However, single error-correcting codes perform poorly when faced with burst errors, such as when a portion of an image is cropped or obscured. When errors occur consecutively and exceed the error correction threshold, the steganalytic payload cannot be resolved. Meanwhile, how to leverage the robustness of low-frequency coefficient WIS while effectively controlling embedding distortion to circumvent expensive steganalysis detection remains a crucial balancing act in the field.

[0005] In summary, while existing technologies have made breakthroughs in single metrics, they still suffer from shortcomings in complex channel environments, such as the inability to balance robustness and reversibility, difficulty in resisting sudden geometric attacks, and limitations due to floating-point transformation rounding errors. Therefore, an error-resistant robust image steganography method based on low-frequency embedding without background noise is proposed. Summary of the Invention

[0006] The main objective of this invention is to provide an error-resistant robust image steganography method based on low-frequency embedding without background noise. Existing technologies still suffer from problems such as inability to balance robustness and reversibility in complex channel environments, difficulty in resisting sudden geometric attacks, and limitations due to floating-point transformation rounding errors.

[0007] Based on the first main aspect of the present invention, an error-resistant robust image steganography method based on low-frequency embedding without background noise is provided, comprising the following steps performed by a computer system:

[0008] Input the original secret image and the carrier image, binarize and serialize the original secret image to obtain the secret image, and generate a one-dimensional binary bit sequence based on the secret image;

[0009] The statistical fingerprint is eliminated by double randomization of the one-dimensional binary bit sequence, and burst errors in the channel are transformed into random errors by global interleaving to generate the final secret embedding stream.

[0010] Perform a 5 / 3 integer wavelet transform on the carrier image to decompose the frequency domain into two integer operation steps to obtain a first-level low-frequency sub-band. Calculate the texture activity of the first-level low-frequency sub-band into several blocks divided according to a preset side length, and generate an embedding mask based on the texture activity.

[0011] Based on quantization index modulation, the quantization step size and jitter offset are set. The average value of the block to be embedded is adjusted according to the bit to be embedded in the final secret embedding stream and the offset is added to complete the embedding and obtain the received image.

[0012] Integer wavelet transform and block statistics are performed on the received image, and hard decision is made on the feature mean using the minimum distance rule of quantization index modulation to obtain the initial bit stream;

[0013] Based on the initial bitstream, the secret image is reconstructed and output through deinterleaving, error correction decoding, inverse whitening, and inverse scrambling.

[0014] As a further preferred embodiment, in the aforementioned method, the steps for generating a one-dimensional binary bit sequence based on the secret image are as follows:

[0015] Based on the secret image, a robust threshold is used to discretize the image pixels into a one-dimensional binary bit sequence containing only 0 or 1;

[0016] The dual randomization includes spatial scrambling and statistical whitening;

[0017] The execution steps for spatial scrambling are as follows:

[0018] A global random permutation is generated using the first key, and the spatial order of the one-dimensional binary bit sequence is shuffled based on the global random permutation.

[0019] The steps for statistical whitening are as follows:

[0020] A pseudo-random mask sequence following a uniform distribution is generated using the second key, and an XOR operation is performed with a spatially scrambled one-dimensional binary bit sequence.

[0021] The steps for converting burst errors in the channel into random errors through global interleaving include error correction coding and global interleaving processing, as detailed below:

[0022] The statistically whitened one-dimensional binary bit sequence is mapped to a byte stream, and redundant check bits are added to it using Reed-Solomon coding and interleaving techniques to generate an encoded sequence.

[0023] The encoded sequence is scrambled globally using a global interleaving map to generate the final secret embedding stream.

[0024] As a further preferred embodiment, in the aforementioned method, the two integer operation steps include a prediction step and an update step;

[0025] The execution steps of the prediction step are as follows:

[0026] By using the average of two adjacent even-numbered pixels, the pixel at the middle odd-numbered position is predicted, thus obtaining the residual that reflects local abrupt changes in the image.

[0027] The execution steps of the update step are as follows:

[0028] Based on the residuals that reflect local abrupt changes in the image, even-numbered pixels are smoothed to obtain the low-frequency components of the corresponding image.

[0029] As a further preferred embodiment, in the aforementioned method, the steps for adjusting the mean of the bit to be embedded block and adding an offset to complete the embedding are as follows:

[0030] For the first-level low-frequency sub-band selected by the embedded mask, the blocks are divided according to the custom side length. The average value of all wavelet coefficients in the block is calculated to obtain the block mean of the current block.

[0031] Based on the bits to be embedded, set the corresponding jitter offset and obtain the corrected block average using the following formula;

[0032] The formula is:

[0033]

[0034] in, This represents the corrected block mean. Indicates the quantization step size. The quantization factor represents the quantization step size. Represents the block average. This indicates the jitter offset. Indicates the bit to be embedded;

[0035] Subsequently, based on the corrected block mean and jitter offset, the offset is calculated and superimposed on all pixels of the current block to complete the embedding.

[0036] As a further preferred embodiment, in the aforementioned method, the execution step of obtaining the initial bitstream is calculated as shown in the following formula:

[0037]

[0038] in, Indicates the initial bit stream, This indicates taking the bit index corresponding to the minimum distance. Indicates bit index, Represents the characteristic mean. The quantization factor represents the quantization step size. The residual represents the difference between the predicted value and the actual value. Indicates the quantization step size.

[0039] As a further preferred embodiment, in the aforementioned method, the deinterleaving execution step is as follows:

[0040] Based on the initial bit stream, an estimate of the final secret embedding stream is obtained, and the errors in the set are discretized through the inverse mapping of the global interleaving map;

[0041] The execution steps of the error correction decoding are as follows:

[0042] After deinterleaving, the Reed-Solomon algorithm is used to complete the first round of error correction decoding to obtain the intermediate sequence.

[0043] As a further preferred embodiment, in the aforementioned method, the reverse whitening includes:

[0044]

[0045] in, This represents the spatially scrambled sequence after inverse whitening. Indicates an intermediate sequence. This represents the XOR operation. Represents a pseudo-random mask sequence;

[0046] The reverse scrambling process includes:

[0047]

[0048] in, The secret image bits represent the final reconstructed image. Indicates bit index, This represents the spatially scrambled sequence after inverse whitening. This represents the inverse mapping of the global mapping for space scrambling.

[0049] Based on a second key aspect of the present invention, an error-resistant robust image steganography system based on low-frequency embedding without background noise is provided for implementing the aforementioned method, comprising:

[0050] The preprocessing and hybrid coding stream module is used to eliminate statistical fingerprints through double randomization and to transform burst errors in the channel into random errors through global interleaving;

[0051] A reversible embedding domain construction module is used to perform a 5 / 3 integer wavelet transform on the carrier image to obtain the first-level low-frequency subband, and generate an embedding mask based on the texture activity.

[0052] The quantization embedding module is used to set the quantization step size and jitter offset based on the quantization index modulation, and to complete the embedding by adjusting the block mean and superimposing the offset according to the bit to be embedded.

[0053] The bar-blind extraction and timing verification logic module is used to perform integer wavelet transform and block statistics on the received image, obtain the initial bit stream through the minimum distance method of quantization index modulation, and perform deinterleaving, error correction decoding, inverse whitening, and inverse scrambling to reconstruct and output the secret image.

[0054] According to a third key aspect of the present invention, an electronic device is provided, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0055] The memory stores a computer program that, when executed by the processor, causes the processor to perform the error-resistant robust image steganography method based on low-frequency embedding without background noise, as described above.

[0056] Based on a fourth key aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed, implements the aforementioned error-resistant robust image steganography method based on low-frequency embedding without background noise.

[0057] Compared with existing technologies, this invention provides an error-resistant robust image steganography method based on low-frequency embedding without background noise. First, this invention adopts a 5 / 3 integer wavelet transform (IWT) combined with a low-frequency built-in embedding scheme, replacing the floating-point operations of traditional frequency domain transform with integer operations. This eliminates the cumulative rounding error basis generated by traditional floating-point transform at the underlying algorithm level, solving the problem that traditional frequency domain steganography methods cause the error basis to interfere with the decision of secret information, leading to an increased bit error rate and destruction of reversibility. This achieves lossless reconstruction of the carrier image and accurate decision of secret information.

[0058] Secondly, this invention constructs a cascaded anti-interference processing architecture of spatial scrambling, statistical whitening, RS coding, and global interleaving. By eliminating the spatial correlation and statistical characteristics of the secret sequence through dual randomization, and by leveraging the synergistic effect of RS error correction coding and global interleaving, the burst errors caused by geometric attacks on the channel transmitter are discretized into single-bit errors. This solves the difficulties of insufficient ability of single error correction coding to resist burst errors and the ease with which geometric attacks can cause the complete loss of secret information in parts, thus improving the survivability of the steganography scheme in complex channel environments such as lossy compression and pruning.

[0059] Finally, this invention designs a quantization index modulation embedding strategy based on jitter offset, which completes the secret embedding by means of block mean correction and superposition offset, enhancing the steganography of this invention to be practically applied in real-world environments. During transmission, it can resist various JPEG, noise, geometric scaling and combined attacks, giving it the physical characteristics to survive in harsh environments. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention.

[0061] Figure 1 The following is a flowchart illustrating the execution of an error-resistant robust image steganography method based on low-frequency embedding without background noise, according to one embodiment of the present invention.

[0062] Figure 2 This diagram illustrates the preprocessing and hybrid coding of the secret information in an error-resistant robust image steganography method based on low-frequency embedding without background noise, according to one embodiment of the present invention.

[0063] Figure 3 This illustration shows a schematic diagram of the embedding domain construction and adaptive block selection in an error-resistant robust image steganography method based on low-frequency embedding without background noise in one embodiment of the present invention;

[0064] Figure 4 This diagram illustrates the principle of quantization index modulation embedding based on robust threshold in an error-resistant robust image steganography method based on low-frequency embedding without background noise, according to one embodiment of the present invention.

[0065] Figure 5 This illustration shows a schematic diagram of the time-dependent robust blind extraction and verification logic in an error-resistant robust image steganography method based on low-frequency embedding without background noise, according to an embodiment of the present invention. Detailed Implementation

[0066] The preferred embodiments of the present invention will be described in detail below to provide a clearer understanding of the purpose, features, and advantages of the invention. It should be understood that the following embodiments are not intended to limit the scope of the invention, but are merely illustrative of the essential spirit of the technical solution of the invention.

[0067] In the following description, certain specific details are set forth for the purpose of illustrating various disclosed embodiments in order to provide a thorough understanding of the various disclosed embodiments. However, those skilled in the art will recognize that embodiments may be practiced without one or more of these specific details. In other instances, well-known techniques associated with the invention may not have been shown or described in detail to avoid unnecessarily obscuring the description of the embodiments.

[0068] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0069] The following is a description of the specific meanings of technical terms, English abbreviations, and formula parameters that may be used in this invention:

[0070] Statistical fingerprint: A unique identifier or feature representation constructed by extracting a set of quantitative statistical features of a time object, used to identify, classify, track or verify the identity, origin, integrity or attributes of a data object.

[0071] Global interleaving: A technique that rearranges the positions of bits or symbols in a encoded data sequence using a pre-defined global random permutation rule. Its core function is to disperse the continuous burst errors that may occur during channel transmission into discrete random errors, making the error distribution adaptable to the error correction capabilities of error-correcting codes, thereby improving the system's robustness against continuous errors.

[0072] Burst errors: Errors occurring in a continuous sequence of bits or symbols during data transmission. These errors are characterized by their continuity and are typically caused by interference such as channel fading, impulse noise, partial clipping or smearing, and lossy compression block effects.

[0073] Random errors: Independent, unrelated, and randomly distributed bit or symbol errors in data transmission. The error locations have no obvious continuity and are usually caused by random interference such as Gaussian white noise.

[0074] Discrete Wavelet Transform (DWT) is a time-frequency analysis technique that decomposes a signal into multi-scale frequency sub-bands. It maps the signal into low-frequency approximate components and high-frequency detail components through wavelet basis functions and is widely used in image compression, feature extraction and steganography.

[0075] Integer Wavelet Transform (IWT): An integer implementation of the discrete wavelet transform. By introducing integer operation constraints such as floor function, the transform coefficients are limited to integers, completely avoiding rounding errors caused by floating-point operations, ensuring the reversibility of the transform, and enabling distortion-free reconstruction of signals. It is one of the core technologies for reversible image steganography and lossless compression.

[0076] Texture activity: an indicator that quantifies the texture complexity of a local region of an image, usually obtained by calculating statistical features such as pixel gradient magnitude, variance, entropy, or high-frequency capability within the region.

[0077] Quantization Index Modulation (QIM) is a quantization-based information embedding technique that achieves hidden information embedding by quantizing the sampled values ​​of the carrier signal to different quantization intervals and mapping secret bits to the corresponding interval indices.

[0078] The Reed-Solomon algorithm is the core technology system for implementing Reed-Solomon encoding and decoding. It is based on polynomial operations in finite fields and is a symbol-level linear block error correction algorithm.

[0079] Minimum distance method: A pattern recognition and signal decision method based on distance metric criteria. Its core logic is to select the template with the smallest distance from the sample to be identified to multiple reference templates as the sample's category or decision result.

[0080] like Figure 1 As shown, in one embodiment, the error-resistant robust image steganography method based on low-frequency embedding without background noise of the present invention includes the following steps S1-S6 performed by a computer system:

[0081] S1, Input the original secret image and the carrier image, binarize and serialize the original secret image to obtain the secret image, and generate a one-dimensional binary bit sequence based on the secret image;

[0082] S2, the statistical fingerprint is eliminated by double randomization of the one-dimensional binary bit sequence, and the burst errors in the channel are transformed into random errors by global interleaving to generate the final secret embedding stream;

[0083] S3, perform a 5 / 3 integer wavelet transform on the carrier image, decompose the frequency domain into two integer operation steps to obtain a first-level low-frequency sub-band, calculate the texture activity of the first-level low-frequency sub-band divided into several blocks according to a preset side length, and generate an embedding mask based on the texture activity.

[0084] S4, based on quantization index modulation, set the quantization step size and jitter offset, adjust the block mean according to the bit to be embedded in the final secret embedding stream and add the offset to complete the embedding, and obtain the received image;

[0085] S5, perform integer wavelet transform and block statistics on the received image, and use the minimum distance rule of quantization index modulation to make a hard decision on the feature mean to obtain the initial bit stream;

[0086] S6. Based on the initial bitstream, the secret image is reconstructed and output through deinterleaving, error correction decoding, inverse whitening, and inverse scrambling.

[0087] In one feasible embodiment, the present invention provides an error-resistant robust image steganography method based on low-frequency embedding without background noise, through a detailed description of the following steps.

[0088] Combination Figure 2 As shown, the preprocessing stage of this invention is not just a simple bit conversion, but a three-level security protection system of "statistical whitening - redundancy correction - spatial discretization" is designed.

[0089] In the data information processing method, a defense system against channel interference was constructed. Its core is to eliminate statistical fingerprints through double randomization and to transform burst errors in the channel into random errors through global interleaving.

[0090] First, binarize and serialize the original secret image to obtain a size of [size missing]. Secret Images , here This indicates the side length of the image.

[0091] To facilitate streaming processing and improve robustness, a robustness threshold is used. Discretize the image pixels into a one-dimensional binary bit sequence. The mathematical expression is as follows:

[0092]

[0093] in, Represents a one-dimensional binary bit sequence. Represents the pixel grayscale value. Represents the pixel coordinates arranged in raster scan order. The function representing the pixel grayscale value of an image. Indicates the robustness threshold. Indicates bit index, Indicates the total length of the sequence. This indicates the side length of the image.

[0094] Secret Images They typically have strong spatial correlation. If directly embedded in the carrier image, the statistical features of the steganographic image are easily detected by a steganalysis analyzer.

[0095] Therefore, the present invention performs the following two steps.

[0096] First, spatial permutation: using the first key. Generate a global random permutation This disrupts the spatial order of bits.

[0097] This step ensures that if an attack damages a local area of ​​the carrier, the damaged bits will be distributed throughout the entire secret image during recovery, avoiding local blind spots.

[0098] Secondly, statistical whitening: utilizing a second key. Generate a pseudo-random mask sequence that follows a uniform distribution. The scrambled sequence is then XORed with the scrambled sequence.

[0099] This step forces the 0 / 1 distribution of the sequence to be embedded to approach 50%, making it statistically exhibit white noise characteristics.

[0100]

[0101] in, This represents the first bit of a one-dimensional binary bit sequence after statistical whitening. bit Represents the spatially scrambled one-dimensional binary bit sequence. bit This represents the XOR operation. Represents the pseudo-random mask sequence number 1 bit Indicates bit index.

[0102] Finally, completing error correction coding and global interleaving will cause issues with JPEG compression. The characteristic of consecutive bit errors within a block.

[0103] This invention introduces Reed-Solomon (RS) coding and interleaving techniques to statistically whiten the one-dimensional binary bit sequence, i.e., the whitened sequence. The mapping is converted into a byte stream, generating a sequence with redundant check bits, i.e., a sequence of encoded lines. .

[0104] This embodiment sets the number of verification symbols. To enable it to correct The ability to handle byte errors, among which... This indicates the number of byte errors that can be corrected.

[0105] For global interleaving techniques, global interleaving mapping is used. For the encoded sequence Perform a global shuffle to generate the final secret embedding stream. .

[0106] The physical significance of interleaving is to "dilute" the concentrated block errors in the channel into discrete single-bit errors in the decoding domain, thereby ensuring that the number of errors is always within the error correction tolerance of the RS code.

[0107] The specific formula is as follows:

[0108]

[0109] in, The final secret embedding stream is represented by the first... bit Represents the encoded sequence. Represents the first in the global interleaving map Index of a position, This represents the bit index in the final secret embedded stream.

[0110] Through the above preprocessing steps, the present invention has transformed the original binary secret image into a highly processed final secret embedding stream. .

[0111] This bitstream eliminates spatial correlation through scrambling, acquires pseudo-random statistical characteristics through whitening, and possesses the ability to correct sudden errors through RS coding and interleaving.

[0112] After completing the preprocessing and hybrid encoding of the secret information, combined with Figure 3 As shown, this invention constructs an embedding domain that possesses both mathematical invertibility and physical stability.

[0113] To fundamentally eliminate the floating-point rounding error problem commonly found in transform domain steganography, this invention abandons the traditional DWT and adopts a 5 / 3 integer wavelet transform based on the Lifting Scheme.

[0114] The lossless decomposition of the number domain performs 5 / 3 IWT on the carrier image C. The core of this method is to break down the complex frequency domain decomposition into two simple, reversible integer operation steps: prediction and update.

[0115] High-frequency details are captured by the prediction step, while the low-frequency profile is preserved by the update step.

[0116] Unlike traditional floating-point conversions, IWT introduces a floor operator to ensure... Transformation coefficients Transformation coefficients Transformation coefficients All transformation coefficients are integers.

[0117] This ensures that the reconstruction of the carrier data is bit-accurate in the absence of attacks, avoiding hash verification failures caused by transformation errors.

[0118] The specific formula for improvement is as follows:

[0119] Predict step:

[0120]

[0121] in, The residual represents the difference between the predicted value and the actual value. Indicates adjacent even-numbered pixel positions; Represents pixels at odd positions. Indicates bit index.

[0122] Update step:

[0123]

[0124] in, Indicates low-frequency components, Indicates even-numbered pixel positions. This represents the prediction residual that is spatially adjacent to the current processing node. The residual represents the difference between the predicted value and the actual value. Indicates bit index.

[0125] The corresponding low-frequency components of an image, namely the image's contours, smooth background, and subject structure, ultimately constitute the image. Sub-belt.

[0126] Utilizing residuals Even-numbered pixel positions Smoothing correction is performed not only to eliminate jagged edges, but more importantly to maintain the average energy conservation of the signal before and after the transformation.

[0127] In order to utilize the texture masking effect of the human visual system (HVS), this invention calculates... Texture activity of each B×B block in the subband Where B represents the block side length, based on texture activity. Arrange the regions in descending order and prioritize generating the embedded mask from regions with high activity.

[0128] This allows embedded noise to be hidden in areas with complex textures, significantly improving its imperceptibility.

[0129] The specific formula is as follows:

[0130]

[0131] in, Indicates texture activity, B represents the block edge length. Indicates the first Each block, Represents pixel coordinates, This represents the pixel coordinate index in the horizontal direction. This represents the pixel coordinate index in the vertical direction. Represents the horizontal gradient. Represents the vertical gradient. Represents the horizontal gradient operator. Represents the vertical gradient operator. Indicates the horizontal direction. Indicates the vertical direction. This indicates the inherent coefficient at low frequencies. This indicates a block index.

[0132] After the above processing, the present invention obtains a decoupled signal frequency band, namely the first-level low-frequency sub-band. It also obtained guidance on embedded spatial navigation, which is also an embedded mask.

[0133] After completing the adaptive embedding of the secret information, the generated steganographic image is transmitted through the channel.

[0134] To highlight the robustness of this invention and enhance its practical application in real-world environments, this invention must be able to resist various JPEG, noise, geometric scaling, and combined attacks encountered during transmission, thus possessing the physical characteristics to survive in harsh environments. Therefore, this invention needs to establish physical tolerance against channel distortion at this stage.

[0135] Combination Figure 4 As shown, Quantization Index Modulation (QIM) is used to establish a strong binding relationship between data and coefficient statistics by setting a specific quantization step size. This ensures that the embedded data feature points maintain sufficient Euclidean distance in the signal space to resist attacks. The critical value setting of the embedding parameters is used to calculate the block mean for a selected block. .

[0136] According to the bit to be embedded Introducing jitter offset :

[0137]

[0138] in, This indicates the jitter offset. Indicates the quantization step size. This represents the bit to be embedded.

[0139] Corrected block mean for:

[0140]

[0141] in, This represents the corrected block mean; Indicates the quantization step size; Quantization factor representing the quantization step size ; This represents the original value to be quantized; Indicates the quantized output value; This function represents rounding to the nearest integer. Indicates the block average; Indicates the jitter offset; This represents the bit to be embedded.

[0142] Then calculate the offset:

[0143]

[0144] in, Indicates the offset. This represents the corrected block mean. This represents the block mean. The offset is then added to all pixels within the block to complete the embedding.

[0145] This stage introduces a significant quantization isolation band into the low-frequency subband coefficients of the carrier, ensuring that the signal distortion of the dense image is always controlled within the fault tolerance range of the QIM decision when it subsequently undergoes an uncontrollable channel transmission process (i.e., an attack), thus laying the physical foundation for zero-error extraction in subsequent steps.

[0146] Combination Figure 5 As shown, after receiving the Stego image obtained through channel transmission, the computer system no longer faces a perfect steganographic image. Instead, it is a received image carrying channel interference such as quantization noise and interpolation errors. .

[0147] To ensure the complete extraction of secret information, it is necessary to complete the blind detection and high-precision reconstruction of noisy secret information under the condition of no original carrier reference in this stage. This invention achieves the above goal by constructing a dual recovery mechanism of physical layer feature demodulation and logical layer timing verification.

[0148] First, at the signal demodulation level, this invention demodulates the received image. Perform an integer wavelet transform and block statistics that are completely mirrored in the aforementioned process, and use the minimum distance rule of Quantization Index Modulation (QIM) to calculate the feature mean. Make a hard decision, that is, by calculation:

[0149]

[0150] in, Indicates the initial bit stream, This indicates taking the bit index corresponding to the minimum distance. Indicates bit index, Represents the characteristic mean. The quantization factor represents the quantization step size. The residual represents the difference between the predicted value and the actual value. Indicates the quantization step size.

[0151] By obtaining the initial bit stream as described above, preliminary filtering of physical channel noise can be achieved, based on the initial secret bits. To obtain an estimate of the final secret embedding stream. .

[0152] Furthermore, to address the burst block error problem caused by channel attacks (such as JPEG compression), this invention employs the inverse mapping of global interleaving mapping. Perform deinterleaving operation.

[0153]

[0154] in, This represents the error correction coding sequence after deinterleaving. This represents the estimate of the final secret embedding stream. This represents the inverse mapping of the global interleaving mapping.

[0155] The errors in the set are discretized, and the Reed-Solomon algorithm is used to complete the first round of error correction decoding to obtain the intermediate sequence. .

[0156] Based on this, the present invention solves the decoding failure problem caused by the non-commutativity of operations by strictly defined post-processing timing logic, that is, according to the whitening operation's dependence on the index position, it is mandatory to first use the formula:

[0157]

[0158] in, This represents the spatially scrambled sequence after inverse whitening. Indicates an intermediate sequence. This represents the XOR operation. This represents a pseudo-random mask sequence.

[0159] Remove the pseudo-random mask used to conceal statistical characteristics, and then apply the formula:

[0160]

[0161] in, The secret image bits represent the final reconstructed image. Indicates bit index, This represents the spatially scrambled sequence after inverse whitening. This represents the inverse mapping of the global mapping for space scrambling.

[0162] The above formula is used to restore the original spatial topology of the data. This specific timing design utilizes the mathematical irreversibility of linear operations and scrambling mapping, effectively avoiding the risk of index misalignment caused by reversed order, thereby ensuring the system can reconstruct secret images with zero errors under complex attack environments.

[0163] The technical terms, principles, or means related to the technical solutions of the present invention mentioned in the above embodiments, which are not described in detail above, are all well-known technologies or common practices that are known to those skilled in the art.

[0164] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A robust image steganography method based on low-frequency embedding with no background noise, characterized in that, This includes the computer system performing the following steps: Input the original secret image and the carrier image, binarize and serialize the original secret image to obtain the secret image, and generate a one-dimensional binary bit sequence based on the secret image; The statistical fingerprint is eliminated by double randomization of the one-dimensional binary bit sequence, and burst errors in the channel are transformed into random errors by global interleaving to generate the final secret embedding stream. Perform a 5 / 3 integer wavelet transform on the carrier image to decompose the frequency domain into two integer operation steps to obtain a first-level low-frequency sub-band. Calculate the texture activity of the first-level low-frequency sub-band into several blocks divided according to a preset side length, and generate an embedding mask based on the texture activity. Based on quantization index modulation, the quantization step size and jitter offset are set. The average value of the block to be embedded is adjusted according to the bit to be embedded in the final secret embedding stream and the offset is added to complete the embedding and obtain the received image. Integer wavelet transform and block statistics are performed on the received image, and hard decision is made on the feature mean using the minimum distance rule of quantization index modulation to obtain the initial bit stream; Based on the initial bitstream, the secret image is reconstructed and output through deinterleaving, error correction decoding, inverse whitening, and inverse scrambling. The execution steps for generating a one-dimensional binary bit sequence based on the secret image are as follows: Based on the secret image, a robust threshold is used to discretize the image pixels into a one-dimensional binary bit sequence containing only 0 or 1; The dual randomization includes spatial scrambling and statistical whitening; The execution steps for spatial scrambling are as follows: A global random permutation is generated using the first key, and the spatial order of the one-dimensional binary bit sequence is shuffled based on the global random permutation. The steps for statistical whitening are as follows: A pseudo-random mask sequence following a uniform distribution is generated using the second key, and an XOR operation is performed with a spatially scrambled one-dimensional binary bit sequence. The steps for converting burst errors in the channel into random errors through global interleaving include error correction coding and global interleaving processing, as detailed below: The statistically whitened one-dimensional binary bit sequence is mapped to a byte stream, and redundant check bits are added to it using Reed-Solomon encoding and interleaving techniques to generate an encoded sequence; The encoded sequence is scrambled globally using a global interleaving map to generate the final secret embedding stream. The steps for adjusting the mean of the bit to be embedded and adding the offset to complete the embedding are as follows: For the first-level low-frequency sub-band selected by the embedded mask, the blocks are divided according to the custom side length. The average value of all wavelet coefficients in the block is calculated to obtain the block mean of the current block. Based on the bits to be embedded, set the corresponding jitter offset and obtain the corrected block average using the following formula; The formula is: in, This represents the corrected block mean. Indicates the quantization step size. The quantization factor represents the quantization step size. Represents the block average. This indicates the jitter offset. Indicates the bit to be embedded; Subsequently, based on the corrected block mean and jitter offset, the offset is calculated and superimposed on all pixels of the current block to complete the embedding.

2. The error-resistant robust image steganography method based on low-frequency embedding without background noise as described in claim 1, characterized in that, The two integer operation steps include a prediction step and an update step; The execution steps of the prediction step are as follows: By using the average of two adjacent even-numbered pixels, the pixel at the middle odd-numbered position is predicted, thus obtaining the residual that reflects local abrupt changes in the image. The execution steps of the update step are as follows: Based on the residuals that reflect local abrupt changes in the image, even-numbered pixels are smoothed to obtain the low-frequency components of the corresponding image.

3. The error-resistant robust image steganography method based on low-frequency embedding without background noise according to claim 1, characterized in that, The steps for obtaining the initial bit stream are calculated as follows: in, Indicates the initial bit stream, This indicates taking the bit index corresponding to the minimum distance. Indicates bit index, Represents the characteristic mean. The quantization factor represents the quantization step size. The value represents the difference between the predicted value and the actual value, i.e., the residual, and Δ represents the quantization step size.

4. The error-resistant robust image steganography method based on low-frequency embedding without background noise as described in claim 1, characterized in that, The de-interleaving execution steps are as follows: Based on the initial bit stream, an estimate of the final secret embedding stream is obtained, and the errors in the set are discretized through the inverse mapping of the global interleaving map; The execution steps of the error correction decoding are as follows: After deinterleaving, the Reed-Solomon algorithm is used to complete the first round of error correction decoding to obtain the intermediate sequence.

5. The error-resistant robust image steganography method based on low-frequency embedding without background noise according to claim 1, characterized in that, The reverse whitening includes: in, This represents the spatially scrambled sequence after inverse whitening. Indicates an intermediate sequence. This represents the XOR operation. Represents a pseudo-random mask sequence; The reverse scrambling process includes: in, The secret image bits represent the final reconstructed image. Indicates bit index, This represents the spatially scrambled sequence after inverse whitening. This represents the inverse mapping of the global mapping for space scrambling.

6. A robust image steganography system based on low-frequency embedding with no background noise for implementing the method of any one of claims 1-5, characterized in that, include, The preprocessing and hybrid coding stream module is used to eliminate statistical fingerprints through double randomization and to transform burst errors in the channel into random errors through global interleaving; A reversible embedding domain construction module is used to perform a 5 / 3 integer wavelet transform on the carrier image to obtain the first-level low-frequency subband, and generate an embedding mask based on the texture activity. The quantization embedding module is used to set the quantization step size and jitter offset based on the quantization index modulation, and to complete the embedding by adjusting the block mean and superimposing the offset according to the bit to be embedded. The bar-blind extraction and timing verification logic module is used to perform integer wavelet transform and block statistics on the received image, obtain the initial bit stream through the minimum distance method of quantization index modulation, and perform deinterleaving, error correction decoding, inverse whitening, and inverse scrambling to reconstruct and output the secret image.

7. An electronic device, comprising: The processor, communication interface, memory, and communication bus are connected, with the processor, communication interface, and memory communicating with each other via the communication bus. The feature is that the memory stores a computer program, which, when executed by the processor, causes the processor to perform the error-resistant robust image steganography method based on low-frequency embedding without background noise as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the error-resistant robust image steganography method based on low-frequency embedding without background noise as described in any one of claims 1-5.

Images