Multi-domain simultaneous embedding video steganography method and system based on motion vector index and PU segmentation

Abstract

The invention provides a multi-domain simultaneous embedding video steganography method and system based on a motion vector index and PU segmentation, and belongs to the technical field of video steganography. On the basis of a traditional motion vector domain embedding method, utilization of a Merge mode PU is added. According to the method, the secret information is embedded by modifying a predicted motion vector flag bit of the PU in the AMVP mode and a segmentation method of the PU in the Merge mode, the mutual influence condition of two embedding domains is considered, the embedding method is reasonably limited, disturbance brought by the embedded secret information is analyzed, a joint distortion function is designed, and the overall distortion of embedded information is minimized through STC coding. Experiments show that in combination with embedding of the Merge mode PU, the steganography capacity is effectively improved, the embedding capacity can be kept stable in videos with various contents, the quality of the embedded video can be flexibly adjusted through a control factor of a distortion function, and in combination with STC coding, disturbance of the algorithm to the quality of the video is small. For a targeted steganography analysis method, good resistance can be kept, and steganography safety is reliable.

Description

Technical Field

[0001] This invention belongs to the field of video steganography technology, specifically relating to a method and system for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation. Background Technology

[0002] With the development of information technology, modern steganography tends to hide secret information in ubiquitous digital media, including text, images, and videos. Video, due to its substantial redundancy and resistance to detection, has become an ideal medium for information steganography. To save storage space and improve transmission efficiency over networks, most videos we watch in daily life are compressed using video coding technology. Since the ciphertext hidden in the original video is easily lost after compression, video steganography is closely related to video coding technology. H.265 / HEVC has higher compression efficiency than H.264 / AVC and has been widely used; its market share is expected to exceed that of H.264 / AVC. Therefore, it is necessary to study video steganography technology based on the HEVC standard.

[0003] Single-domain video steganography algorithms possess different modification characteristics, such as transform coefficients, pre-interpretation modes, and inter-prediction modes. In the transform coefficient domain, Chang et al. modified the Discrete Cosine Transform (DCT) coefficients according to five different PU partitioning modes to eliminate intra-frame distortion drift. Liu et al. encoded the message into n sub-secrets through (t,n) threshold secret sharing and embedded the secrets into multiple coefficients of the selected 4×4 Luminance Discrete Sine Transform (DST) block to improve robustness. Yang et al. designed an adaptive cost function by considering factors such as intra-block distortion, inter-frame distortion accumulation, and inter-block distortion. These algorithms are based on DCT / DST modulation secret information. However, in actual compression, most transform domain residual coefficients are zero, which limits the number of carriers.

[0004] For pre-prediction modes, Xu et al. established a mapping between the overlay sequence and 33 internal prediction modes to embed secret information. Wang et al. established a three-layer isolation channel based on the properties of internal prediction mode (IPM) coding to avoid interaction between adjacent IPMs. Yang et al. designed a method to embed multi-bit secret information within a single coding unit (CU) while maintaining the four-part tree structure. These algorithms are suitable for Intra(I) frames, but the number of I-frames in a video sequence is relatively small.

[0005] Algorithms employing inter-frame prediction modes can be further divided into two types: Prediction Unit Partitioning (PUPM) and Motion Vector (MV) modes. For the PUPM domain, Yang et al. analyzed the block structure and distribution of PUs in HEVC and embedded secret information by defining a PUPM mapping. Zhang et al. proposed EMD encoding to increase embedding capacity. Liu et al. used diamond encoding to allow CUs to carry more information. PUPM-based algorithms need to consider the impact on invisibility and encoding cost when modifying the segmentation mode. For the MV domain, Guo et al. established a mapping strategy between MVs and binary bitstreams based on motion trends. Liu et al. modified the MVP index to maintain consistent visual quality before and after the change. Li et al. improved the algorithm's security through MVP index and motion vector difference (MVD). MV domain steganography is based on the AMVP mode PU, which limits the embedding capacity.

[0006] The main metrics for evaluating steganography algorithms are security and embedding capacity. Security refers to the concealment of communication and its resistance to steganalysis. For video, this means preserving the original coding structure as much as possible while maintaining high video quality and resistance to steganalysis. Embedding capacity focuses on whether the algorithm can provide enough space for secret information. However, high security usually means less modification, while large embedding capacity usually requires more modification, thus striking a balance between the two.

[0007] A review of the aforementioned literature reveals that while single-domain algorithms have their advantages, they also possess inherent limitations. On one hand, when using a single domain, other domains remain idle. On the other hand, single-domain methods are susceptible to targeted steganalysis. To achieve a better balance between security and embedding capability, researchers are now exploring multi-domain algorithms, aiming to improve performance by combining the advantages of single-domain methods. To our knowledge, Zhai et al. proposed the first H.264-based multi-domain embedding video steganography algorithm. They used macroblock segmentation patterns and motion vector embedding information, thereby increasing the steganography capacity. Xing et al. proposed an H.265-based multi-domain method, using TU partitioning and intra-frame prediction patterns with i-frames, which can increase the embedding capacity without compromising security. Generally, videos contain more p-frames than i-frames. Therefore, working with p-frames, MV-domain-based algorithms can maintain relatively good video quality because perturbations caused by MV modification can be offset by subsequent prediction residual transformation and quantization processing. However, the carrier of secret information can only be the pu in AMVP mode. To overcome this limitation, a multi-domain simultaneous embedding video steganography method based on motion vector indexing and PU segmentation is proposed. Summary of the Invention

[0008] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a method and system for simultaneous embedding of video steganography in multiple domains based on motion vector indexing and PU segmentation.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation, comprising the following steps: Inter-frame prediction is performed on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and the carrier sequence is constructed. Based on the constructed carrier sequence, the distortion cost of the carrier elements is calculated by combining the joint distortion function of the segmentation mode domain and the motion vector domain. The carrier elements are modified by the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret-carrying sequence. The video is re-encoded based on the encrypted sequence and mapping rules, and the encrypted video is output.

[0010] In the step of performing inter-frame prediction on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and constructing the carrier sequence, the specific method for constructing the carrier sequence is as follows: Perform inter-frame prediction on the video sequence to obtain segmentation mode domain and motion vector domain information for all P frames; Process each P-frame sequentially according to the frame arrangement in the video, and obtain CU information in order within each P-frame; Determine whether the acquired CU information contains elements that can be used as carriers, generate carrier elements according to the mapping rules, and write them into the carrier sequence in sequence.

[0011] The mapping rule for the segmentation pattern domain is as follows: the Merge pattern CU of N×2N is mapped to 1, and the Merge pattern CU of 2N×N is mapped to 0. The mapping rule for the motion vector domain is as follows: for a PU in AMVP mode, if the index of the predicted motion vector is 1, it is mapped to 1; if the index of the predicted motion vector is 0, it is mapped to 0.

[0012] The Merge mode CU, which acts as the carrier, is divided into 8×8 segments; the AMVP mode PU, which acts as the carrier, does not include the Merge mode PU with N×2N and 2N×N segmentation patterns of 8×8 size in its motion vector candidate list.

[0013] To avoid perturbation of the segmentation mode domain by modifying motion vectors, the motion vector residual MVD of the selected AMVP mode PU acting as the carrier is compensated, and the specific formula is expressed as follows:

[0014]

[0015] in, This represents the original, unmodified motion vector. This represents the predicted motion vector when it is in its original, unmodified state. Represents the residual of the original motion vector; This represents the modified motion vector. This represents the modified predicted motion vector.

[0016] In the step of calculating the distortion cost of carrier elements based on the constructed carrier sequence, combining the joint distortion function of the segmentation mode domain and the motion vector domain, and modifying the carrier elements using the STC encoding method to embed the secret message into the carrier sequence to obtain the secret-carrying sequence, the distortion function of the segmentation mode domain is as follows:

[0017] Among them, D P This represents the distortion cost of segmenting the pattern domain. This represents the difference between the current block after steganography and its original pixel value. This represents the original difference between the current block and the original pixel. This represents the number of bits required for steganography. This represents the number of raw bits required for encoding; β is half the number of frames contained in the GOP, and n is the position of the current frame in the GOP; This is a Lagrange control factor used to control the quality of the encoded video and the encoding cost.

[0018] The distortion function in the motion vector domain is as follows:

[0019] in, This represents the distortion cost in the motion vector domain. and These are the modified number of encoded bits and the original number of encoded bits, respectively. This is a variable parameter used to control the cost of the number of encoded bits.

[0020] A joint distortion function is constructed based on the distortion function in the segmentation mode domain and the distortion function in the motion vector domain. The specific formula is as follows:

[0021] in, To ensure the reasonable range of distortion values ​​for the final combined distortion cost, the exp function is designed to guarantee the distortion value.

[0022] Based on the vector sequence and the joint distortion function, the distortion cost of each vector element is calculated, and a vector-dense sequence is generated according to STC embedding, as shown in the following formula:

[0023] Where C is the vector sequence, For confidential information, C (m) It is a coset of m, T takes the minimum distortion, and S represents the cipher sequence.

[0024] Secondly, the present invention provides a multi-domain simultaneous embedding video steganography system based on motion vector indexing and PU segmentation, comprising: The carrier sequence construction module is used to perform inter-frame prediction on the video sequence, obtain the segmentation mode domain and motion vector domain information of P frames, and construct the carrier sequence. The secret message sequence generation module is used to calculate the distortion cost of the carrier elements based on the constructed carrier sequence, combined with the joint distortion function of the segmentation mode domain and motion vector domain, and to modify the carrier elements by using the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret message sequence. The video encoding module is used to re-encode the video according to the encrypted sequence and mapping rules, and output the encrypted video.

[0025] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a multi-domain simultaneous embedding video steganography method based on motion vector indexing and PU segmentation, comprising the following steps: performing inter-frame prediction on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and constructing a carrier sequence; based on the constructed carrier sequence, calculating the distortion cost of carrier elements by combining the joint distortion function of the segmentation mode domain and motion vector domain, modifying the carrier elements using the STC encoding method to embed the secret information into the carrier sequence, obtaining a steganography sequence; re-encoding the video according to the steganography sequence and mapping rules, and outputting the steganography video. By utilizing the segmentation mode domain and motion vector domain for information embedding, the limitations of single-domain embedding in traditional steganography methods are overcome. The correlation between the two domains is dynamically adjusted through the joint distortion function to adjust the embedding strategy. Combined with STC encoding, steganography traces are effectively dispersed, significantly reducing the statistical anomalies of the steganography video, and its resistance to steganography analysis is superior to single-domain embedding methods.

[0026] Furthermore, by combining the embedding of Merge mode PU, the steganography capacity is effectively improved, and the embedding capacity can be kept stable in videos with various content. Attached Figure Description

[0027] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a comparison chart of the results in Embodiment 2 of the present invention. Detailed Implementation

[0028] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0029] Example 1 like Figure 1 As shown, a multi-domain simultaneous embedding video steganography method based on motion vector indexing and PU segmentation includes the following steps: S1: Perform inter-frame prediction on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and construct the carrier sequence; S2: Based on the constructed carrier sequence, the distortion cost of the carrier elements is calculated by combining the joint distortion function of the segmentation mode domain and the motion vector domain. The carrier elements are modified by the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret message sequence. S3 re-encodes the video according to the encrypted sequence and mapping rules, and outputs the encrypted video.

[0030] Specifically, in S1, the video sequence is pre-compressed once, that is, the complete inter-frame prediction is performed to obtain the segmentation mode domain and motion vector domain information of all P frames. The P frames are processed sequentially according to the order of the frames in the video. In each P frame, the CU (Coding Unit) information is obtained in sequence, and it is determined whether the CU contains elements that can be used as carriers. The carrier elements are generated according to the mapping rules and written into the carrier sequence in sequence.

[0031] The vector sequence consists of vector elements c1 to c L composition, .

[0032] The mapping rule for the segmentation mode domain of the carrier element is as follows: the Merge mode CU, which is divided into N×2N, is mapped to 1, and the Merge mode CU, which is divided into 2N×N, is mapped to 0.

[0033] The mapping rule for the motion vector domain of the carrier element is as follows: the CU is further divided into one or more PUs (Prediction Units). For the PU in the AMVP (Advanced Motion Vector Prediction) mode, if the index of the predicted motion vector is 1, it is mapped to 1; if the index of the predicted motion vector is 0, it is mapped to 0.

[0034] Specifically, the Merge pattern CU, which can serve as the carrier, should be divided into 8×8 blocks, which is equivalent to a block depth of 3 in HEVC (High Efficiency Video Coding). This is because, on the one hand, modifying the segmentation pattern of a larger CU would significantly disrupt the original HEVC coding structure; on the other hand, modifying the segmentation pattern of a larger block could easily disturb the motion vector information of the PU blocks in the surrounding AMVP pattern, thus affecting the simultaneous embedding of multiple domains and causing inaccurate embedding and extraction.

[0035] The motion vector candidate list of the AMVP pattern PU that can serve as a carrier cannot include the Merge pattern PU with 8×8 size N×2N and 2N×N segmentation patterns. This is to avoid the modification of the segmentation pattern domain embedding affecting the motion vector domain embedding.

[0036] Furthermore, during the encoding process, the encoder always chooses the compression method with the lowest rate-distortion cost. Therefore, the predicted motion vectors of the AMVP mode PU, which acts as the carrier element, should satisfy the principle of local optima. In the candidate list, there are two candidate predicted motion vectors. The encoder selects the one with the lower encoding cost as the predicted motion vector for that block. The predicted motion vector index identifies these two predicted motion vectors, as shown in the following formula:

[0037] Where R represents the number of bits required for encoding, MVP mod This represents the modified predicted motion vector encoding information, MVP. ori This represents the original predicted motion vector encoding information. The number of bits required to encode the modified predicted motion vector cannot exceed the number of bits required for the original encoding, so as to keep its encoding cost low, thereby maintaining local optima, avoiding providing evidence for steganalysis, and ensuring the security of the steganalysis algorithm.

[0038] To avoid disturbing the segmentation mode domain by modifying the motion vectors, it is necessary to compensate for the motion vector residual MVD of the selected AMVP mode PU that serves as the carrier, so as to keep the motion vectors unchanged. The formula is as follows:

[0039]

[0040] in, This represents the original, unmodified motion vector. This represents the predicted motion vector when it is in its original, unmodified state. This represents the residual of the original motion vector, which is the difference between the original motion vector and the predicted motion vector. This represents the modified motion vector. This represents the modified predicted motion vector. Modifying the predicted motion vector index will cause a change in the MVP, thereby altering the MV value and affecting the MV values ​​and segmentation patterns of surrounding blocks. Therefore, after modifying the predicted motion vector index, the MVD should be modified accordingly to keep the MV of the current block unchanged and avoid affecting the steganography and information extraction of the segmentation pattern domain.

[0041] Specifically, in S2, based on the constructed carrier sequence, the distortion cost of the carrier elements is calculated using a distortion function. The carrier elements are then modified using the STC encoding method to embed the secret message into the carrier sequence, resulting in a secret-carrying sequence. The specific method is as follows: In the motion vector domain, after modifying the predicted motion vector index, compensation is applied to the motion vector difference, thus keeping the motion vector unchanged. At this point, the pixel difference remains unchanged, and the distortion function only needs to consider the encoding cost, as shown in the formula:

[0042] in, This represents the distortion cost in the motion vector domain. and These are the modified number of encoded bits and the original number of encoded bits, respectively. This is a variable parameter used to control the cost of the number of encoded bits. When the modification of the predicted motion vector index satisfies local optima, The value is close to 0; if the modification of the predicted motion vector index does not satisfy local optima, The value is based on The settings are amplified, and STC will prioritize other modification paths.

[0043] In the segmentation mode domain, modifications to the segmentation mode affect pixel values ​​and the number of encoded bits, as shown in the formula:

[0044] Among them, D P This represents the distortion cost of segmenting the pattern domain. This represents the difference between the current block after steganography and its original pixel value. This represents the original difference between the current block and the original pixel, and is generally measured using mean square error or absolute error. This represents the number of bits required for steganography. Indicates the number of raw bits required for encoding; Half the number of frames contained in a GOP. The position of the current frame in the GOP. These two parameters make it possible for the distortion to be relatively large when modifying earlier frames. STC will prioritize modifying later frames to avoid the accumulation of distortion between frames and cause errors. This is a Lagrange control factor used to control the quality of the encoded video and the encoding cost.

[0045] The joint distortion function over the two domains is given by the following formula:

[0046] in, To determine the final joint distortion cost, the exp function ensures a reasonable range for the distortion value. When modifying the predicted motion vector index satisfies local optima, the distortion cost approaches 0, and STC will prioritize this approach. If local optima are not satisfied, STC will prioritize modifying the segmented mode domain.

[0047] Based on the carrier sequence and the joint distortion function, the distortion cost of each carrier element is calculated, and then the carrier-encrypted sequence is generated according to STC embedding, as shown in the following formula:

[0048] Where C is the carrier sequence, m is the secret information, and C (m) It is a coset of m. T means taking the minimum distortion, and S represents the cipher sequence.

[0049] The STC embedding rules are as follows: When the current PU is in Merge mode and its size is 8×8, the partitioning mode is modified, allowing 1 bit of secret information to be embedded in every two consecutive PUs within the same CU. When the current PU is in AMVP mode and its candidate list does not contain any 8×8-sized Merge-mode PUs with 2N×N or N×2N partitioning modes, the MVP index is modified, allowing each PU to embed 1 bit of secret information. Within a CU, the N×2N partitioning method of two consecutive Merge-mode PUs is mapped to 1, and the 2N×N partitioning method of two consecutive Merge-mode PUs is mapped to 0. For AMVP mode PUs, MVP index 1 is mapped to 1, and 0 is mapped to 0.

[0050] Specifically, in S3, after obtaining the encrypted sequence, according to the above mapping rules and the information shown in the encrypted sequence, inter-frame prediction is performed again to complete the video compression encoding and output the encrypted video.

[0051] Example 2 In this embodiment, the method is implemented on the HEVC reference software HM16.26. The experimental data consists of 10 standard YUV test video sequences, including 3 different resolutions, as shown in Tables 1 and 2. Each video is encoded for 50 frames, with the first 30 frames used for information hiding. The frame rate is 30fps, and the GOP is configured as IPPP. The distortion function parameter α is set to 10.

[0052] Table 1. Video sequences used in the experiment

[0053] Table 2 PSNR and SSIM performance

[0054] Based on the results obtained from the above experiments, as follows: Figure 2As shown, Figure 2 (a) is a frame from the original cover video. Figure 2 (b) A frame of the cover video after steganography using this method. By observation, it is difficult to distinguish the modified video from the original video with the naked eye. The steganographic video does not exhibit blurring, loss of detail, or visible pixelation, jagged edges, or artifacts. It achieves the visual quality of the original video and demonstrates excellent visual invisibility.

[0055] Example 3 A multi-domain simultaneous embedding video steganography system based on motion vector indexing and PU segmentation includes: The carrier sequence construction module is used to perform inter-frame prediction on the video sequence, obtain the segmentation mode domain and motion vector domain information of P frames, and construct the carrier sequence. The secret message sequence generation module is used to calculate the distortion cost of the carrier elements based on the constructed carrier sequence, combined with the joint distortion function of the segmentation mode domain and motion vector domain, and to modify the carrier elements by using the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret message sequence. The video encoding module is used to re-encode the video according to the encrypted sequence and mapping rules, and output the encrypted video.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation, characterized in that, Includes the following steps: Inter-frame prediction is performed on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and the carrier sequence is constructed. Based on the constructed carrier sequence, the distortion cost of the carrier elements is calculated by combining the joint distortion function of the segmentation mode domain and the motion vector domain. The carrier elements are modified by the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret-carrying sequence. The video is re-encoded based on the encrypted sequence and mapping rules, and the encrypted video is output.

2. The method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation according to claim 1, characterized in that, In the step of performing inter-frame prediction on the video sequence to obtain the segmentation mode domain and motion vector domain information of P frames, and constructing the carrier sequence, the specific method for constructing the carrier sequence is as follows: Perform inter-frame prediction on the video sequence to obtain segmentation mode domain and motion vector domain information for all P frames; Process each P-frame sequentially according to the frame arrangement in the video, and obtain CU information in order within each P-frame; Determine whether the acquired CU information contains elements that can be used as carriers, generate carrier elements according to the mapping rules, and write them into the carrier sequence in sequence.

3. The method for simultaneous embedding of video steganography in multiple domains based on motion vector indexing and PU segmentation according to claim 2, characterized in that, The mapping rule for the segmentation pattern domain is as follows: the Merge pattern CU of N×2N is mapped to 1, and the Merge pattern CU of 2N×N is mapped to 0. The mapping rule for the motion vector domain is as follows: for a PU in AMVP mode, if the index of the predicted motion vector is 1, it is mapped to 1; if the index of the predicted motion vector is 0, it is mapped to 0.

4. The method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation according to claim 3, characterized in that, The Merge mode CU, which acts as the carrier, is divided into 8×8 segments; the AMVP mode PU, which acts as the carrier, does not include the Merge mode PU with N×2N and 2N×N segmentation patterns of 8×8 size in its motion vector candidate list.

5. The method for simultaneous embedding of video steganography in multiple domains based on motion vector indexing and PU segmentation according to claim 4, characterized in that, To avoid perturbation of the segmentation mode domain by modifying motion vectors, the motion vector residual MVD of the selected AMVP mode PU acting as the carrier is compensated, and the specific formula is expressed as follows: in, This represents the original, unmodified motion vector. This represents the predicted motion vector when it is in its original, unmodified state. Represents the residual of the original motion vector; This represents the modified motion vector. This represents the modified predicted motion vector.

6. The method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation according to claim 1, characterized in that, In the step of calculating the distortion cost of carrier elements based on the constructed carrier sequence, combining the joint distortion function of the segmentation mode domain and the motion vector domain, and modifying the carrier elements using the STC encoding method to embed the secret message into the carrier sequence to obtain the secret-carrying sequence, the distortion function of the segmentation mode domain is as follows: Among them, D P This represents the distortion cost of segmenting the pattern domain. This represents the difference between the current block after steganography and its original pixel value. This represents the original difference between the current block and the original pixel. This represents the number of bits required for steganography. This represents the number of raw bits required for encoding; β is half the number of frames contained in the GOP, and n is the position of the current frame in the GOP; This is a Lagrange control factor used to control the quality of the encoded video and the encoding cost.

7. The method for simultaneous embedding of video steganography in multiple domains based on motion vector indexing and PU segmentation according to claim 6, characterized in that, The distortion function in the motion vector domain is as follows: in, This represents the distortion cost in the motion vector domain. and These are the modified number of encoded bits and the original number of encoded bits, respectively. This is a variable parameter used to control the cost of the number of encoded bits.

8. The method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation according to claim 6, characterized in that, A joint distortion function is constructed based on the distortion function in the segmentation mode domain and the distortion function in the motion vector domain. The specific formula is as follows: in, To ensure the reasonable range of distortion values ​​for the final combined distortion cost, the exp function is designed to guarantee the distortion value.

9. A method for simultaneous embedding of multiple domains in video steganography based on motion vector indexing and PU segmentation according to claim 8, characterized in that, Based on the vector sequence and the joint distortion function, the distortion cost of each vector element is calculated, and a vector-dense sequence is generated according to STC embedding, as shown in the following formula: Where C is the vector sequence, For confidential information, C (m) It is a coset of m, T takes the minimum distortion, and S represents the cipher sequence.

10. A multi-domain simultaneous embedding video steganography system based on motion vector indexing and PU segmentation as described in any one of claims 1 to 9, characterized in that, include: The carrier sequence construction module is used to perform inter-frame prediction on the video sequence, obtain the segmentation mode domain and motion vector domain information of P frames, and construct the carrier sequence. The secret message sequence generation module is used to calculate the distortion cost of the carrier elements based on the constructed carrier sequence, combined with the joint distortion function of the segmentation mode domain and motion vector domain, and to modify the carrier elements by using the STC encoding method to embed the secret message into the carrier sequence, thus obtaining the secret message sequence. The video encoding module is used to re-encode the video according to the encrypted sequence and mapping rules, and output the encrypted video.

Images