A large-capacity information hiding method based on short video communication media

Abstract

In view of the shortcomings of existing video information hiding methods, the present invention provides a large-capacity information hiding method based on short video information dissemination media. The method makes full use of the characteristics of short video dissemination media with rich scene switching, combines the different tolerance of human eyes to scene switching images and general images that can be modified, uses a scene detection algorithm to determine the transition frame interval, and performs adaptive hierarchical embedding based on scene switching for I4x4 luminance block QDCT coefficients and motion vector difference MVD in different types of frames. At the same time, in order to ensure video quality, the I4x4 luminance block QDCT coefficient is combined with the technology of preventing intra-frame drift distortion, thereby improving the embedding capacity while ensuring video quality.

Description

Technical Field

[0001] The present invention belongs to the technical field of information security and relates to a method for hiding large-capacity information in a short video communication medium based on the H.264 / AVC standard. Background Art

[0002] With the rapid development of network technology and the widespread application of multimedia technology, the application and dissemination of multimedia information has become an indispensable part of people's lives, enriching people's daily leisure life and entertainment mode. Multimedia information has the characteristics of distortion-free copying, dissemination and easy modification. While providing convenience for people's life and entertainment, it also hides many security risks. Information hiding technology can effectively curb the illegal use, tampering and forgery of digital information, thereby ensuring the security of digital information content.

[0003] As one of the mainstream transmission media, video information hiding can be divided into spatial domain video steganography and compression domain video steganography. Spatial domain video steganography directly modifies the color space RGB or YUV of the video before encoding to embed secret information. However, video compression encoding is often required during storage and transmission, so the embedded secret information is easily lost in the subsequent compression encoding process. Compression domain video steganography is the mainstream of video steganography today. Secret information is embedded in the compression encoding process, mainly using prediction mode, motion vector (MV) and QDCT coefficient (Quantized Discrete Cosine Transform, non-zero DCT coefficient) as embedding carriers. These embedding carriers provide more embedding methods and embedding space options for steganography. Among them, modifying QDCT coefficients to embed information has been widely used, because discrete cosine transform is an indispensable step in the video compression encoding process; secondly, the method based on QDCT coefficients has little effect on video compression quality (including reconstruction of visual quality and compression efficiency).

[0004] At present, some relevant patent applications for compressed domain video steganography using QDCT coefficients as embedding carriers are as follows:

[0005] (1) Chinese patent application No. 201911260045.1 discloses “an embedding method and extraction method for hiding HEVC video information”. The embedding method obtains a video as a carrier and receives secret information to be embedded by a user; encrypts the secret information using a single-letter substitution cipher algorithm to obtain encrypted secret information; decodes the video and selects a 4x4 luminance DST coefficient block whose I-frame intra-frame residual absolute value is greater than a threshold T as a candidate embedding block; if the intra-frame prediction mode of the neighboring blocks of the candidate embedding block meets the preset condition of preventing intra-frame distortion drift, the encrypted secret information is embedded in the candidate embedding block using a matrix coding embedding method.

[0006] (2) Chinese patent application No. 201810699822.1 discloses “A video information hiding method based on inter-block decoupling”. The method first obtains the channel parameters of the carrier video, wherein the channel parameters include whether each block is a non-reference block, the non-zero DCT coefficient value in each non-reference block and the embedding cost, and then performs steganographic encoding on the non-zero DCT coefficient value of the non-reference block. Then, according to the encoding result, the non-zero DCT coefficients in all non-reference blocks are replaced and modified by the lowest bit to obtain a steganographic video file.

[0007] (3) Chinese patent application No. 201710501846.7 discloses “a video steganography method based on H.265 standard distortion-free drift”, which performs preprocessing such as binarization and information termination bit on secret information, determines the embedding position according to the prediction mode during video encoding, and embeds the preprocessed secret information.

[0008] (4) Chinese patent application No. 201610554530.X discloses “an H.264 video information hiding method tightly coupled with compression coding process”. The core idea is to achieve high-security or high-capacity information hiding without drift distortion based on full coding by means of a method tightly coupled with compression coding.

[0009] (5) Chinese patent application No. 201510015101.0 discloses “A method for embedding and extracting video watermarks in H.264 / AVC compressed domain”, which divides the 4x4 luminance blocks in the I frame into texture blocks and non-texture blocks according to the coefficients of the 4x4 luminance blocks after integer discrete cosine transform, and then embeds secret information according to the local energy difference of each macroblock modulated by binary watermark.

[0010] The above-mentioned video steganography method based on QDCT coefficients mainly screens and embeds I4x4 luminance blocks in I frames. Due to the limited number of I frames, the embedding capacity is limited, and P frames and B frames are not effectively utilized. In addition, the existing video steganography methods do not take into account the characteristics of short videos, a new type of communication medium, with rich scene switching. Due to the large pixel changes in adjacent frames during scene switching, it is difficult for a considerable number of macroblocks in P frames and B frames to find corresponding matching blocks when performing motion estimation during inter-frame prediction. At this time, intra-frame prediction is performed, and many macroblocks are encoded as I macroblocks. At the same time, since the scene switching image is more chaotic than the general image, the human eye has different tolerance for modifications to the scene switching image and the general image. Therefore, the transition effect frame interval can be defined in combination with the scene switching, and the I4x4 luminance block QDCT coefficients and motion vector difference MVD in different types of frames are embedded in an adaptive level based on scene switching, thereby improving the embedding capacity. Summary of the invention

[0011] The purpose of the present invention is to address the shortcomings of existing video information hiding methods and provide a large-capacity information hiding method based on short video information dissemination media. The method makes full use of the characteristics of short video dissemination media with rich scene switching, combines the different tolerance of the human eye to the modifiable scene switching images and general images, uses a scene detection algorithm to determine the transition frame interval, and performs adaptive hierarchical embedding based on scene switching for the I4x4 luminance block QDCT coefficients and motion vector difference MVD in different types of frames. At the same time, in order to ensure video quality, the I4x4 luminance block QDCT coefficients are combined with the technology to prevent intra-frame drift distortion, thereby improving the embedding capacity while ensuring video quality.

[0012] The present invention includes information embedding and information extraction;

[0013] Information embedding specifically includes the following steps:

[0014] Step 1, obtaining the H.264 original video as a carrier and the information bit stream to be embedded;

[0015] Step 2: For the H.264 original video, the average values ​​C1, C2, and C3 of the single pixel value changes in each dimension of the two previous and next frames in the HSV space are calculated, and then the average values ​​of C1, C2, and C3 are taken to obtain the scene change value C. The scene change value C is compared with the set transition threshold V. If the change value C is greater than the transition threshold V, it is determined to be a transition frame, and the continuous transition frames among all transition frames are defined as the transition frame interval.

[0016] Step 3: decode the H.264 original video, obtain the QDCT coefficients of the I4x4 luminance block and its neighboring block luminance prediction mode, motion vector difference MVD, and define the screening conditions for each frame according to the scene change value C and transition threshold V of each frame calculated in step 2, and screen the candidate embedded blocks and candidate motion vector difference MVD in a hierarchical manner.

[0017] The hierarchical screening conditions depend on the scene change value C and the transition threshold V of each frame. A large scene change value indicates that the current frame is not strongly correlated with the previous frame, and the current frame is significantly modified by reducing the screening conditions to increase the number of candidate embedding blocks and candidate motion vector differences MVD.

[0018] The hierarchical screening conditions for candidate embedding blocks are as follows:

[0019] When the current frame is an I frame, an I4x4 luminance block that meets the following conditions is a candidate embedding block: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T1, T1>0, and the average sum of the absolute values ​​of the QDCT coefficients is greater than or equal to the threshold T1', T1'>0.

[0020] Among them, the values of the thresholds T1 and T1' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T1 = 1 and T1' = 1; otherwise, T1 = 2 and T1' = 2.

[0021] When the current frame is a P frame or a B frame, an I4x4 luminance block is an alternative embedding block if it satisfies the following conditions: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T2, T2 > 0, and the average value of the absolute values of each term of the QDCT coefficient is greater than or equal to the threshold T2', T2 > 0.

[0022] Among them, the values of the thresholds T2 and T2' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T2 = 0 and T2' = 0.5; if it is not in the transition frame interval, when the scene change value C satisfies C >= V or 0 <= C < V / 2, T2 = 2 and T2' = 2; otherwise, T2 = 1 and T2' = 1.

[0023] Among them, the DC coefficient is the first coefficient in the QDCT coefficient matrix.

[0024] The screening condition for the motion vector difference MVD is that the horizontal component MVDx or the vertical component MVDy of the motion vector difference MVD satisfies the condition MVDx ≥ 2 n ||MVDy ≥ 2 n ; if the current frame is not in the transition frame interval and the scene change value C satisfies 0 <= C < V / 2 or C >= V, then n = 3; if the scene change value C satisfies V / 2 <= C < V, then n = 2; if the current frame is in the transition frame interval, then n = 1.

[0025] Step 4. Embed the QDCT coefficients and the motion vector difference MVD of the alternative embedding block:

[0026] During the process of embedding the QDCT coefficients, by analyzing the reference pixel types of the neighboring blocks of the alternative embedding block, it is determined which pixels of the alternative embedding block are referenced by subsequent predictive coding, and then the embedding type is determined. According to the embedding type, a specific QDCT coupling coefficient is obtained, and the least significant bit of the first coefficient in the specific QDCT coupling coefficient is embedded with information. The non-first coefficients in the coupling coefficient are modified accordingly according to the embedding type, so as to achieve the purpose of preventing I-frame intra-drift distortion.

[0027] Among them, the specific QDCT coupling coefficient is a coefficient group composed of QDCT coefficients at several specific positions in the QDCT coefficient matrix, and is called the specific QDCT coupling coefficient.

[0028] In the process of embedding the motion vector difference MVD, only the least significant bit LSB of the horizontal component MVDx or the vertical component MVDy that meets the MVD screening condition in step 4 is embedded with bit information, and 1 bit can be embedded at this time; if both the horizontal component MVDx and the vertical component MVDy meet the embedding conditions, they are both embedded, and 2 bits can be embedded at this time.

[0029] Step 5: Re-encode each frame of the embedded information using H.264 to obtain the carrier video of the embedded information.

[0030] Information extraction specifically includes the following steps:

[0031] Step 1: Get the H.264 carrier video with embedded information;

[0032] Step ②, for the H.264 original video, the average values ​​C1, C2, and C3 of the single pixel value changes in each dimension of the two frames before and after the HSV space are calculated, and then the average values ​​of C1, C2, and C3 are taken to obtain the scene change value C. Compare with the set transition threshold V, if the change value C is greater than the transition threshold V, it is determined to be a transition frame, and the continuous transition frames among all transition frames are defined as the transition frame interval.

[0033] Step ③, decode the H.264 carrier video, obtain the QDCT coefficients of the I4x4 luminance block and its neighboring block luminance prediction mode, motion vector difference MVD, and define different screening conditions according to the scene change value C and threshold V of each frame calculated in step ②, and screen the candidate extraction blocks and candidate motion vector difference MVD in a hierarchical manner.

[0034] The hierarchical screening conditions are consistent with the screening conditions in step 3 of the information embedding part.

[0035] Step 4, extracting the information bits embedded in the QDCT coefficients of the candidate extraction blocks and the candidate motion vector difference MVD;

[0036] The extraction process for the candidate extraction block needs to determine the extraction type of the candidate extraction block, and the determination of the extraction type is consistent with the determination of the embedding type in step 4 of the specific QDCT coupling coefficient and information embedding part of each extraction type and the specific QDCT coupling coefficient of each embedding type. According to the extraction type, the least significant bit of the first coefficient in the specific QDCT coupling coefficient can be extracted to extract the bit information.

[0037] The extraction process of the candidate motion vector difference MVD only extracts the least significant bit of the horizontal component MVDx or the vertical component MVDy of the motion vector difference that meets the screening condition in step ③. If both the horizontal component MVDx and the vertical component MVDy meet the screening condition, 2 bits of information can be extracted.

[0038] Another object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the above method.

[0039] The present invention aims at the deficiencies of the existing video steganography method and proposes a video steganography method using short video communication media. Since short videos often have rich scene switching, and the pixel changes of adjacent frames in the scene switching process are large, part of the macroblocks in the P frame and the B frame are encoded as I macroblocks. The video steganography method for embedding QDCT coefficients often does not utilize this part of the I macroblock. In addition, the scene switching detection algorithm is combined to determine the transition frame sequence number and define the transition frame interval. Combined with the different tolerances of the human eye to the scene switching image and the general image, the human eye generally has a higher tolerance to the scene switching image. The QDCT coefficients and motion vector difference MVD of the I4x4 luminance block in different types of frames are embedded in an adaptive level based on scene switching. At the same time, in order to ensure the video quality, the I4x4 luminance block QDCT coefficient is combined with the technology to prevent intra-frame drift distortion, and the embedding capacity is improved under the premise of ensuring the video quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the embedding part of a large-capacity information hiding method based on short video scene switching provided by an embodiment of the present invention;

[0041] Figure 2a It is a 4x4 intra prediction mode in an embodiment of the present invention;

[0042] Figure 2b It is the 16x16 intra prediction mode in the embodiment of the present invention;

[0043] Figure 3 is a schematic diagram of reference pixel types in an embodiment of the present invention;

[0044] Figure 4a Schematic diagram of a specific coupling coefficient of a Cat1 information embedding type modulation in an embodiment of the present invention;

[0045] Figure 4b Schematic diagram of a specific coupling coefficient modulated by a Cat2 information embedding type in an embodiment of the present invention;

[0046] Figure 4c Schematic diagram of a specific coupling coefficient of a Cat3 information embedding type modulation in an embodiment of the present invention;

[0047] Figure 4d Schematic diagram of a specific coupling coefficient of a Cat4 information embedding type modulation in an embodiment of the present invention;

[0048] Figure 4eSchematic diagram of a specific coupling coefficient of a Cat5 information embedding type modulation in an embodiment of the present invention;

[0049] Figure 5 It is a flow chart of the extraction part of a large-capacity information hiding method based on short video scene switching provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The embodiment of the present invention provides a method for hiding large-capacity information using short video communication media, including information embedding and information extraction.

[0052] like Figure 1 As shown, the specific steps of information embedding are:

[0053] Step 1, obtaining the H.264 original video as a carrier and the information bit stream to be embedded;

[0054] Step 2: Determine whether there is a scene switch based on the content transformation of the previous and next two frames of the H.264 original video, calculate the scene change value C of each frame, and compare it with the set transition threshold V. If the change value C is greater than the transition threshold V, it is determined to be a transition frame, and define the continuous transition frames among all transition frames as the transition frame interval. It should be noted that a single discontinuous transition frame is not located in the transition frame interval.

[0055] Specifically, calculate the average value C1, C2, C3 of the change in the single pixel value in each dimension of the two frames in the HSV space, and then take the average value of C1, C2, C3 as the final scene change value C. The scene change value C is used to measure the difference between adjacent frames. The greater the difference, the greater the scene change value C, indicating that the correlation between adjacent frames is not strong, so the image can be significantly modified. Generally, the transition threshold V is 30.

[0056] Scene Change Value

[0057] Among them, w and h are the width and height of the video respectively. Where c represents the current frame. Where f represents the frame before the current frame, d represents the three channels H, S, and V, and (i, j) represents the index of the pixel value. is the pixel value component of the dth channel index (i, j) in the current frame, It is the pixel value component of the d-th channel at the (i, j) position in the previous frame of the current frame.

[0058] Step 3: Entropy decode the original H.264 video to obtain the QDCT coefficients of the I4x4 luminance block, its neighboring block intra prediction mode, and the motion vector difference MVD. Define a hierarchical screening condition based on the scene transformation value C and the threshold V calculated in Step 2, and hierarchically screen the alternative embedding blocks and the alternative motion vector difference MVD.

[0059] Among them, the intra prediction mode is divided into 4x4 intra prediction mode and 16x16 intra prediction mode. The 4x4 intra prediction mode is as Figure 2a shown, and the 16x16 intra prediction mode is as Figure 2b shown.

[0060] The hierarchical screening condition depends on the scene transformation value C and the transition threshold V of each frame. A larger scene transformation value indicates that the correlation between the current frame and the previous frame is not strong, and the number of alternative embedding blocks and alternative motion vector differences MVD can be increased by reducing the screening condition to greatly modify the current frame.

[0061] The hierarchical screening condition for the alternative embedding blocks is as follows:

[0062] When the current frame is an I frame, the I4x4 luminance block is an alternative embedding block if it satisfies the following conditions: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T1, T1>0, and the mean value of the absolute values of each QDCT coefficient is greater than or equal to the threshold T1', T1'>0. Among them, the values of the thresholds T1 and T1' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T1 = 1, T1' = 1; otherwise, T1 = 2, T1' = 2.

[0063] When the current frame is a P frame or a B frame, the I4x4 luminance block is an alternative embedding block if it satisfies the following conditions: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T2, T2>0, and the mean value of the absolute values of each QDCT coefficient is greater than or equal to the threshold T2', T2>0. Among them, the values of the thresholds T2 and T2' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T2 = 0, T2' = 0.5. If it is not in the transition frame interval, when the scene transformation value C satisfies C >= V or 0 <= C < V / 2, T2 = 2, T2' = 2; otherwise, T2 = 1, T2' = 1.

[0064] Among them, the DC coefficient is the first coefficient in the QDCT coefficient matrix.

[0065] The screening condition for the motion vector difference MVD is that the horizontal component MVDx or the vertical component MVDy of the motion vector difference MVD satisfies the condition MVDx ≥ 2n ||MVDy≥2 n ; If the current frame is not in the transition frame interval and the scene change value C satisfies 0 <= C < V / 2 or C >= V, then n = 3; if the scene change value C satisfies V / 2 <= C < V, then n = 2; if the current frame is in the transition frame interval, then n = 1.

[0066] Step 4, embed the QDCT coefficients and the motion vector difference MVD of the alternative embedding block into the information bitstream;

[0067] During the process of embedding the QDCT coefficients, in order to further ensure the video quality, the alternative embedding block is combined with the technology of preventing in-frame drift distortion. By analyzing the adjacent block reference pixel types of the alternative embedding block, it is determined which pixels of the alternative embedding block are referenced by subsequent predictive coding, and then the embedding type is determined. The least significant bit (LSB, Least Significant Byte) of the first coefficient in the specific QDCT coupling coefficient is embedded with information, and the non-first coefficients in the coupling coefficient are modified accordingly according to the embedding type.

[0068] In this operation, changing the specific QDCT coupling coefficient will not affect the pixel values referenced by the predictive coding of the adjacent blocks of the alternative embedding block, and finally the purpose of preventing in-frame drift distortion in I frames is achieved.

[0069] Among them, the specific QDCT coupling coefficient is a coefficient group composed of several QDCT coefficients at specific positions in the QDCT coefficient matrix, and is called the specific QDCT coupling coefficient.

[0070] Such as Figure 3 shown, in this embodiment, Case1, Case2, and Case3 in the adjacent block reference pixel types respectively represent whether the adjacent block predictive coding references the rightmost column pixels, the bottommost row pixels, and the bottom-right corner pixels of the alternative embedding block. The specific definition of the adjacent block reference pixel type is as follows:

[0071] Case1: (Right-Mode ∈ {1, 2, 4, 5, 6, 8} 4x4 U {1, 2, 3} 16x16 ), where Right-Mode is the intra-frame prediction mode of the right adjacent block of the alternative embedding block, and 4x4 or 16x16 is the coding mode of the adjacent block. That is, the prediction mode of the right adjacent block of the alternative embedding block is 1, 2, 4, 5, 6, 8 in the 4x4 coding mode or 1, 2, 3 in the 16x16 coding mode.

[0072] Case2: (Under-Mode ∈ {0, 2, 3, 4, 5, 6, 7} 4x4 U {0, 2, 3} 16x16 || Under-Left-Mode ∈ {3, 7} 4X4), where Under-Mode is the intra-frame prediction mode of the lower neighboring block of the candidate embedded block, Under-Left-Mode is the intra-frame prediction mode of the lower left neighboring block of the candidate embedded block, and 4x4 or 16x16 is the coding mode of the neighboring block. That is, the prediction mode of the lower neighboring block of the candidate embedded block is 0, 2, 3, 4, 5, 6, 7 in the 4x4 coding mode or 0, 2, 3 in the 16x16 coding mode; or the prediction mode of the lower left neighboring block is 0, 2, 3 in the 4x4 coding mode.

[0073] Case3: (Under-Right-Mode∈{4,5,6} 4x4 ), where Under-Right-Mode is the intra-frame prediction mode of the right neighboring block under the candidate embedded block, and 4x4 is the coding mode of the neighboring block. That is, the prediction mode of the right neighboring block under the candidate embedded block is 4, 5, and 6 in the 4x4 coding mode.

[0074] The embedding type is determined by the neighboring block reference pixel type. The embedding types Cat1, Cat2, Cat3, Cat4, and Cat5 represent the rightmost column pixel, bottommost row pixel, no pixel, bottom right corner pixel, rightmost column pixel, and bottommost row pixel of the candidate embedding block as references for subsequent encoding. The embedding types are specifically defined as follows:

[0075] like Figure 4a As shown, Cat1: (Case1==True and Case2==False), in this case, the rightmost column pixels of the candidate embedded block are referenced by subsequent prediction coding.

[0076] like Figure 4b As shown, Cat2: (Case1 == False and Case2 == True), in this case, the bottom row of pixels of the candidate embedded block is referenced by subsequent prediction coding.

[0077] like Figure 4c As shown, Cat3: (Case1==False and Case2==False and Case3==False), in this case, no pixel of the candidate embedded block is referenced by subsequent prediction coding.

[0078] like Figure 4d As shown, Cat4: (Case1==True and Case2==False and Case3==True), in this case, the lower right corner pixel of the candidate embedded block is referenced by the subsequent prediction coding;

[0079] like Figure 4eAs shown, Cat5: (Case1==True and Case2==True), in this case, the rightmost column pixel and the bottommost row pixel of the candidate embedded block are referenced by the subsequent prediction coding;

[0080] The specific QDCT coupling coefficients and embedding operations for each embedding type are as follows:

[0081] where Y ij The QDCT coefficient indexed by (i, j) in the QDCT coefficient matrix of the I4x4 luminance block represented.

[0082] (1) Cat1 = Cat11 ∪ Cat12;

[0083] Where Cat11 = {(Y 00 , Y 02 ), (Y 10 , Y 12 ), (Y 20 , Y 22 ), (Y 30 , Y 32 )},

[0084] Cat12={(Y 01 , Y 03 ), (Y 11 , Y 13 ), (Y 21 ,Y 23 ), (Y 31 ,Y 33 )}.

[0085] In the case of Cat1, the embedding operation is divided into two categories: ij , Y mn )∈Cat11;

[0086] If the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required. Otherwise, the coupling coefficient (Y ij , Y mn ). If Y ij ≥0, modulation (Y ij , Y mn ) is (Y ij +1, Y mn -1), otherwise (Y ij -1, Y mn +1). When the coupling coefficient (Y ij , Y mn )∈Cat12, if the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required, otherwise the coupling coefficient (Yij , Y mn ). If Y ij ≥0, modulation (Y ij , Y mn ) is (Y ij +1, Y mn -2), otherwise (Y ij -1, Y mn +2).

[0087] (2) Cat2 = Cat21∪Cat22;

[0088] Where Cat21 = {(Y 00 ,Y 20 ),(Y 01 ,Y 21 ),(Y 02 ,Y 22 ),(Y 03 ,Y 23 )},

[0089] Cat22={(Y 10 ,Y 30 ),(Y 11 ,Y 31 ),(Y 12 ,Y 32 ),(Y 13 ,Y 33 )}.

[0090] In the case of Cat2, the embedding operation is divided into two categories: ij , Y mn )∈Cat21, if the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required, otherwise the coupling coefficient (Y ij , Y mn ). If Y ij ≥0, modulation (Y ij , Y mn ) is (Y ij +1, Y mn -1), otherwise (Y ij -1, Y mn +1). When the coupling coefficient (Y ij , Y mn )∈Cat22, if the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required, otherwise the coupling coefficient (Y ij , Y mn ). If Y ij ≥0, modulation (Y ij , Ymn ) is (Y ij +1, Y mn -2), otherwise (Y ij -1, Y mn +2).

[0091] (3)Cat3={Y ij |i,j∈(0,1,2,3),Y ij !=0};

[0092] In the case of Cat3, the embedding operation is to embed the bit information to be embedded into the non-zero QDCT coefficient through the least significant bit LSB algorithm, and it is necessary to ensure that the QDCT coefficient after embedding is not zero.

[0093] (4) Cat4 = Cat2;

[0094] The coupling coefficients and embedding operation specific to the Cat4 embedding type are the same as those in Cat2.

[0095] (5) Cat5 = Cat51 ∪ Cat52;

[0096] Where Cat51 = {(Y 00 ,Y 02 ,Y 20 ,Y 22 )}, Cat52={(Y 10 ,Y 12 ,Y 30 ,Y 32 )}.

[0097] In the case of Cat5, the embedding operation is divided into two categories: ij , Y mn , Y pq , Y xy )∈Cat51, if the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required, otherwise the coupling coefficient (Y ij , Y mn , Y pq , Y xy ). If Y ij ≥0, modulation (Y ij , Y mn , Y pq , Y xy ) is (Y ij +1, Y mn -1, Y pq -1, Y xy +1), otherwise (Y ij -1, Y mn +1, Ypq +1, Y xy -1). When the coupling coefficient (Y ij , Y mn , Y pq , Y xy )∈Cat52, if the bit information to be embedded is the same as the least significant bit of the first coefficient of the coupling coefficient, no modification is required, otherwise the coupling coefficient (Y ij , Y mn , Y pq , Y xy ). If Y ij ≥0, modulation (Y ij , Y mn , Y pq , Y xy ) is (Yij+1, Ymn-1, Ypq-2, Yxy+2), otherwise it is (Yij-1, Ymn+1, Ypq+1, Yxy-1).

[0098] Finally, the specific QDCT coupling coefficient is obtained according to the embedding type, the least significant bit (LSB) of the first coefficient in the specific QDCT coupling coefficient is embedded with bit information, and the non-first coefficients in the coupling coefficient are changed accordingly to achieve the purpose of preventing drift distortion within the I frame.

[0099] During the embedding process of the motion vector difference MVD, only the least significant bit LSB of the horizontal component MVDx or the vertical component MVDy that meets the MVD screening condition is embedded, and 1 bit can be embedded at this time; if both the horizontal component MVDx and the vertical component MVDy meet the embedding conditions, they are both embedded, and 2 bits can be embedded at this time.

[0100] Step 5: re-encode each frame of the embedded information to obtain the H264 carrier video of the embedded information.

[0101] The following is an example of the embedding process of the candidate embedding block: Assuming that a candidate embedding block satisfies the embedding type Cat1, the QDCT coefficient matrix is ​​as follows: The bit information to be embedded is 01100101, then according to the embedding rule, we get:

[0102] Original coupling coefficient Bit information to be embedded Coupling coefficient after modulation (4，-4) 0 (4，-4) (0，2) 1 (1，1) (5，0) 1 (5，0) (-4，3) 0 (-4，3) (-5，3) 0 (-6，2) (-2，0) 1 (-3，1) (2，0) 0 (2，0) (2，0) 1 (3，-1)

[0103] The final modulated QDCT coefficient matrix is:

[0104] like Figure 5 As shown in the figure, the specific steps of information extraction are:

[0105] Step 1: Get the H.264 carrier video with embedded information;

[0106] Step ②, use the scene switching detection algorithm to calculate the scene change value C of each frame of the H.264 original video, and compare it with the set transition threshold V. If the change value C is greater than the transition threshold V, it is determined to be a transition frame, and the continuous transition frames among all transition frames are defined as the transition frame interval.

[0107] The details are the same as step 2.

[0108] Step ③, decode the H.264 carrier video, obtain the QDCT coefficients of the I4x4 luminance block and its neighboring block luminance prediction mode, motion vector difference MVD, and define different screening conditions according to the scene change value C and threshold V of each frame calculated in step ②, and screen the candidate extraction blocks and candidate motion vector difference MVD in a hierarchical manner.

[0109] The details are the same as step 3.

[0110] Step ④, extracting the information bits embedded in the QDCT coefficients of the candidate extraction blocks and the candidate motion vector difference MVD;

[0111] For the extraction process of the candidate extraction block, it is necessary to determine the extraction type of the candidate extraction block, and the determination of the extraction type is consistent with the determination of the embedding type and the specific QDCT coupling coefficient of each embedding type in step 4 of embedding information. According to the extraction type, the least significant bit of the first coefficient in the specific QDCT coupling coefficient is extracted to extract the bit information.

[0112] For the extraction process of the candidate motion vector difference MVD, only the least significant bit of the horizontal component MVDx or the vertical component MVDy of the motion vector difference that meets the screening conditions in step ③ is extracted. If both the horizontal component MVDx and the vertical component MVDy meet the screening conditions, 2 bits of information can be extracted.

[0113] The following is an example of the extraction process of the candidate extraction block: Assuming that a candidate extraction block satisfies the extraction type Cat1, the QDCT coefficient matrix is ​​as follows:

[0114] The extraction process is:

[0115] Coupling coefficient to be extracted Extracting Bit Information (4，-4) 0 (1，1) 1 (5，0) 1 (-4，3) 0 (-6，2) 0 (-3，1) 1 (2，0) 0 (3，-1) 1

[0116] The final bit information extracted is 01100101.

[0117] In order to verify the feasibility and performance of the method, the method of this embodiment is implemented based on the experimental platform JM8.6, the video encoding parameters are consistent with the original video, the number of test frames is 100 frames, and the remaining parameters are set to default parameters. In order to better evaluate the experimental effect, 5 videos with rich transition effects were selected online, and the resolution was 320x240. Compared with the conventional solution, the screening threshold in the screening process of the alternative embedded blocks and the alternative motion vector difference in the conventional solution is a determined value, and the embedding method is consistent with the embedding method of the solution in this paper. The embedding capacity comparison is as follows:

[0118]

[0119] The embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and the computer program can be executed by a processor to implement any of the above methods. Its implementation principle and technical effect are similar to those of the above methods, and will not be repeated here.

[0120] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for hiding a large amount of information using a short video communication medium, including information embedding and information extraction; characterized in that: The information embedding specifically includes the following steps: Step 1, obtain the original H.264 video as the carrier and the information bit stream to be embedded; Step 2, judge whether there is a scene transition for the original H.264 video based on the content transformation between the previous and the next frames. By calculating the average values C1, C2, and C3 of the change in the single-pixel value of each dimension in the HSV space between the previous and the next frames, and then taking the average value of C1, C2, and C3 to obtain the scene transformation value C; compare it with the set transition threshold V. If the transformation value C is greater than the transition threshold V, it is determined as a transition frame, and all consecutive transition frames in all transition frames are defined as a transition frame interval; Step 3, perform entropy decoding on the original H.264 video to obtain the QDCT coefficients of the I4x4 luminance block, its neighboring block intra prediction mode, and the motion vector difference MVD, and define a hierarchical screening condition according to the scene transformation value C and the threshold V calculated in Step 2 for each frame, and hierarchically screen the alternative embedding blocks and the alternative motion vector differences MVD; The hierarchical screening conditions for the alternative embedding blocks are as follows: When the current frame is an I frame, the I4x4 luminance block is an alternative embedding block if it satisfies the following conditions: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T1, T1>0, and the average value of the absolute values of each item of the QDCT coefficients is greater than or equal to the threshold T1', T1'>0; Among them, the values of the thresholds T1 and T1' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T1 = 1, T1' = 1; otherwise T1 = 2, T1' = 2; When the current frame is a P frame or a B frame, the I4x4 luminance block is an alternative embedding block if it satisfies the following conditions: the absolute value of the DC coefficient of the I4x4 luminance block is greater than or equal to the threshold T2, T2>0, and the average value of the absolute values of each item of the QDCT coefficients is greater than or equal to the threshold T2', T2>0; Among them, the values of the thresholds T2 and T2' depend on whether the current frame is in the transition frame interval. If the current frame is in the transition frame interval, T2 = 0, T2' = 0.5; if not in the transition frame interval, when the scene transformation value C satisfies C >= V or 0 <= C < V / 2, T2 = 2, T2' = 2; otherwise T2 = 1, T2' = 1; Among them, the DC coefficient is the first coefficient in the QDCT coefficient matrix; The screening condition for the motion vector difference MVD is that the horizontal component MVDx or the vertical component MVDy of the motion vector difference MVD satisfies the condition MVDx≥2 n ||MVDy≥2 n ; if the current frame is not located in the transition frame interval and the scene transformation value C satisfies 0 <= C < V / 2 or C >= V, then n = 3; if the scene transformation value C satisfies V / 2 <= C < V, then n = 2; if the current frame is located in the transition frame interval, then n = 1; Step 4, embed the information bit stream into the QDCT coefficients and the motion vector difference MVD of the alternative embedding blocks; During the process of embedding the QDCT coefficients, determine which pixels of the alternative embedding block are referenced by subsequent predictive coding by analyzing the neighboring block reference pixel types of the alternative embedding block, and then determine the embedding type. Obtain a specific QDCT coupling coefficient according to the embedding type, perform least significant bit embedding of information on the first coefficient in the specific QDCT coupling coefficient, and make corresponding changes to the non-first coefficients in the coupling coefficient according to the embedding type to achieve the purpose of preventing intra-frame drift distortion of the I frame; Among them, the specific QDCT coupling coefficient is a coefficient group composed of several specific-position QDCT coefficients in the QDCT coefficient matrix, and is called a specific QDCT coupling coefficient; In the process of embedding the motion vector difference MVD, only the least significant bit LSB of the horizontal component MVDx or the vertical component MVDy that meets the MVD screening condition is embedded, and 1 bit can be embedded at this time; if both the horizontal component MVDx and the vertical component MVDy meet the embedding condition, they are both embedded, and 2 bits can be embedded at this time; Step 5, re-encode each frame of the embedded information to obtain the H264 carrier video of the embedded information; The specific steps of information extraction are: Step 1: Get the H.264 carrier video with embedded information; Step ②, for the H.264 original video, the average values ​​C1, C2, and C3 of the single pixel value changes in each dimension of the two frames before and after the HSV space are calculated, and then the average values ​​of C1, C2, and C3 are taken to obtain the scene change value C; compared with the set transition threshold V, if the change value C is greater than the transition threshold V, it is determined to be a transition frame, and the continuous transition frames in all transition frames are defined as the transition frame interval; Step ③, decode the H.264 carrier video, obtain the QDCT coefficient of the I4x4 luminance block and its neighboring block luminance prediction mode, motion vector difference MVD, and define different screening conditions according to the scene change value C and threshold V of each frame calculated in step ②, and screen the candidate extraction blocks and candidate motion vector difference MVD in a hierarchical manner; the screening conditions are consistent with step 3; Step ④, extracting the information bits embedded in the QDCT coefficients of the candidate extraction blocks and the candidate motion vector difference MVD; For the extraction process of the candidate extraction block, it is necessary to determine the extraction type of the candidate extraction block, and the determination of the extraction type is consistent with the determination of the embedding type in the specific QDCT coupling coefficient and information embedding step 4 of each extraction type and the specific QDCT coupling coefficient of each embedding type; according to the extraction type, the least significant bit of the first coefficient in the specific QDCT coupling coefficient is extracted to extract the bit information; For the extraction process of the candidate motion vector difference MVD, only the least significant bit of the horizontal component MVDx or the vertical component MVDy of the motion vector difference that meets the screening conditions in step ③ is extracted. If both the horizontal component MVDx and the vertical component MVDy meet the screening conditions, 2 bits of information can be extracted.

2. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, which, when executed on a computer, cause the computer to perform the method of claim 1 .

Images