Re-compression Forensics Method for Information Fusion of Chrominance Domain and Luminance Domain in Video Bitstream

Abstract

A video stream recompression forensics method for fusing chroma domain and luminance domain information comprises the steps of extracting primary stream features of the video stream, determining a chroma domain attribute matrix of the video stream, determining a luminance domain attribute matrix of the video stream, recompressing the video stream, extracting primary stream features of the recompressed video stream, determining a chroma domain attribute matrix of the recompressed video stream, determining a chroma domain attribute tag of the video stream, determining a ratio of coding units with different attributes in the chroma domain of the video stream, determining a luminance domain attribute matrix of the recompressed video stream, determining a luminance domain attribute tag of the video stream, determining a ratio of coding units with different attributes in the luminance domain of the video stream, determining advanced stream features of the video stream, and determining a recompression forensics result. The present invention has the advantages of high accuracy and low complexity in video recompression forensics and can be used in the field of video stream recompression forensics technology for standards such as H.266 / VVC, H.265 / HEVC, and H.264 / AVC that support coding unit division types and prediction mode technologies.

Description

Technical Field

[0001] The present invention belongs to the technical field of video evidence collection, and in particular relates to re-compression evidence collection of video code streams after video coding compression. Background Art

[0002] In recent years, with the increasing power of image editing software, video tampering has become increasingly easy, and ensuring the security of video data has become a widespread consensus. Consequently, video forensics techniques for detecting tampering have also garnered significant attention. In video communication systems, video signals undergo a series of operations, including acquisition, encoding and compression, transmission, storage, decoding, and display. The encoded and compressed video stream can be compromised during network transmission and storage. Tamperers then manipulate the stolen video data, and the altered video content must be re-encoded and compressed before it can be effectively transmitted and stored. Therefore, the number of encoding and compression cycles becomes a crucial criterion for determining video tampering. Consequently, video re-compression forensics methods designed to detect video encoding and compression cycles have emerged.

[0003] Current approaches to video recompression forensics suffer from the following issues: 1) Most methods only consider the impact of luminance-domain coding information on the number of compressions performed on the video bitstream, but fail to consider the impact of chrominance-domain information on the number of compressions performed. 2) Most methods are designed for earlier video coding standards, such as MPEG-2, MPEG-4, H.264 / AVC, and H.265 / HEVC, and are not applicable to the latest H.266 / VVC standard. The latest international video coding standard is H.266 / VVC, released in 2020. Compared to earlier video coding standards, H.266 / VVC utilizes more advanced coding techniques, significantly improving coding efficiency while also significantly altering the syntax, semantics, and compression artifacts of the video bitstream. With its superior coding efficiency and versatility, the H.266 / VVC standard holds broad market potential. 3) These methods are generally complex. For example, some methods require recompression of the video bitstream being recompressed more than once during the forensic process, significantly increasing the complexity and limiting their practical application. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a video recompression forensics method that fuses the chroma domain and luminance domain information of the video code stream with high accuracy and low complexity.

[0005] The technical solution adopted to solve the above technical problems consists of the following steps:

[0006] (1) Extracting primary stream features of video stream

[0007] The collected video stream is decoded and analyzed using decoding software to extract primary stream features in the chroma and luma domains. The primary chroma domain features include the division type and prediction mode of the chroma domain coding units of the intra-frame prediction frame. The primary luma domain features include the division type and prediction mode of the luma domain coding units of the intra-frame prediction frame. The decoding and analysis software is publicly available at https: / / vcgit.hhi.fraunhofer.de / jvet / VVCSoftware_VTM.

[0008] (2) Determine the chromaticity domain attribute matrix of the video stream

[0009] According to formula (1), the attribute matrix A of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x :

[0010] A i,x =[E i,x ,F i,x ] (1)

[0011] Among them, E i,x 、F i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream, i∈{1,2,…,I}, I represents the total number of intra-frame prediction frames contained in the video code stream, I is a finite positive integer, and x∈{1,2,…,M i}, M i Indicates the total number of chroma domain coding units contained in the i-th intra-frame prediction frame in the video stream, M i is a finite positive integer. The chroma domain coding units are numbered from left to right and from top to bottom. i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ represent the optional partition types and total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, respectively, and θ and ξ are finite positive integers.

[0012] (3) Determine the brightness domain attribute matrix of the video stream

[0013] Determine the attribute matrix B of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream according to formula (2): i,y :

[0014] B i,y =[G i,y ,H i,y ] (2)

[0015] Among them, G i,y 、H i,yThey represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream, y∈{1,2,…,N i}, N i Indicates the total number of luminance domain coding units contained in the i-th intra-frame prediction frame in the video stream, N i is a finite positive integer, and the luminance domain coding units are numbered from left to right and from top to bottom. H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε is a finite positive integer.

[0016] (4) Recompress the video stream

[0017] The video code stream is decoded into a decoded video using a decoder supported by the video coding standard, and the decoded video is re-encoded and compressed once using an encoder supported by the video coding standard to obtain a re-compressed video code stream.

[0018] (5) Extracting primary stream features of the recompressed video stream

[0019] The primary code stream features in the chroma domain and luminance domain are extracted from the re-compressed video code stream using decoding analysis software. The primary code stream features in the chroma domain include the division type and prediction mode of the chroma domain coding unit of the intra-frame prediction frame, and the primary code stream features in the luminance domain include the division type and prediction mode of the luminance domain coding unit of the intra-frame prediction frame.

[0020] (6) Determine the chroma domain attribute matrix of the re-compressed video stream

[0021] According to formula (3), the attribute matrix C of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined: i,x :

[0022] C i,x =[K i,x ,P i,x ] (3)

[0023] Among them, K i,x 、P i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream, K i,x ∈{1,2,…,θ},P i,x ∈{1,2,…,ξ}.

[0024] (7) Determine the chroma domain attribute mark of the video stream

[0025] According to formula (4), the attribute flag β of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x,α :

[0026]

[0027] Among them, α∈{1,2,…,θ}, A i,x ≠C i,x The coding units are different attribute coding units in the chroma domain.

[0028] (8) Determine the proportion of different attribute coding units in the chroma domain of the video stream

[0029] According to formula (5), the proportion of different attribute coding units of the αth chroma domain coding unit division type in the video stream is determined as W. α :

[0030]

[0031] Among them, M i,α Indicates the total number of coding units of the αth chroma domain coding unit partition type selected in the i-th intra-frame prediction frame in the video stream, M i,α is a finite positive integer.

[0032] (9) Determine the brightness domain attribute matrix of the re-compressed video stream

[0033] According to formula (6), the attribute matrix D of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined i,y :

[0034] D i,y =[Q i,y ,R i,y ] (6)

[0035] Among them, Q i,y 、R i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream, respectively. R i,y ∈{1,2,…,ε}.

[0036] (10) Determine the brightness domain attribute mark of the video stream

[0037] According to formula (7), the attribute flag δ of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,y,γ :

[0038]

[0039] in, Bi,y ≠D i,y The coding units are different attribute coding units in the brightness domain.

[0040] (11) Determine the proportion of different attribute coding units in the brightness domain of the video stream

[0041] According to formula (8), the proportion of different attribute coding units of the γth type of luminance domain coding unit division type in the video code stream is determined as V γ :

[0042]

[0043] Among them, N i,γ Indicates the total number of coding units of the γth luminance domain coding unit partitioning type selected in the i-th intra-frame prediction frame in the video code stream, N i,γ is a finite positive integer.

[0044] (12) Determine the high-level stream characteristics of the video stream

[0045] According to formula (9), the high-level stream feature matrix T of the video stream is determined:

[0046]

[0047] (13) Determine the results of recompression forensics

[0048] The high-level bitstream feature matrix T is input into a support vector machine pattern recognition and regression software package. Based on the binary classification data output by the support vector machine pattern recognition and regression software package, the recompression forensics result corresponding to the video bitstream is finally obtained. The support vector machine pattern recognition and regression software package is publicly available at https: / / www.csie.ntu.edu.tw / ~cjlin / libsvm / .

[0049] In the formula (1) for determining the chroma domain attribute matrix of the video code stream in (2) of the present invention, the E i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ represent the optional partition types and total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, θ∈[1,19] and ξ∈[1,70].

[0050] In the formula (2) of the step (3) of determining the brightness domain attribute matrix of the video code stream of the present invention, the H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε∈[1,67].

[0051] Since the present invention adopts the steps of determining the chroma domain attribute matrix of the video stream, determining the chroma domain attribute matrix of the re-compressed video stream, and determining the chroma domain attribute marking of the video stream, it solves the problem that the existing technology does not consider the influence of chroma domain coding information and the step of determining the proportion of different attribute coding units in the chroma domain of the video stream. It adopts the step of determining the advanced stream characteristics of the video stream, and solves the problem of not being applicable to the latest H.266 / VVC standard and the high complexity of evidence collection. The present invention has the advantages of high accuracy and low complexity in video recompression forensics, and can be used in the field of video stream recompression forensics technology of standards such as H.266 / VVC, H.265 / HEVC, H.264 / AVC, etc. that support coding unit division type and prediction mode technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flowchart of Example 1 of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described in detail below with reference to the accompanying drawings and examples, but the present invention is not limited to the following embodiments.

[0054] Example 1

[0055] The video code stream chroma domain and luminance domain information fusion recompression evidence collection method of this embodiment consists of the following steps (see Figure 1 ):

[0056] (1) Extracting primary stream features of video stream

[0057] The collected video stream is decoded and analyzed using decoding software to extract primary stream features in the chroma and luma domains. The primary chroma domain features include the division type and prediction mode of the chroma domain coding units of the intra-frame prediction frame. The primary luma domain features include the division type and prediction mode of the luma domain coding units of the intra-frame prediction frame. The decoding and analysis software is publicly available at https: / / vcgit.hhi.fraunhofer.de / jvet / VVCSoftware_VTM.

[0058] (2) Determine the chromaticity domain attribute matrix of the video stream

[0059] According to formula (1), the attribute matrix A of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x :

[0060] A i,x =[E i,x ,F i,x ] (1)

[0061] Among them, Ei,x 、F i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream, i∈{1,2,…,I}, I represents the total number of intra-frame prediction frames contained in the video code stream, I is a finite positive integer, and x∈{1,2,…,M i}, M i Indicates the total number of chroma domain coding units contained in the i-th intra-frame prediction frame in the video stream, M i is a finite positive integer. The chroma domain coding units are numbered from left to right and from top to bottom. i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ represent the optional partition types and total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, θ∈[1,19] and ξ∈[1,70]. In this embodiment, θ is 9 and ξ is 35.

[0062] (3) Determine the brightness domain attribute matrix of the video stream

[0063] Determine the attribute matrix B of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream according to formula (2): i,y :

[0064] B i,y =[G i,y ,H i,y ] (2)

[0065] Among them, G i,y 、H i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream, y∈{1,2,…,N i}, N i Indicates the total number of luminance domain coding units contained in the i-th intra-frame prediction frame in the video stream, N i is a finite positive integer, and the luminance domain coding units are numbered from left to right and from top to bottom. H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε∈[1,67], in this embodiment The value is 9, and the value of ε is 33.

[0066] (4) Recompress the video stream

[0067] A decoder supported by the video coding standard is used to decode the video stream into a decoded video, and an encoder supported by the video coding standard is used to re-encode and compress the decoded video to obtain a re-compressed video stream.

[0068] (5) Extracting primary stream features of the recompressed video stream

[0069] The primary code stream features in the chroma domain and luminance domain are extracted from the re-compressed video code stream using decoding analysis software. The primary code stream features in the chroma domain include the division type and prediction mode of the chroma domain coding unit of the intra-frame prediction frame, and the primary code stream features in the luminance domain include the division type and prediction mode of the luminance domain coding unit of the intra-frame prediction frame.

[0070] (6) Determine the chroma domain attribute matrix of the re-compressed video stream

[0071] According to formula (3), the attribute matrix C of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined: i,x :

[0072] C i,x =[K i,x ,P i,x ] (3)

[0073] Among them, K i,x 、P i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream, K i,x ∈{1,2,…,θ},P i,x ∈{1,2,…,ξ}.

[0074] (7) Determine the chroma domain attribute mark of the video stream

[0075] According to formula (4), the attribute flag β of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x,α :

[0076]

[0077] Among them, α∈{1,2,…,θ}, A i,x ≠C i,x The coding units are different attribute coding units in the chroma domain.

[0078] (8) Determine the proportion of different attribute coding units in the chroma domain of the video stream

[0079] According to formula (5), the proportion of different attribute coding units of the αth chroma domain coding unit division type in the video stream is determined as W. α :

[0080]

[0081] Among them, M i,α Indicates the total number of coding units of the αth chroma domain coding unit partition type selected in the i-th intra-frame prediction frame in the video stream, M i,α is a finite positive integer.

[0082] (9) Determine the brightness domain attribute matrix of the re-compressed video stream

[0083] According to formula (6), the attribute matrix D of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined i,y :

[0084] D i,y =[Q i,y ,R i,y ] (6)

[0085] Among them, Q i,y 、R i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream, respectively. , R i,y ∈{1,2,…,ε}.

[0086] (10) Determine the brightness domain attribute mark of the video stream

[0087] According to formula (7), the attribute flag δ of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,y,γ :

[0088]

[0089] in, B i,y ≠D i,y The coding units are different attribute coding units in the brightness domain.

[0090] (11) Determine the proportion of different attribute coding units in the brightness domain of the video stream

[0091] According to formula (8), the proportion of different attribute coding units of the γth type of luminance domain coding unit division type in the video code stream is determined as V γ :

[0092]

[0093] Among them, N i,γ Indicates the total number of coding units of the γth luminance domain coding unit partitioning type selected in the i-th intra-frame prediction frame in the video code stream, N i,γ is a finite positive integer.

[0094] (12) Determine the advanced stream characteristics of the video stream

[0095] According to formula (9), the high-level stream feature matrix T of the video stream is determined:

[0096]

[0097] (13) Determine the results of recompression forensics

[0098] The high-level bitstream feature matrix T is input into a support vector machine pattern recognition and regression software package. Based on the binary classification data output by the support vector machine pattern recognition and regression software package, the recompression forensics result corresponding to the video bitstream is finally obtained. The support vector machine pattern recognition and regression software package is publicly available at https: / / www.csie.ntu.edu.tw / ~cjlin / libsvm / .

[0099] A recompression forensics method that integrates chroma domain and luminance domain information of video streams is completed.

[0100] Example 2

[0101] The recompression evidence collection method for fusion of chroma domain and luminance domain information of a video stream in this embodiment comprises the following steps:

[0102] (1) Extracting primary stream features of video stream

[0103] This step is the same as in Example 1.

[0104] (2) Determine the chromaticity domain attribute matrix of the video stream

[0105] According to formula (1), the attribute matrix A of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x :

[0106] A i,x =[E i,x ,F i,x ] (1)

[0107] Among them, E i,x 、F i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream, i∈{1,2,…,I}, I represents the total number of intra-frame prediction frames contained in the video code stream, I is a finite positive integer, and x∈{1,2,…,M i}, M i Indicates the total number of chroma domain coding units contained in the i-th intra-frame prediction frame in the video stream, M i is a finite positive integer. The chroma domain coding units are numbered from left to right and from top to bottom. i,x∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ respectively represent the optional partition type and the total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, θ∈[1,19], ξ∈[1,70]. In this embodiment, θ takes a value of 1 and ξ takes a value of 1.

[0108] (3) Determine the brightness domain attribute matrix of the video stream

[0109] Determine the attribute matrix B of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream according to formula (2): i,y :

[0110] B i,y =[G i,y ,H i,y ] (2)

[0111] Among them, G i,y 、H i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream, y∈{1,2,…,N i}, N i Indicates the total number of luminance domain coding units contained in the i-th intra-frame prediction frame in the video stream, N i is a finite positive integer, and the luminance domain coding units are numbered from left to right and from top to bottom. H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε∈[1,67], in this embodiment The value is 1, and the value of ε is 1.

[0112] The other steps are the same as those in Example 1. A recompression evidence collection method for fusing chroma domain and luminance domain information of a video stream is completed.

[0113] Example 3

[0114] The recompression evidence collection method for fusion of chroma domain and luminance domain information of a video stream in this embodiment comprises the following steps:

[0115] (1) Extracting primary stream features of video stream

[0116] This step is the same as in Example 1.

[0117] (2) Determine the chromaticity domain attribute matrix of the video stream

[0118] According to formula (1), the attribute matrix A of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determinedi,x :

[0119] A i,x =[E i,x ,F i,x ] (1)

[0120] Among them, E i,x 、F i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream, i∈{1,2,…,I}, I represents the total number of intra-frame prediction frames contained in the video code stream, I is a finite positive integer, and x∈{1,2,…,M i}, M i Indicates the total number of chroma domain coding units contained in the i-th intra-frame prediction frame in the video stream, M i is a finite positive integer. The chroma domain coding units are numbered from left to right and from top to bottom. i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ respectively represent the optional partition types and the total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, θ∈[1,19], ξ∈[1,70]. In this embodiment, θ is 19 and ξ is 70.

[0121] (3) Determine the brightness domain attribute matrix of the video stream

[0122] Determine the attribute matrix B of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream according to formula (2): i,y :

[0123] B i,y =[G i,y ,H i,y ] (2)

[0124] Among them, G i,y 、H i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream, y∈{1,2,…,N i}, N i Indicates the total number of luminance domain coding units contained in the i-th intra-frame prediction frame in the video stream, N i is a finite positive integer, and the luminance domain coding units are numbered from left to right and from top to bottom. H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε∈[1,67], in this embodiment The value is 17 and the value of ε is 67.

[0125] The other steps are the same as those in Example 1. A recompression evidence collection method for fusing chroma domain and luminance domain information of a video stream is completed.

[0126] In order to verify the beneficial effects of the present invention, the inventors conducted an experiment on a test video using the method of Example 3 of the present invention. The experimental results are as follows:

[0127] 27 standard videos recommended by the coding standard formulation organization are selected as test videos, and the color space of the video is YCbCr4:2:0. The reference software VTM-11.0 recommended by the H.266 / VVC standard is used to encode and compress the test video once and twice to obtain the corresponding video stream, and the video stream is input into the method of embodiment 3 of the present invention for recompression and evidence collection. The encoding level is main_10, the encoding structure is a low-delay encoding structure, and the quantization parameters are set to 22, 25, 27, 29, 32, 35, 37, 39, and 42 respectively. For other encoding parameters, refer to the default settings in the configuration file encoder_lowdelay_P_vtm.cfg. The video recompression and evidence collection accuracy s corresponding to the method of embodiment 3 of the present invention is calculated according to formula (10):

[0128] s=(u+j) / 2 (10)

[0129] Wherein, u represents the accuracy of the present invention in identifying single-compressed videos, and j represents the accuracy of the present invention in identifying heavily compressed videos. The results are shown in Table 1.

[0130] Table 1 Video recompression forensics accuracy (s) of the method in Example 3 of the present invention

[0131]

[0132] As can be seen from Table 1, under each test quantization parameter, the video recompression forensics accuracy of the method in Example 3 of the present invention is greater than or equal to 92.59%. For all test quantization parameters, the video recompression forensics accuracy of the method in Example 3 of the present invention is 96.50% on average, indicating that the video recompression forensics accuracy of the method in Example 3 of the present invention is high.

[0133] In the technical solution steps of the present invention, the video stream to be recompressed for evidence only needs to be encoded and compressed once in the step of recompressing the video stream, which shows that the video recompression forensics method of embodiment 3 of the present invention has low complexity.

Claims

1. A video code stream chroma domain and luminance domain information fusion recompression forensics method, characterized by It consists of the following steps: (1) Extracting primary stream features of video stream The primary code stream features of the chroma domain and luma domain in the collected video code stream are extracted using decoding analysis software. The primary code stream features of the chroma domain include the division type and prediction mode of the chroma domain coding unit of the intra-frame prediction frame, and the primary code stream features of the luma domain include the division type and prediction mode of the luma domain coding unit of the intra-frame prediction frame. (2) Determine the chromaticity domain attribute matrix of the video stream According to formula (1), the attribute matrix A of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x : Among them, E i,x 、F i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream, i∈{1,2,…,I}, I represents the total number of intra-frame prediction frames contained in the video code stream, I is a finite positive integer, and x∈{1,2,…,M i }, M i Indicates the total number of chroma domain coding units contained in the i-th intra-frame prediction frame in the video stream, M i is a finite positive integer. The chroma domain coding units are numbered from left to right and from top to bottom. i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ represent the optional partition type and total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, and θ and ξ are finite positive integers; (3) Determine the brightness domain attribute matrix of the video stream Determine the attribute matrix B of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream according to formula (2): i,y : Among them, G i,y 、H i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream, y∈{1,2,…,N i }, N i Indicates the total number of luminance domain coding units contained in the i-th intra-frame prediction frame in the video stream, N i is a finite positive integer, and the luminance domain coding units are numbered from left to right and from top to bottom. H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε is a finite positive integer; (4) Recompress the video stream Decoding the video stream into a decoded video using a decoder supported by the video coding standard, and re-encoding and compressing the decoded video once using an encoder supported by the video coding standard to obtain a re-compressed video stream; (5) Extracting primary stream features of the recompressed video stream The primary code stream features in the chroma domain and luma domain are extracted from the recompressed video code stream using decoding analysis software. The primary code stream features in the chroma domain include the division type and prediction mode of the chroma domain coding unit of the intra-frame prediction frame, and the primary code stream features in the luma domain include the division type and prediction mode of the luma domain coding unit of the intra-frame prediction frame. (6) Determine the chroma domain attribute matrix of the re-compressed video stream According to formula (3), the attribute matrix C of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined: i,x : Among them, K i,x 、P i,x They represent the partition type and prediction mode number of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream, K i,x ∈{1,2,…,θ},P i,x ∈{1,2,…,ξ}; (7) Determine the chroma domain attribute mark of the video stream According to formula (4), the attribute flag β of the xth chroma domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,x,α : Among them, α∈{1,2,…,θ}, A i,x ≠C i,x The coding unit is a coding unit of different attributes in the chroma domain; (8) Determine the proportion of different attribute coding units in the chroma domain of the video stream According to formula (5), the proportion of different attribute coding units of the αth chroma domain coding unit division type in the video stream is determined as W. α : Among them, M i,α Indicates the total number of coding units of the αth chroma domain coding unit partition type selected in the i-th intra-frame prediction frame in the video stream, M i,α is a finite positive integer; (9) Determine the brightness domain attribute matrix of the re-compressed video stream According to formula (6), the attribute matrix D of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video code stream is determined i,y : Among them, Q i,y 、R i,y They represent the partition type and prediction mode number of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the re-compressed video stream, respectively. R i,y ∈{1,2,…,ε}; (10) Determine the brightness domain attribute mark of the video stream According to formula (7), the attribute flag δ of the yth luminance domain coding unit of the i-th intra-frame prediction frame in the video code stream is determined i,y,γ : in, B i,y ≠D i,y The coding units are different attribute coding units in the brightness domain; (11) Determine the proportion of different attribute coding units in the brightness domain of the video stream According to formula (8), the proportion of different attribute coding units of the γth type of luminance domain coding unit division type in the video code stream is determined as V γ : Among them, N i,γ Indicates the total number of coding units of the γth luminance domain coding unit partitioning type selected in the i-th intra-frame prediction frame in the video code stream, N i,γ is a finite positive integer; (12) Determine the advanced stream characteristics of the video stream According to formula (9), the high-level bitstream feature matrix T of the video bitstream is determined: (13) Determine the results of recompression forensics The high-level bitstream feature matrix T is input into the support vector machine pattern recognition and regression software package, and the heavy compression forensics result corresponding to the video bitstream is finally obtained based on the two-classification data output by the support vector machine pattern recognition and regression software package.

2. The recompression evidence collection method for fusion of chroma domain and luminance domain information of a video stream according to claim 1 is characterized by: In the formula (1) for determining the chroma domain attribute matrix of the video code stream (2), the E i,x ∈{1,2,…,θ},F i,x ∈{1,2,…,ξ}, θ and ξ represent the optional partition types and total number of prediction modes of the intra-frame prediction frame chroma domain coding unit supported by the video coding standard, θ∈[1,19] and ξ∈[1,70].

3. The recompression evidence collection method for fusion of chroma domain and luminance domain information of a video stream according to claim 1 is characterized by: In the formula (2) of the step (3) of determining the brightness domain attribute matrix of the video code stream, the H i,y ∈{1,2,…,ε}, ε represents the optional partition type and total number of prediction modes of the intra-frame prediction frame luminance domain coding unit supported by the video coding standard, ε∈[1,67].

Images