CN104853186A - Improved video steganalysis method based on motion vector reply - Google Patents
Improved video steganalysis method based on motion vector reply Download PDFInfo
- Publication number
- CN104853186A CN104853186A CN201510309316.3A CN201510309316A CN104853186A CN 104853186 A CN104853186 A CN 104853186A CN 201510309316 A CN201510309316 A CN 201510309316A CN 104853186 A CN104853186 A CN 104853186A
- Authority
- CN
- China
- Prior art keywords
- video
- motion vector
- motion
- steganalysis
- parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The invention relates to an improved video steganalysis method based on a motion vector reply. According to the method provided by the invention, video compression parameters are classified into two types, i.e., obtainable parameters (parameters which can be directly obtained from a video decompression process) and unobtainable parameters (parameters which cannot be obtained from decompression), and an improved method is brought forward. The obtainable parameters are obtained from the decompression process for recompression, the unobtainable parameters are simulated by use of a motion prediction matching method during the recompression, and thus a video compression process is reconstructed. The method guarantees parameter consistency, and characteristics obtained by use of an optimized calibration method can effectively improve the accuracy of a detection result. The calibration-based video analysis method is optimized through reconstruction of a primary video compression process, and on the basis of a conventional motion vector reply based method, the validity of characteristics and the correct rate of steganalysis are improved.
Description
Technical field
The present invention relates to steganalysis (Steganalysis) method, particularly relate to a kind of steganalysis method based on calibration, and the application of the method in the video steganography detecting motion vector field, the method belongs to the sub-field of Information hiding in field of information security technology.
Background technology
In this day and age, high performance network and compress technique facilitate a large amount of transmission of multimedia file.The object of steganography is hidden in carrier by secret information, makes it not easily to be discovered, and steganalysis is then the technology whether existed for detecting classified information.As one of the most influential daily multimedia file, compressed video can provide sufficient redundant information available for steganography and steganalysis.
Motion vector, as the important parameter of video compression, has been widely used in steganography and has embedded.As Kutter (F.Jordan, M.Kutter, and T.Ebrahimi.Proposal of a watermarking technique for hiding data in compressed anddecompressed video, ISO/IEC Doc, JTC1/SC29/QWG11, Tech.Rep.M2281, Jul.1997.), Xu (C.Xu, X.Ping, and T.Zhang.Steganography in compressed video stream, in Proc.1st Int.Conf.InnovComput.Inf.Control, vol.1, pp.269 – 272, Sep.2006.), Aly (H.Aly, Data hiding in motion vectors ofcompressed video based on their associated prediction error, IEEE Trans.Inf.Forensics Security., vol.6, no.1, pp.14 – 18, Mar.2011.), Cao (Y.Cao, X.Zhao, D.Feng, and R.Sheng.Videosteganography with perturbed motion estimation, in Proc.13th Int.Conf.IH, vol.6958, no.1, pp.193 – 207, 2011.), Fang (D.Y.Fang and L.W.Chang.Data hiding for digital video with phase ofmotion vector, in Proc.IEEE Int.Symp.Circuits Syst., pp.1422 – 1425, etc. May.2006.) video steganography method proposed, make steganography method one of the video steganography method becoming main flow based on motion vector.
In order to detect the steganography based on motion vector, many steganalysis method are suggested in recent years, as Su (Y.Su, C.Zhang, and C.Zhang, A video steganalysis algorithm against motion-vector-based steganography, SignalProcess., vol.91, no.8, pp.1901 – 1909, 2011.), Ren (Y.Ren, L.Zhai, and L.Wang, Videosteganalysis based on subtractive probability of optimal matching feature, in Proc.2nd ACMWorkshop on IH & MMSec., pp.83 – 90, 2014.), Cao (Y.Cao, X.Zhao, and D.Feng, Videosteganalysis exploiting motion vector reversion-based features, Signal Process.Lett., vol.19, no.1, pp.35 – 38, Jan.2012.), Wang (K.Wang, H.Zhao, and H.Wang, Video steganalysis against motionvector-based steganography by adding or subtracting one motion vector value, " Trans.Inf.Forensics Security, vol.9, no.5, pp.741 – 751, Feb.2014.), wherein, the method effect that Cao and Wang proposes is best.Cao proposes a kind of method based on calibration and devises and replys (MVRB based on motion vector, motion vectorreversion based) feature, this feature is by calculating the motion vector (MV before and after weight contracting, motion vector) and the difference of corresponding residual absolute value sum (SAD, sum of absolute difference) obtain.Because this algorithm is realized by video calibration, the validity of MVRB feature depends on that whether the parameter of recodification is identical with the parameter of encoding before.Wang carries out plus-minus one to motion vector and operates (AoSO, adding-or-subtracting-one), and the difference of the optimum SAD obtained and actual SAD is used for feature calculation.When using MVRB feature to analyze, if when twice motion forecast method is different, this signature analysis effect acute exacerbation, in this case, the calculating due to AoSO feature does not relate to weight contracting so can ensure good effect.But, when using the uncertain information of motion prediction to revise motion vector, the local optimality of motion vector is kept, in this case, the analytical effect of AoSO feature is with regard to degradation, because embedding operation changes the information of motion prediction, MVRB feature can have good analytical effect.Therefore, on the basis of existing MVRB feature, reappear original compression process, the feature based on motion vector reply proposing a kind of improvement is significant.
Summary of the invention
The process that the object of the invention is by rebuilding video compression first optimizes the methods of video analyses based on calibration, on original basis based on motion vector answering method, improves the validity of feature and the accuracy of steganalysis.
Calibration is a kind of typical method of image induct, for compressed video, calibrate by by video compression to spatial domain, then by video when without carry out when messages embedding weight contracting realize.In order to detect the steganography method based on motion vector, inter macroblocks (MB, macroblock) and the corresponding vector that runs will by as analytic targets.As shown in Figure 1, in normal compression, in order to encoding current macroblock MB
c, the similar macro blocks MB before in coded frame
ras a reference searched, the SAD between two macro blocks is used for weighing prediction residual.Embed message in mv after, it becomes points to MB
r'mv'.Prediction residual is transmitted after dct transform, quantification and entropy code, again obtains prediction residual in decoding end through entropy decoding, inverse quantization and inverse DCT conversion.Original analytical method of replying based on motion vector comprises two steps: first decompressing video rebuild macro block
then video is carried out weight contracting, in this process,
will as reference macro block.Therefore motion vector mv can be obtained
*with corresponding SAD, their value is all close to value when normal compression.Therefore for steganography video, motion vector and the sad value more non-steganography video variance of weight contracting front and back are larger.
The prerequisite of the method is the macro block searched in weight contracting is the macro block searched in first-time compression just.But in practical operation, this condition is difficult to ensure, content of the present invention is each compression parameters being obtained first-time compression by the method for parameter acquiring and coupling, for the video weight contracting in calibrating.
The technical solution adopted in the present invention mainly comprises the following steps (if no special instructions, following steps perform by the software and hardware of computer and electronic equipment), Fig. 2 is the video steganalysis flow chart of replying based on motion vector that the present invention improves, and comprises the following steps:
(1) original video and steganography video set is prepared.User carries out the unified process of compression parameters (as size, length, resolution etc.) to original video as required, obtains one or more original video collection.Based on one group of original video collection, steganographic algorithm to be analyzed is adopted to generate corresponding one group of steganography video set.
(2) the MVRB steganalysis feature extraction improved.Its idiographic flow is as follows:
A) video is carried out decompress(ion), collect available (effectively) compression parameters in this process and be used for weight contracting, record the motion vector of each macro block for coupling afterwards and feature calculation, and the SAD recording each macro block is for feature calculation;
B) the some original videos obtained in step (1) are used, each method for searching motion in traversal video compression obtains motion vector, similitude statistics is carried out to the motion vector that each method for searching motion obtains, selects method for searching motion that motion vector difference is larger representatively for matching operation afterwards;
Video is carried out weight contracting by the parameter in c) using step a), traversal is the middle representational method for searching motion selected b), by the motion vector obtained with a) in the motion vector of corresponding macro block that obtains compare, the motion vector that coupling is the most approximate is used for macroblock motion prediction, and records the motion vector of each macro block and SAD for feature calculation;
D) motion vector in using step a) and c) and SAD information carry out feature calculation, thus extract video steganalysis feature.
(3) training of steganalysis grader and configuration.By characteristic vector input SVMs (Support Vector Machine, the SVM) grader extracted from original video collection and steganography video set in step (2), grader is trained, generates steganalysis grader.
(4) video to be measured is analyzed.Receive video to be measured, first use the step in (2) to carry out feature extraction to this video, then analyze in the feature of acquisition input steganalysis grader, repeatedly, average result is finally differentiated as foundation.
The beneficial effect of steganalysis method to video steganalysis field of above-mentioned improvement is the accuracy rate that improve video steganalysis.In original analytical method, fully cannot ensure the consistency of second-compressed parameter, and then the validity of feature is not high.Video compression parameter is classified as two classes by the present invention, can obtain parameter (parameter that directly can obtain from video compression process) and non-availability parameter (parameter that namely cannot obtain from decompress(ion)), and propose the method for improvement.From decompression procedure, obtain available parameter contract for weight, use the matching process simulation non-availability parameter of motion prediction when weight contracting, thus reconstruction video compression process.Compared with former feature, this method ensure that the consistency of parameter, the feature using the calibration steps optimized to obtain effectively can improve the accuracy of testing result.
Accompanying drawing explanation
Fig. 1 be based on motion vector steganography and original based on motion vector reply video steganalysis schematic diagram;
Fig. 2 is the video steganalysis flow chart of replying based on motion vector that the present invention improves;
Fig. 3 is the feature extraction flow chart of replying based on motion vector that the present invention improves;
Fig. 4 is the motion vector difference rate comparison diagram of the 8 kinds of searching methods using EPZS;
Fig. 5 is the motion vector difference rate comparison diagram of the 8 kinds of searching methods using MID;
Fig. 6 is the motion vector difference rate comparison diagram of 8 kinds of searching methods of EPZS and MID;
Fig. 7 is that the MVRB that the present invention improves contrasts schematic diagram to the verification and measurement ratio of the steganography method (Tar1 ~ Tar5) of Xu from former MVRB, AoSO under different embedding rate;
Fig. 8 is that the present invention improves MVRB and under different embedding rate, contrasts schematic diagram from former MVRB, AoSO to the verification and measurement ratio of the steganography method (Tar6 ~ Tar10) of Aly.
Embodiment
Below in conjunction with the drawings and specific embodiments, the inventive method is further described.
The present embodiment is to the analysis that the steganography based on motion vector is carried out under MPEG4 video encoding standard, and it is only the application of method in mpeg 4 standard of replying based on motion vector of the improvement that the present invention proposes, and can absolutely prove the effect of the method.But what the present invention proposed is a general framework, and except the present embodiment, the method can be applicable to the steganalysis under other video compression standards.Therefore based on other embodiments that framework of the present invention proposes, all belong to protection scope of the present invention.
As everyone knows, video compression parameter, comprise bit rate, frame per second, resolution, frame number, image sets (GOP, group ofpicture), compression parameters (QP, quantification parameter), motion forecast method, macro-block partition mode etc., directly determine the effect of compression.Although some can obtain parameter and can obtain from decompression procedure, the still non-availability of the key parameter in motion prediction.Therefore, in order to improve the validity of MVRB feature, the present invention improves video calibration from two aspects, and first aspect from decompression procedure, obtains available parameter contract for weight, and second aspect is the matching process simulation non-availability parameter using motion prediction when weight contracting.
The improvement that the present invention realizes on MPEG4 based on motion vector reply embodiment, its method mainly comprises the following steps:
(1) original video and steganography video set is prepared.
The standard C IF sequence of 25 4:2:0YUV forms is used to the experiment of this embodiment.The frame per second of video sequence is 25fps, and mean bit rate is 4Mbps.Because video is the different length that 100 frames change to 2000 frames, therefore each sequence is divided into the subsequence of 100 frame lengths, and the sum of subsequence reaches 120.The steganography method of the Xu mentioned in background technology and the method for Aly will be used to preparation steganography video set, in order to take into full account that motion forecast method is on the impact embedding effect, full search is used respectively in telescopiny, hex search, small diamond search, funny diamond search, umh search searching method, the corresponding steganography method of Xu represents with Tar1 to Tar5 respectively, and the corresponding steganography method of Aly represents with Tar6 and Tar10 respectively.Motion vector ratio (the Corrupted MV Ratio of amendment, CMVR) ratio of the number of motion vector and the total motion vector be modified in every frame is represented, in the experiment of this embodiment, CMVR is used to Metric Embedding intensity, and use 0.2 respectively, the CMVR value of 0.1,0.05 embeds original video.
(2) the MVRB steganalysis feature extraction improved, its idiographic flow following (as shown in Figure 3):
A) video is carried out decompress(ion), collect effective compression parameters and record motion vector and SAD information.
During video decode, compressed video obtains residual error (PE, Prediction Error) and motion vector through entropy decoding, inverse quantization, iDCT conversion, and utilizes the macroblock decode in reference frame to obtain current macro by motion compensation.Available compression parameter can be collected contract for weight in decode procedure.
Important available parameter can be roughly divided into three classes according to their effect: the first kind is relevant to the fundamental characteristics of video, comprises resolution, frame number and frame per second; Equations of The Second Kind is relevant with video quality, as compression parameters and bit rate; Last class determines the position of macro block, as image sets and macroblock partitions mode.In order to ensure the consistency of above important parameter in weight compression process, collect above video compression parameter in this step.In addition, the motion vector of each macro block and SAD information is recorded for coupling afterwards and feature calculation.
B) carry out similitude statistics to each method for searching motion, the representative searching method that selection differences is larger is used for coupling.
Because the motion vector obtained by different motion searching method has bigger difference in statistical property, in order to analyze the otherness of motion vector further, use the 8 kinds of different motion searching methods (comprising full search, hex search, small diamond search, various diamond search, 12s search, funny diamond, sad diamond and umhsearch) in MPEG4 to test 10 original video sequences obtained in (1), represent by numeral 1 to 8 in Fig. 4, Fig. 5, Fig. 6 respectively.In addition, also contemplating 2 kinds of different methods obtaining initial search point, is EPZS (Enhanced PredictiveZonal Search) and MID (Median of left, top, top-right MV values) respectively.As shown in figures 4-6, although any two kinds of searching methods all have difference more or less, still Some features can be drawn by statistics.Can be obtained by Fig. 4 and Fig. 5, when using full search, hex search and various diamond search, the ratio of the different motion vector of they and additive method is all higher than 0.3, and the result of hex search is substantially identical with the result of various diamond search.In addition, the ratio of the different motion vector between other method for searching motion is all lower than 0.15, and wherein small diamond is default methods.Therefore, full search, hex search and small diamond search can representatively method for searching motion.Fig. 6 compares the difference of 8 kinds of searching methods of EPZS and MID, and because the value on diagonal is all lower than 0.15, the impact that the impact that the different start point search method of deducibility is brought brings than different motion searching method is little.
C) video weight contracted, the most similar motion forecast method of coupling also records motion vector and SAD information.
The flow process of video weight contracting is carry out motion prediction to each macro block of present frame, search and obtain reference block in reference frame, and obtain residual sum motion vector.Residual error through dct transform, quantize after, entropy code process, motion vector, after entropy code process, obtains compressed video and transmits.After current block encoded, it still may as with reference to block, therefore need to obtain rebuilding macro block to current block decoding (inverse quantization, iDCT convert) in compression process.
Based on the statistical analysis in b), the present invention is that the analytical method of replying based on motion vector proposes a kind of motion forecast method matching strategy.In the motion prediction of weight contracting, traversal full search, hex search and small diamond search, using the motion vector that obtained by the above method candidate item as coupling.In the matching process, video offset distance (SD, shiftdistance) is used to measure that whether this motion vector obtain to decoding is similar.
SD(mv,mv
i)＝|h-h
i|+|v-v
i|
Wherein mv
i=(h
i, v
i), i=1,2,3 is by a kind of motion vector obtained in above-mentioned three kinds of searching methods, and mv=(h, v) is the motion vector obtained in decoding.The mv be similar to the most with the corresponding motion vector that obtains of decoding
iusing the motion vector as use during weight contracting.In addition, the motion vector of each macro block and SAD information be recorded for feature calculation afterwards.
D) motion vector in using step a) and c) and SAD information carry out feature calculation.
Use the computational methods of original MVRB feature, use motion vector and SAD information to carry out feature calculation.
(3) training and configuration steganalysis grader.
In embodiments of the present invention, use the SVMs of Gaussian kernel to classify, select arbitrarily the video sequence of 60% to use step (2) to extract feature, feature is inputted grader and trains, remaining subsequence is used to test.The parameter configuration of SVM classifier can be determined by cross validation, namely travels through all optional parameters combinations, selects the parameter that wherein repeatedly the average correct classification rate of cross validation is the highest to be configured grader.The concrete grammar of cross validation is, by original and set of eigenvectors that is steganography video set, random division is training set and test set by a certain percentage, and use grader to carry out judgement of training and classify, the accuracy of grader to test set is the result of cross validation.
(4) video to be measured is detected.
The steganalysis grader of having trained is used to analyze remaining 40% video sequence, first the step in (2) is used to carry out feature extraction to this video, then test in the feature of acquisition input steganalysis grader, repeatedly, average detected result is finally differentiated as foundation.
In order to the method for more different steganalysis, the AoSO feature (as described in the background art) that the MVRB characteristic sum Wang that the present embodiment uses Cao to propose proposes contrasts for analytical effect.The Detection results of different analytical method is listed in Table 1, clearly illustrates AoSO analytical characteristic, the verification and measurement ratio contrast of MVRB feature that original MVRB characteristic sum improves in Fig. 7 and Fig. 8.
Table 1. is when different searching method and embedding rate, and AoSO, original MVRB and improvement MVRB feature are to the testing result of Xu and Aly steganography method
Original MVRB feature effect in three kinds of analytical characteristics is the poorest, AoSO feature is better than original MVRB feature, but when using full search in the embedding grammar at Aly, the verification and measurement ratio of this feature acutely declines, because maintain local optimality at this better off.In all cases, the MVRB feature of raising is all better than original MVRB and AoSO feature, and its average detected rate is up to 0.895.Fig. 7 and Fig. 8 also illustrates along with embedding rate reduces, and each accuracy rate also declines thereupon.When analysis embedding rate is the Tar6 of 0.05, the accuracy rate of the MVRB of raising is still higher than 0.75.
From the embodiment in above embodiment, the video steganalysis method based on motion vector reply that the present invention improves can significantly improve the validity of feature and the accuracy rate of analysis, detects efficiently the video steganography based on motion vector.
Above embodiment is only in order to illustrate technical scheme of the present invention but not to be limited; those of ordinary skill in the art can modify to technical scheme of the present invention or equivalent replacement; and not departing from the spirit and scope of the present invention, protection scope of the present invention should be as the criterion with described in claims.
Claims (7)
1. the video steganalysis method of replying based on motion vector improved, its step comprises:
1) prepare original video collection, and generate corresponding steganography video set based on original video centralized procurement steganographic algorithm;
2) to step 1) original video that obtains and steganography video carry out decompress(ion), and in decompression procedure, collect available compression parameters, and record each macroblock motion vectors and SAD information;
3) to step 1) some original videos of obtaining, motion vector is obtained by each method for searching motion in traversal video compression, and similitude statistics is carried out to the motion vector that each method for searching motion obtains, select the method for searching motion representatively method for searching motion that the otherness of motion vector is larger;
4) step 2 is used) described available compression parameters carries out weight contracting to original video and steganography video, using step 3) described representational motion search mode obtains each motion vector, by itself and step 2) motion vector of corresponding macro block that obtains compares, the macroblock motion prediction that the motion vector that coupling is similar to the most contracts for weight, and record each macroblock motion vectors and SAD information;
5) use step 2) and 4) in motion vector and SAD information carry out feature calculation, extract video steganalysis feature;
6) feature extracted from original video collection and steganography video set input grader is trained, generate steganalysis grader;
7) step 2 is used to video to be measured) ~ 4) described in method carry out feature extraction, the feature of the video to be measured obtained is inputted in described steganalysis grader and analyzes, to differentiate whether video to be measured exists secret information.
2. the method for claim 1, is characterized in that: step 1) unified by carrying out compression parameters to original video, obtain one or more original video collection, described compression parameters comprises size, length, resolution.
3. the method for claim 1, is characterized in that: step 2) described effective compression parameters is divided three classes: the first kind is relevant to the fundamental characteristics of video, comprises resolution, frame number and frame per second; Equations of The Second Kind is relevant with video quality, comprises quantization parameter and bit rate; Last class determines the position of macro block, comprises image sets and macroblock partitions mode.
4. the method for claim 1, it is characterized in that: step 4) in, by video offset distance SD measure the motion vector that obtains of described representational method for searching motion whether to step 2) motion vector that obtains of decompress(ion) is similar, its computing formula is:
SD(mv,mv
i)＝|h-h
i|+|v-v
i|，
Wherein mv
i=(h
i, v
i), i=1,2,3 is a kind of motion vectors obtained by described representational method for searching motion, and mv=(h, v) is the motion vector obtained in decompress(ion), the mv be similar to the most with mv
iusing the motion vector as use during weight contracting.
5. the method for claim 1, it is characterized in that: step 6) use the SVMs of Gaussian kernel as steganalysis grader, the parameter configuration of support vector machine classifier is determined by cross validation, namely travel through all optional parameters combinations, select the parameter that wherein repeatedly the average correct classification rate of cross validation is the highest to be configured grader.
6. method as claimed in claim 5, it is characterized in that: described cross validation is the set of eigenvectors by original video and steganography video set, random division is training set and test set by a certain percentage, use grader to carry out judgement of training and classify, the accuracy of grader to test set is the result of cross validation.
7. the method for claim 1, is characterized in that: step 7) repeatedly, average result is finally differentiated as foundation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510309316.3A CN104853186B (en) | 2015-06-08 | 2015-06-08 | A kind of improved video steganalysis method that is replied based on motion vector |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510309316.3A CN104853186B (en) | 2015-06-08 | 2015-06-08 | A kind of improved video steganalysis method that is replied based on motion vector |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104853186A true CN104853186A (en) | 2015-08-19 |
| CN104853186B CN104853186B (en) | 2017-03-08 |
Family
ID=53852504
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510309316.3A Expired - Fee Related CN104853186B (en) | 2015-06-08 | 2015-06-08 | A kind of improved video steganalysis method that is replied based on motion vector |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104853186B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105915916A (en) * | 2016-05-12 | 2016-08-31 | 中国科学院信息工程研究所 | Video steganalysis method based on motion vector rate-distortion performance estimation |
| CN105979269A (en) * | 2016-06-03 | 2016-09-28 | 中国科学院信息工程研究所 | Motion vector domain video steganography method based on novel embedding cost |
| CN106101713A (en) * | 2016-07-06 | 2016-11-09 | 武汉大学 | A kind of video steganalysis method based on the calibration of window optimum |
| CN106131553A (en) * | 2016-07-04 | 2016-11-16 | 武汉大学 | A kind of video steganalysis method based on motion vector residual error dependency |
| CN107197297A (en) * | 2017-06-14 | 2017-09-22 | 中国科学院信息工程研究所 | A kind of video steganalysis method of the detection based on DCT coefficient steganography |
| CN107613303A (en) * | 2017-09-08 | 2018-01-19 | 中国科学院信息工程研究所 | Video steganalysis method and device based on quantization parameter singular values standard form |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
| CN103281473A (en) * | 2013-06-09 | 2013-09-04 | 中国科学院自动化研究所 | General video steganalysis method based on video pixel space-time relevance |
| US20130230099A1 (en) * | 2004-07-30 | 2013-09-05 | Euclid Discoveries, Llc | Standards-compliant model-based video encoding and decoding |
| CN103338376A (en) * | 2013-07-02 | 2013-10-02 | 中国科学技术大学 | Video steganography method based on motion vector |
| CN103888773A (en) * | 2014-02-19 | 2014-06-25 | 南京邮电大学 | Video steganography analysis method based on mutual information and motion vectors |
-
2015
- 2015-06-08 CN CN201510309316.3A patent/CN104853186B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130230099A1 (en) * | 2004-07-30 | 2013-09-05 | Euclid Discoveries, Llc | Standards-compliant model-based video encoding and decoding |
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
| CN103281473A (en) * | 2013-06-09 | 2013-09-04 | 中国科学院自动化研究所 | General video steganalysis method based on video pixel space-time relevance |
| CN103338376A (en) * | 2013-07-02 | 2013-10-02 | 中国科学技术大学 | Video steganography method based on motion vector |
| CN103888773A (en) * | 2014-02-19 | 2014-06-25 | 南京邮电大学 | Video steganography analysis method based on mutual information and motion vectors |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105915916A (en) * | 2016-05-12 | 2016-08-31 | 中国科学院信息工程研究所 | Video steganalysis method based on motion vector rate-distortion performance estimation |
| CN105915916B (en) * | 2016-05-12 | 2019-02-22 | 中国科学院信息工程研究所 | Video steganalysis method based on the estimation of motion vector distortion performance |
| CN105979269A (en) * | 2016-06-03 | 2016-09-28 | 中国科学院信息工程研究所 | Motion vector domain video steganography method based on novel embedding cost |
| CN105979269B (en) * | 2016-06-03 | 2019-06-04 | 中国科学院信息工程研究所 | Motion vector field video steganography method based on novel insertion cost |
| CN106131553A (en) * | 2016-07-04 | 2016-11-16 | 武汉大学 | A kind of video steganalysis method based on motion vector residual error dependency |
| CN106101713A (en) * | 2016-07-06 | 2016-11-09 | 武汉大学 | A kind of video steganalysis method based on the calibration of window optimum |
| CN106101713B (en) * | 2016-07-06 | 2018-10-09 | 武汉大学 | A kind of video steganalysis method based on the optimal calibration of window |
| CN107197297A (en) * | 2017-06-14 | 2017-09-22 | 中国科学院信息工程研究所 | A kind of video steganalysis method of the detection based on DCT coefficient steganography |
| CN107197297B (en) * | 2017-06-14 | 2019-12-10 | 中国科学院信息工程研究所 | Video steganalysis method for detecting steganalysis based on DCT coefficient steganalysis |
| CN107613303A (en) * | 2017-09-08 | 2018-01-19 | 中国科学院信息工程研究所 | Video steganalysis method and device based on quantization parameter singular values standard form |
| CN107613303B (en) * | 2017-09-08 | 2019-10-22 | 中国科学院信息工程研究所 | Video steganalysis method, device, equipment and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104853186B (en) | 2017-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104853186A (en) | Improved video steganalysis method based on motion vector reply | |
| CN102137263B (en) | Distributed video coding and decoding methods based on classification of key frames of correlation noise model (CNM) | |
| Sun et al. | Exposing video forgeries by detecting MPEG double compression | |
| CN102724554B (en) | Scene-segmentation-based semantic watermark embedding method for video resource | |
| CN103988500A (en) | Video quality measurement | |
| WO2001057802A1 (en) | Method for describing motion activity in video | |
| CN101478691B (en) | Non-reference evaluation method for Motion Jpeg2000 video objective quality | |
| He et al. | Double compression detection based on local motion vector field analysis in static-background videos | |
| CN105933711B (en) | Neighborhood optimum probability video steganalysis method and system based on segmentation | |
| CN104837011A (en) | Content self-adaptive video steganalysis method | |
| CN105979269B (en) | Motion vector field video steganography method based on novel insertion cost | |
| CN105915916B (en) | Video steganalysis method based on the estimation of motion vector distortion performance | |
| CN103020138A (en) | Method and device for video retrieval | |
| CN103237209A (en) | H264 video watermarking method based on regional DCT (discrete cosine transform) coefficients | |
| CN103034993A (en) | Digital video transcode detection method | |
| Wang et al. | Motion vector reversion-based steganalysis revisited | |
| CN102547371B (en) | Secondary compression detection method based on H.264/AVC (Advanced Video Coding) video | |
| Ma et al. | Reduced reference video quality assessment based on spatial HVS mutual masking and temporal motion estimation | |
| CN110324634A (en) | It is a kind of to be embedded in the video steganography method that distortion is decomposed based on motion vector | |
| Jing et al. | Motion vector based information hiding algorithm for H. 264/AVC against motion vector steganalysis | |
| He et al. | A two-stage cascaded detection scheme for double HEVC compression based on temporal inconsistency | |
| CN108024114A (en) | A kind of lossless HEVC information concealing methods of large capacity based on the modification of flag bit parameter | |
| Liu et al. | A MPEG-2 video watermarking algorithm with compensation in bit stream | |
| Pankajakshan et al. | Detection of motion-incoherent components in video streams | |
| Ma et al. | Error sensitivity model based on spatial and temporal features |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170308 Termination date: 20170608 |