CN103384331A - Video inter-frame forgery detection method based on light stream consistency - Google Patents
Video inter-frame forgery detection method based on light stream consistency Download PDFInfo
- Publication number
- CN103384331A CN103384331A CN2013103074816A CN201310307481A CN103384331A CN 103384331 A CN103384331 A CN 103384331A CN 2013103074816 A CN2013103074816 A CN 2013103074816A CN 201310307481 A CN201310307481 A CN 201310307481A CN 103384331 A CN103384331 A CN 103384331A
- Authority
- CN
- China
- Prior art keywords
- light stream
- video
- frame
- mvsum
- layer
- 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.)
- Pending
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides a passive video inter-frame forgery detection method based on the light stream consistency in the field of video content safety. The detection method mainly comprises the steps of extracting light stream features of two adjacent frames in a surveillance video shot by a static camera, adding light stream absolute values in horizontal and vertical directions to generate two light stream sequences of the whole video in horizontal and vertical directions, analyzing the light stream sequences, and performing video inter-frame forgery detection for features of different inter-frame forgery modes. According to the detection method, through the light stream features which are high-robustness video features, five types of video inter-frame forgeries, including video frame insertion, frame deletion, repeated duplication of single frames, non-adjacent duplication of continuous frames and adjacent duplication of continuous frames can be detected effectively, and a forceful detection weapon is provided for the integrity and the authenticity of court testimony videos.
Description
Technical field
What the present invention relates to is a kind of method that video is distorted detection field, specifically altering detecting method between the conforming frame of video of a kind of optical flow-based.
Background technology
Growing along with Internet technology and digital equipment, digital video is dissolved in people's life widely, and an important evidence source that becomes gradually that court authenticates.Yet along with various video editing software functions are more and more perfect, a lot of videos are tampered and spread in network, cause video information no longer reliable.If the video that is tampered is provided evidence as court, be difficult to guarantee integrality and the authenticity of video.How to judge exactly whether a given video is tampered, become an important topic of information security field.
At present, common video is distorted can be divided into distorting in frame with interframe and is distorted dual mode.Distort in frame is mainly that the regional area in a frame of video (such as certain object) is modified and edited; It is that a plurality of frame of video to complete in video are inserted, delete and copy that interframe is distorted.
Video is distorted detection algorithm can be divided into two classes: video distorts the active detecting algorithm and video is distorted the passive detection algorithm.It is mainly to utilize the digital watermarking of embedding or the integrality of digital signature to judge whether video is tampered that video is distorted the active detecting algorithm, relies on the preliminary treatment to video.It is mainly that the features such as the texture that utilizes video itself, noise, motion vector judge whether video is tampered that video is distorted the passive detection algorithm, and because it need not priori, active detecting algorithm practicality is stronger relatively.
Find through the retrieval to prior art, Chinese patent literature CN102413327A, that open day 2012-04-11 put down in writing is a kind of " based on the video altering detecting method of the semi-fragile watermarking of compressed sensing ", this technology is for coded system and the DCT coefficient characteristics of I two field picture in the MPEG-2 video, extract the characteristics of image of I frame by the compressed sensing technology, generate and embed content authentication watermark and Watermarking for Integrity, carry out distorting between video integrality initial survey, frame of video and detect and the interior content authentication of frame of video.
This patent has proposed a kind of video based on watermark and has distorted the active detecting method, the present invention proposes between a kind of new frame of video and distort passive detection method, detect by extracting Optical-flow Feature in video that frame of video insertion, frame deletion, single frames multiple copies, multiframe are non-adjacent to be copied copy this video of five type adjacent with multiframe and distort mode, the method is very effective for the integrality and the authenticity that detect monitor video, is identifying that court's video provides evidence and have significant impact aspect authenticity.This as an emerging research direction, is still had very large development space.
Summary of the invention
The present invention proposes between the conforming frame of video of a kind of optical flow-based and distort passive detection method, the monitor video that the method is taken for static camera, by extracting the Optical-flow Feature between consecutive frame in video, whether unanimously analyze in whole video Optical-flow Feature, judge whether video is tampered, and the positioning tampering type.
Altering detecting method between frame of video conforming according to optical flow-based provided by the invention comprises the steps:
Step 1: read in video to be detected, be decoded as independently sequence of frames of video;
Step 2: the sequence of frames of video that decoding is obtained, extract the light stream vector between every two adjacent video frames, all pixel place's horizontal directions of the light stream vector that generates and the absolute value of vertical direction are sued for peace respectively, obtain the horizontal direction of frame and vertical direction two light stream absolute values and, with the light stream absolute value of all frames and according to time sequence generate the light stream sequence of whole video level direction and vertical direction;
Step 3: analyze the light stream sequence signature, according to distorting the characteristics of pattern between different frame of video, utilize light stream rate of change (Optical Flow Change Rate) to judge whether video is tampered, if be tampered, determine to distort type;
Step 4: copy the video that copy adjacent with successive frame to frame insertion, single frames multiple copies, successive frame are non-adjacent, the frame of video that deletion is inserted into or copies, thus carry out video recovery.
Preferably, extract the light stream vector between described two adjacent video frames, specifically comprise the steps:
Step 2.1: obtain two adjacent video frames in video for decompress(ion), each frame of video is carried out four iteration frame compressions, generate 5 layers of pyramid that comprise original video frame, in pyramid, every one deck is to be obtained through compression by its lower one deck, wherein, pyramidal bottom, namely layer 5, be original video frame;
Step 2.2: from the pyramid top to bottom, calculate simultaneously the motion vector of horizontal direction and vertical direction between two video frame images of corresponding identical level between two adjacent video frames, wherein, pyramidal top is pyramidal ground floor;
Step 2.3: 5 layer motion vectors that obtain by step 2.2 are denoted as from top to bottom successively: layer (1), layer (2), layer (3), layer (4) and layer (5);
At first, generate the motion vector figure of a sky, its size is consistent with layer (1), and the light stream value of all pixels of the inside is 0, is denoted as ExpMV (0);
ExpMV (0) and layer (1) addition are obtained MVSum (1), then MVSum (1) is carried out smoothing processing (level and smooth additional be first to compress, rear expansion); MVSum (1) after level and smooth is denoted as MVSmooth (1), then MVSmooth (1) is expanded obtaining ExpMV (1);
ExpMV (1) and layer (2) addition are obtained MVSum (2), then MVSum (2) is carried out smoothing processing (level and smooth additional be first to compress, rear expansion); MVSum (2) after level and smooth is denoted as MVSmooth (2), then MVSmooth (2) is expanded obtaining ExpMV (2);
ExpMV (2) and layer (3) addition are obtained MVSum (3), then MVSum (3) is carried out smoothing processing (level and smooth additional be first to compress, rear expansion); MVSum (3) after level and smooth is denoted as MVSmooth (3), then MVSmooth (3) is expanded obtaining ExpMV (3);
ExpMV (3) and layer (4) addition are obtained MVSum (4), then MVSum (4) is carried out smoothing processing (level and smooth additional be first to compress, rear expansion); MVSum (4) after level and smooth is denoted as MVSmooth (4), then MVSmooth (4) is expanded obtaining ExpMV (4);
ExpMV (4) and layer (5) addition are obtained MVSum (5), then MVSum (5) is carried out smoothing processing (level and smooth additional be first to compress, rear expansion); MVSum (5) after level and smooth is denoted as MVSmooth (5), and MVSmooth (5) is two light stream vectors between adjacent video frames.
Preferably, in described step 2.1, described compression specifically comprises the steps:
Step 2.1.1: spatial sampling: the even number line and the even column that extract video frame images form a new images;
Step 2.1.2: the window that is N with a size in the horizontal direction of new images carries out convolution algorithm, carries out smoothing processing with the pixel to horizontal direction, and N is natural number;
Step 2.1.3: the window that is N with a size in the vertical direction of the image that obtains by step 2.1.2 carries out convolution algorithm, and the pixel of vertical direction is carried out smoothing processing, obtains final compressed image.
Preferably, described expansion specifically comprises the steps:
Steps A: spatial spread: horizontal direction and vertical direction at image are inserted blank line and blank column;
Step B: the window that is N with a size in the horizontal direction of new images carries out convolution algorithm, and the pixel of horizontal direction is carried out smoothing processing;
Step C: the window that the vertical direction of the image that forms in the B step is N with a size carries out convolution algorithm, and the pixel of vertical direction is carried out smoothing processing, obtains final expanded images.
Preferably, the light stream rate of change OFCR between i frame and i+1 frame
(i, i+1)(x) be:
OFCR
(i,i+1)(x)=2×S
(i,i+1)(x)/(S
(i-1,i)(x)+S
(i+1,i+2)(x))
Wherein, S
(i, i+1)(x) be the horizontal direction between i frame and i+1 frame the light stream absolute value and, S
(i-1, i)(x) be the horizontal direction between i-1 frame and i frame the light stream absolute value and, S
(i+1, i+2)(x) be the horizontal direction between i+1 frame and i+2 frame the light stream absolute value and.
Preferably, if there are two very lofty light stream pulses in the light stream sequence, and there is not on all four light stream sequence in video, do not exist continuous light stream value to equate and near 0 situation, distort and be defined as frame of video and insert and distort;
If have a light stream pulse in the light stream sequence, and do not have on all four light stream sequence in video, do not exist continuous light stream value to equate and near 0 situation, determining to distort type is that the frame of video deletion is distorted;
If there is not the light stream pulse in the light stream sequence, but exists continuous light stream value to equate and near 0 light stream sequence, do not have on all four light stream sequence, determining to distort type is that video single frames multiple copies is distorted;
If there are two light stream pulses in the light stream sequence, do not exist continuous light stream to equate in video and near 0 situation, but have other one section light stream subsequence in video, and the light stream subsequence between two pulses is in full accord, to determine to distort type be that the video successive frame is non-adjacent copies;
If there is a light stream pulse in the light stream sequence, do not exist continuous light stream value to equate in video and near 0 light stream sequence, but before and after pulse, two sections light stream subsequences are in full accord, determining to distort type is that adjacent the copying of video successive frame distorted.
Preferably, N=5, wherein,
Determine that the method that exists frame to insert the light stream pulse of distorting type of type is to have two points, its light stream rate of change is greater than threshold value threshold1=20;
The method of determining to exist the light stream pulse of distorting type of frame deletion type is to have a point, and its light stream rate of change is greater than threshold value threshold2=2;
Determine to exist single frames repeatedly the method for the light stream pulse of distorting type of type be to have a plurality of continuous points, its light stream rate of change is 1;
The method of determining to exist the light stream pulse of distorting type of the non-adjacent copy type of successive frame is to have two points, and its light stream rate of change is greater than threshold value threshold2=2;
The method of determining to exist the light stream pulse of distorting type of the adjacent copy type of successive frame is to have a point, and its light stream rate of change is greater than threshold value threshold2=2.
Distort the passive detection algorithm between the frame of video that the present invention proposes, by extracting this video features of light stream, detect that frame of video insertion, frame deletion, single frames multiple copies, successive frame are non-adjacent to be copied copy between this five kind dissimilar frame of video adjacent with successive frame and distort.Optical-flow Feature is one can effectively accumulate the impact of distorting video for distorting highstrung feature between frame of video, and Optical-flow Feature extraction relative noise feature is simple and easy, can effectively improve the time efficiency of algorithm.
Description of drawings
By reading the detailed description of non-limiting example being done with reference to the following drawings, it is more obvious that other features, objects and advantages of the present invention will become:
Fig. 1 distorts the overall flow figure of passive detection method between frame of video of the present invention;
Fig. 2 is that between frame of video of the present invention, Optical-flow Feature extracts flow chart.
Embodiment
The present invention is described in detail below in conjunction with specific embodiment.Following examples will help those skilled in the art further to understand the present invention, but not limit in any form the present invention.Should be pointed out that to those skilled in the art, without departing from the inventive concept of the premise, can also make some distortion and improvement.These all belong to protection scope of the present invention.
The present invention is achieved by the following technical solutions, at first two adjacent frame of video are carried out layer-by-layer contraction and obtain five layers of pyramid, extract the motion vector of every layer, successively motion vector expanded, sued for peace from top to bottom and smoothly obtain two Optical-flow Features between frame of video, respectively the horizontal and vertical direction being carried out light stream absolute value summation statistics.According to said method extract the Optical-flow Feature between all consecutive frames of whole video, each video is generated horizontal direction and two light stream sequences of vertical direction.Then the light stream sequence is analyzed, for the feature of distorting mode between different frame, judge whether video is original video, if not, judge that again it distorts type, and the distorting of some type (frame of video insertion, single frames multiple copies, successive frame be non-adjacent to be copied, successive frame is adjacent copies) carried out video recovery.
Distort the overall flow of passive detection algorithm between frame of video as shown in Figure 1, its concrete steps are:
The first step: video decode, concrete steps are:
For video to be detected, utilize in matlab the mmreader function to decode, obtain independently sequence of frames of video, be convenient to extract in following step the Optical-flow Feature between every two adjacent video frames.
Second step: Optical-flow Feature extracts
To the sequence of frames of video that decoding obtains, extract the Optical-flow Feature between every two consecutive frames:
(1). for each frame of video, it is carried out four iteration frame compressions, generate 5 layers of pyramid that comprise original video frame, in pyramid, every one deck is to be obtained through compression by its lower one deck, and wherein, pyramidal bottom is original video frame.Compression step is as follows:
Steps A: spatial sampling: the even number line and the even column that extract image form a new image.
Step B: be that 5 window carries out convolution algorithm in the horizontal direction of new images with a size, carry out smoothing processing with the pixel to horizontal direction.
Step C: the vertical direction of the image that forms in the B step is that 5 window carries out convolution algorithm with a size, and the pixel of vertical direction is carried out smoothing processing, obtains final compressed image.
(2). from the pyramid top to bottom (being original video frame), calculate simultaneously the motion vector of horizontal direction and vertical direction between two images of corresponding identical level, totally 5 layer motion vectors.
Descriptor has:
Layer (1): the motion vector between two condensed frames of the top layer (being ground floor) of five layers of pyramid of two frame of video
Layer (2): the motion vector between two condensed frames of five layers of pyramid second layer of two frame of video
Layer (3): the motion vector between two condensed frames that five layers of pyramid of two frame of video are the 3rd layer
Layer (4): the motion vector between two condensed frames that five layers of pyramid of two frame of video are the 4th layer
Layer (5): the motion vector (3) between five layers of pyramid layer 5 of two frame of video (two unpressed frames). above-mentioned five layer motion vectors are expanded from top to bottom successively, with lower floor add and, then to add and after motion vector carry out smoothly.The motion vector that last one deck obtains after level and smooth is light stream vector.
Descriptor has:
ExpMV (0): the value of an artificial generation is 0 motion vector entirely, and size is pyramid ground floor size
MVSum (1): ExpMV (0) and layer (1) add and result
MVSmooth (1): MVSum (1) is through the result after smooth operation, and smooth operation is expanded for first compressing afterwards
Result after ExpMV (1): MVSmooth (1) expansion, size are pyramid second layer size
MVSum (2): ExpMV (1) and layer (2) add and result
MVSmooth (2): MVSum (2) is through the result after smooth operation, and smooth operation is expanded for first compressing afterwards
Result after ExpMV (2): MVSmooth (2) expansion, size is the 3rd layer of size of pyramid
MVSum (3): ExpMV (2) and layer (3) add and result
MVSmooth (3): MVSum (3) is through the result after smooth operation, and smooth operation is expanded for first compressing afterwards
Result after ExpMV (3): MVSmooth (3) expansion, size is the 4th layer of size of pyramid
MVSum (4): ExpMV (3) and layer (4) add and result
MVSmooth (4): MVSum (4) is through the result after smooth operation, and smooth operation is expanded for first compressing afterwards
Result after ExpMV (4): MVSmooth (4) expansion, size is pyramid layer 5 (primitive frame) size
MVSum (5): ExpMV (4) and layer (5) add and result
MVSmooth (5): MVSum (5) is through the result after smooth operation, and smooth operation is expanded for first compressing afterwards, and this descriptor is the light stream vector between two final frame of video.
Concrete spread step is as follows:
Steps A: spatial spread: horizontal direction and vertical direction at image are inserted blank line and blank column.
Step B: be that 5 window carries out convolution algorithm in the horizontal direction of new images with a size, the pixel of horizontal direction is carried out smoothing processing.
Step C: the vertical direction of the image that forms in the B step is that 5 window carries out convolution algorithm with a size, and the pixel of vertical direction is carried out smoothing processing, obtains final expanded images.
The 3rd step, the summation of light stream vector horizontal and vertical direction absolute value
All pixel place's horizontal directions of the light stream vector that generates and the absolute value of vertical direction are sued for peace respectively, obtain horizontal direction and two light stream absolute values of vertical direction and.
For given two frame of video i and i+1, the formula calculated level direction below utilizing and the light stream absolute value of all pixels of vertical direction and:
OFX wherein
(i, i+1)(w, h) refers to the light stream absolute value of pixel (w, h) position, and width is the frame of video width, and height is the frame of video height, S
(i, i+1)(x) be the final horizontal direction that generates the light stream absolute value and.With the X in formula change into Y can obtain the corresponding light stream absolute value of vertical direction and.
The 4th step: generate the light stream sequence
Travel through whole video, extract the Optical-flow Feature between all adjacent two frames, calculate horizontal direction and vertical direction between two frames the light stream absolute value and, with all frame light stream absolute values and according to time sequence obtain horizontal direction and two light stream sequences of vertical direction of whole video.
The 5th step: analyze the light stream consistency, type is distorted in judgement
Calculate the light stream rate of change of having a few in the light stream sequence, according to distorting the feature of mode between different frame of video, video light stream sequence is analyzed, judge whether to be tampered, if be tampered, judge that it distorts type.
Light stream rate of change OFCR between i frame and i+1 frame video
(i, i+1)(x) account form is as follows:
OFCR
(i,i+1)(x)=2×S
(i,i+1)(x)/(S
(i-1,i)(x)+S
(i+1,i+2)(x))
S wherein
(i, i+1)(x) be the horizontal direction (directions X) between i frame and i+1 frame the light stream absolute value and, S
(i-1, i)(x) be the horizontal direction between i-1 frame and i frame the light stream absolute value and, S
(i+1, i+2)(x) be the horizontal direction between i+1 frame and i+2 frame the light stream absolute value and.OPCR is the abbreviation (Optical Flow Change Rate) of light stream rate of change.
Concrete light stream consistency detecting method is, calculates the ratio between each light stream and its former and later two light stream averages, is designated as the light stream rate of change.Then light stream rate of change and two fixed thresholds are compared.
(1). there is no the original video through distorting: all light stream rates of change are lower than fixed threshold threshold2=2.
(2). frame of video is inserted and distorted: have two points, then the light stream rate of change relatively, does not exist on all four light stream sequence with other light stream subsequence traversal in the light stream subsequence between two points and light stream sequence higher than fixed threshold threshold1=20.
(3). the frame of video deletion is distorted: have a point, the light stream rate of change is higher than fixed threshold threshold2=2, and relatively greater than two sections light stream subsequences before and after the point of threshold2, there is not on all four light stream subsequence with it in the light stream rate of change.
(4). video single frames multiple copies is distorted: have a plurality of continuous points, its light stream rate of change is 1.
(5). the video successive frame is non-adjacent to be copied: have a point, the light stream rate of change is higher than fixed threshold threshold2=2, and relatively greater than two sections light stream subsequences before and after the point of threshold2, there is on all four light stream subsequence in the light stream rate of change.
(6). adjacent the copying of video successive frame distorted: have two points, the light stream rate of change is higher than fixed threshold threshold2=2, then relatively, there is on all four light stream subsequence with it in other light stream subsequence traversal in the light stream sequence between two points and light stream sequence.
The 6th step, to frame of video insertion, single frames multiple copies, successive frame non-adjacently copy, the adjacent mode of distorting that copies this Four types of successive frame carries out video recovery.
The input of processing: be judged as in the 5th step that frame insertions, single frames multiple copies, successive frame non-adjacently copy, the light stream sequence of the adjacent video that copies of successive frame and correspondence.
Carry out the step of frame of video tamper recovery:
(1) video is decoded, obtain independently frame sequence.
(2) according to the Optical-flow Feature in the light stream sequence, respectively four kinds of modes of distorting are carried out following frame deletion:
A. frame of video is inserted and to be distorted: with the frame deletion between two light stream pulses in the light stream sequence.
B. video single frames multiple copies: with in the light stream sequence continuous that equate, near frame of video deletion corresponding to 0 light stream value.
C. the video successive frame is non-adjacent copies: with the deletion of the frame of video between two light stream pulses in the light stream sequence.
D. the video successive frame is adjacent copies: the light stream sequence in traversal light stream sequence before and after individual pulse, obtain two on all four light stream sequences the longest, and deletion is frame of video corresponding to any one sequence wherein.
(3) video after frame deletion is carried out compressed encoding, revert to former video.
Above specific embodiments of the invention are described.It will be appreciated that, the present invention is not limited to above-mentioned specific implementations, and those skilled in the art can make various distortion or modification within the scope of the claims, and this does not affect flesh and blood of the present invention.
Claims (7)
1. altering detecting method between the conforming frame of video of optical flow-based, is characterized in that, comprises the steps:
Step 1: read in video to be detected, be decoded as independently sequence of frames of video;
Step 2: the sequence of frames of video that decoding is obtained, extract the light stream vector between every two adjacent video frames, all pixel place's horizontal directions of the light stream vector that generates and the absolute value of vertical direction are sued for peace respectively, obtain the horizontal direction of frame and vertical direction two light stream absolute values and, with the light stream absolute value of all frames and according to time sequence generate the light stream sequence of whole video level direction and vertical direction;
Step 3: analyze the light stream sequence signature, according to distorting the characteristics of pattern between different frame of video, utilize the light stream rate of change to judge whether video is tampered, if be tampered, determine to distort type;
Step 4: carry out video recovery.
2. altering detecting method between the conforming frame of video of optical flow-based according to claim 1, is characterized in that, extracts the light stream vector between described two adjacent video frames, specifically comprises the steps:
Step 2.1: obtain two adjacent video frames in video for decompress(ion), each frame of video is carried out four iteration frame compressions, generate 5 layers of pyramid that comprise original video frame, in pyramid, every one deck is to be obtained through compression by its lower one deck, wherein, pyramidal bottom, namely layer 5, be original video frame;
Step 2.2: from the pyramid top to bottom, calculate simultaneously the motion vector of horizontal direction and vertical direction between two video frame images of corresponding identical level between two adjacent video frames, wherein, pyramidal top is pyramidal ground floor;
Step 2.3: 5 layer motion vectors that obtain by step 2.2 are denoted as from top to bottom successively: layer (1), layer (2), layer (3), layer (4) and layer (5);
At first, generate the motion vector figure of a sky, its size is consistent with layer (1), and the light stream value of all pixels of the inside is 0, is denoted as ExpMV (0);
ExpMV (0) and layer (1) addition are obtained MVSum (1), then MVSum (1) is carried out smoothing processing; MVSum (1) after level and smooth is denoted as MVSmooth (1), then MVSmooth (1) is expanded obtaining ExpMV (1);
ExpMV (1) and layer (2) addition are obtained MVSum (2), then MVSum (2) is carried out smoothing processing; MVSum (2) after level and smooth is denoted as MVSmooth (2), then MVSmooth (2) is expanded obtaining ExpMV (2);
ExpMV (2) and layer (3) addition are obtained MVSum (3), then MVSum (3) is carried out smoothing processing; MVSum (3) after level and smooth is denoted as MVSmooth (3), then MVSmooth (3) is expanded obtaining ExpMV (3);
ExpMV (3) and layer (4) addition are obtained MVSum (4), then MVSum (4) is carried out smoothing processing; MVSum (4) after level and smooth is denoted as MVSmooth (4), then MVSmooth (4) is expanded obtaining ExpMV (4);
ExpMV (4) and layer (5) addition are obtained MVSum (5), then MVSum (5) is carried out smoothing processing; MVSum (5) after level and smooth is denoted as MVSmooth (5), and MVSmooth (5) is two light stream vectors between adjacent video frames.
3. altering detecting method between the conforming frame of video of optical flow-based according to claim 2, is characterized in that, in described step 2.1, described compression specifically comprises the steps:
Step 2.1.1: spatial sampling: the even number line and the even column that extract video frame images form a new images;
Step 2.1.2: the window that is N with a size in the horizontal direction of new images carries out convolution algorithm, carries out smoothing processing with the pixel to horizontal direction, and N is natural number;
Step 2.1.3: the window that is N with a size in the vertical direction of the image that obtains by step 2.1.2 carries out convolution algorithm, and the pixel of vertical direction is carried out smoothing processing, obtains final compressed image.
4. altering detecting method between the conforming frame of video of optical flow-based according to claim 3, is characterized in that, described expansion specifically comprises the steps:
Steps A: spatial spread: horizontal direction and vertical direction at image are inserted blank line and blank column;
Step B: the window that is N with a size in the horizontal direction of new images carries out convolution algorithm, and the pixel of horizontal direction is carried out smoothing processing;
Step C: the window that the vertical direction of the image that forms in the B step is N with a size carries out convolution algorithm, and the pixel of vertical direction is carried out smoothing processing, obtains final expanded images.
5. altering detecting method between the conforming frame of video of the described optical flow-based of any one according to claim 1 to 4, it is characterized in that, for given two frame of video i and i+1, the formula calculated level direction below utilizing and the light stream absolute value of all pixels of vertical direction and:
Wherein, OFX
(i, i+1)(w, h) refers to the light stream absolute value of pixel (w, h) position, and width is the frame of video width, and height is the frame of video height, S
(i, i+1)(x) be the final horizontal direction that generates the light stream absolute value and;
Wherein, OFX
(i, i+1)(w, h) refers to the light stream absolute value of pixel (w, h) position, and width is the frame of video width, and height is the frame of video height, S
(i, i+1)(y) be the final vertical direction that generates the light stream absolute value and;
Further, the light stream rate of change OFCR between i frame and i+1 frame
(i, i+1)(x) be:
OFCR
(i,i+1)(x)=2×S
(i,i+1)(x)/(S
(i-1,i)(x)+S
(i+1,i+2)(x))
Wherein, S
(i, i+1)(x) be the horizontal direction between i frame and i+1 frame the light stream absolute value and, S
(i-1, i)(x) be the horizontal direction between i-1 frame and i frame the light stream absolute value and, S
(i+1, i+2)(x) be the horizontal direction between i+1 frame and i+2 frame the light stream absolute value and.
6. altering detecting method between the conforming frame of video of optical flow-based according to claim 5, is characterized in that,
If have two very lofty light stream pulses in the light stream sequence, and do not have on all four light stream sequence in video, do not exist continuous light stream value to equate and near 0 situation, distort and be defined as frame of video and insert and distort;
If have a light stream pulse in the light stream sequence, and do not have on all four light stream sequence in video, do not exist continuous light stream value to equate and near 0 situation, determining to distort type is that the frame of video deletion is distorted;
If there is not the light stream pulse in the light stream sequence, but exists continuous light stream value to equate and near 0 light stream sequence, do not have on all four light stream sequence, determining to distort type is that video single frames multiple copies is distorted;
If there are two light stream pulses in the light stream sequence, do not exist continuous light stream to equate in video and near 0 situation, but have other one section light stream subsequence in video, and the light stream subsequence between two pulses is in full accord, to determine to distort type be that the video successive frame is non-adjacent copies;
If there is a light stream pulse in the light stream sequence, do not exist continuous light stream value to equate in video and near 0 light stream sequence, but before and after pulse, two sections light stream subsequences are in full accord, determining to distort type is that adjacent the copying of video successive frame distorted.
7. altering detecting method between the conforming frame of video of optical flow-based according to claim 6, is characterized in that, N=5, wherein,
Determine that the method that exists frame to insert the light stream pulse of distorting type of type is to have two points, its light stream rate of change is greater than threshold value threshold1=20;
The method of determining to exist the light stream pulse of distorting type of frame deletion type is to have a point, and its light stream rate of change is greater than threshold value threshold2=2;
Determine to exist single frames repeatedly the method for the light stream pulse of distorting type of type be to have a plurality of continuous points, its light stream rate of change is 1;
The method of determining to exist the light stream pulse of distorting type of the non-adjacent copy type of successive frame is to have two points, and its light stream rate of change is greater than threshold value threshold2=2;
The method of determining to exist the light stream pulse of distorting type of the adjacent copy type of successive frame is to have a point, and its light stream rate of change is greater than threshold value threshold2=2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013103074816A CN103384331A (en) | 2013-07-19 | 2013-07-19 | Video inter-frame forgery detection method based on light stream consistency |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013103074816A CN103384331A (en) | 2013-07-19 | 2013-07-19 | Video inter-frame forgery detection method based on light stream consistency |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103384331A true CN103384331A (en) | 2013-11-06 |
Family
ID=49492017
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013103074816A Pending CN103384331A (en) | 2013-07-19 | 2013-07-19 | Video inter-frame forgery detection method based on light stream consistency |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103384331A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103945228A (en) * | 2014-03-28 | 2014-07-23 | 上海交通大学 | Video intra-frame copy-move tampering detection method based on space and time relevance |
| CN103960413A (en) * | 2014-04-17 | 2014-08-06 | 黄叶芳 | Production method of internal heat lowering tea |
| CN104093033A (en) * | 2014-06-12 | 2014-10-08 | 中山大学 | H264/AVC video frame deletion identification method and deleted frame quantity estimation method |
| CN104113737A (en) * | 2014-07-30 | 2014-10-22 | 黄喜荣 | System for preventing monitoring video from being replaced and method thereof |
| CN104270644A (en) * | 2014-09-28 | 2015-01-07 | 上海交通大学 | Video inter-frame tampering detecting method based on velocity field coincidence |
| CN104657994A (en) * | 2015-02-13 | 2015-05-27 | 厦门美图之家科技有限公司 | Image consistency judging method and system based on optical flow method |
| CN105141968A (en) * | 2015-08-24 | 2015-12-09 | 武汉大学 | Video same-source copy-move tampering detection method and system |
| CN105554508A (en) * | 2015-12-11 | 2016-05-04 | 南京昊烽信息科技有限公司 | Method and system for detecting video inter-frame tampering for eliminating GOP periodic noise effects |
| CN106375756A (en) * | 2016-09-28 | 2017-02-01 | 宁波大学 | Single object removing and tampering detection method for monitored video |
| CN106878704A (en) * | 2017-02-14 | 2017-06-20 | 福建师范大学 | Turn altering detecting method on video frame rate based on light stream cyclophysis |
| CN107820075A (en) * | 2017-11-27 | 2018-03-20 | 中国计量大学 | A kind of VR equipment delayed test devices based on light stream camera |
| CN108108699A (en) * | 2017-12-25 | 2018-06-01 | 重庆邮电大学 | Merge deep neural network model and the human motion recognition method of binary system Hash |
| CN110418129A (en) * | 2019-07-19 | 2019-11-05 | 长沙理工大学 | Digital video interframe altering detecting method and system |
| CN110430339A (en) * | 2019-07-19 | 2019-11-08 | 长沙理工大学 | Altering detecting method and system in digital video frame |
| CN111539301A (en) * | 2020-04-20 | 2020-08-14 | 贵州安防工程技术研究中心有限公司 | Scene chaos degree judging method based on video analysis technology |
| CN113660443A (en) * | 2020-05-12 | 2021-11-16 | 武汉Tcl集团工业研究院有限公司 | Video frame insertion method, terminal and storage medium |
| US20220174076A1 (en) * | 2020-11-30 | 2022-06-02 | Microsoft Technology Licensing, Llc | Methods and systems for recognizing video stream hijacking on edge devices |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102413327A (en) * | 2011-11-10 | 2012-04-11 | 中山大学深圳研究院 | Video tampering detection method based on compressed sensing semi fragile watermark |
-
2013
- 2013-07-19 CN CN2013103074816A patent/CN103384331A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102413327A (en) * | 2011-11-10 | 2012-04-11 | 中山大学深圳研究院 | Video tampering detection method based on compressed sensing semi fragile watermark |
Non-Patent Citations (2)
| Title |
|---|
| XINGHAO JIANG, TANFENG SUN*, BING FENG,CHENGMING JIANG: "A Space-Time SURF Descriptor and its Application", 《IEEE 2011 EIGHTH INTERNATIONAL CONFERENCE ON FUZZY SYSTEMS AND KNOWLEDGE DISCOVERY》, 28 July 2011 (2011-07-28), pages 1911 - 1915 * |
| 巢娟: "基于内容的视频帧间篡改被动检测算法研究", 《中国优秀硕士学位论文全文数据库;信息科技辑》, 15 July 2013 (2013-07-15), pages 136 - 466 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103945228B (en) * | 2014-03-28 | 2017-03-15 | 上海交通大学 | Video frame in copy move altering detecting methods based on space-time relationship |
| CN103945228A (en) * | 2014-03-28 | 2014-07-23 | 上海交通大学 | Video intra-frame copy-move tampering detection method based on space and time relevance |
| CN103960413A (en) * | 2014-04-17 | 2014-08-06 | 黄叶芳 | Production method of internal heat lowering tea |
| CN104093033A (en) * | 2014-06-12 | 2014-10-08 | 中山大学 | H264/AVC video frame deletion identification method and deleted frame quantity estimation method |
| CN104093033B (en) * | 2014-06-12 | 2017-04-19 | 中山大学 | H264/AVC video frame deletion identification method and deleted frame quantity estimation method |
| CN104113737A (en) * | 2014-07-30 | 2014-10-22 | 黄喜荣 | System for preventing monitoring video from being replaced and method thereof |
| CN104113737B (en) * | 2014-07-30 | 2017-05-10 | 黄喜荣 | System for preventing monitoring video from being replaced and method thereof |
| CN104270644A (en) * | 2014-09-28 | 2015-01-07 | 上海交通大学 | Video inter-frame tampering detecting method based on velocity field coincidence |
| CN104657994A (en) * | 2015-02-13 | 2015-05-27 | 厦门美图之家科技有限公司 | Image consistency judging method and system based on optical flow method |
| CN104657994B (en) * | 2015-02-13 | 2017-12-19 | 厦门美图之家科技有限公司 | A kind of method and system that image consistency is judged based on optical flow method |
| CN105141968A (en) * | 2015-08-24 | 2015-12-09 | 武汉大学 | Video same-source copy-move tampering detection method and system |
| CN105141968B (en) * | 2015-08-24 | 2016-08-17 | 武汉大学 | A kind of video homology copy-move altering detecting method and system |
| CN105554508B (en) * | 2015-12-11 | 2018-08-03 | 南京昊烽信息科技有限公司 | Eliminate the video interframe altering detecting method and system of GOP periodic noise effects |
| CN105554508A (en) * | 2015-12-11 | 2016-05-04 | 南京昊烽信息科技有限公司 | Method and system for detecting video inter-frame tampering for eliminating GOP periodic noise effects |
| CN106375756B (en) * | 2016-09-28 | 2017-12-19 | 宁波大学 | It is a kind of to remove the detection method distorted for the single object of monitor video |
| CN106375756A (en) * | 2016-09-28 | 2017-02-01 | 宁波大学 | Single object removing and tampering detection method for monitored video |
| CN106878704A (en) * | 2017-02-14 | 2017-06-20 | 福建师范大学 | Turn altering detecting method on video frame rate based on light stream cyclophysis |
| CN107820075A (en) * | 2017-11-27 | 2018-03-20 | 中国计量大学 | A kind of VR equipment delayed test devices based on light stream camera |
| CN108108699A (en) * | 2017-12-25 | 2018-06-01 | 重庆邮电大学 | Merge deep neural network model and the human motion recognition method of binary system Hash |
| CN110418129A (en) * | 2019-07-19 | 2019-11-05 | 长沙理工大学 | Digital video interframe altering detecting method and system |
| CN110430339A (en) * | 2019-07-19 | 2019-11-08 | 长沙理工大学 | Altering detecting method and system in digital video frame |
| CN110418129B (en) * | 2019-07-19 | 2021-03-02 | 长沙理工大学 | Digital video interframe tampering detection method and system |
| CN111539301A (en) * | 2020-04-20 | 2020-08-14 | 贵州安防工程技术研究中心有限公司 | Scene chaos degree judging method based on video analysis technology |
| CN111539301B (en) * | 2020-04-20 | 2023-04-18 | 贵州安防工程技术研究中心有限公司 | Scene chaos degree discrimination method based on video analysis technology |
| CN113660443A (en) * | 2020-05-12 | 2021-11-16 | 武汉Tcl集团工业研究院有限公司 | Video frame insertion method, terminal and storage medium |
| US20220174076A1 (en) * | 2020-11-30 | 2022-06-02 | Microsoft Technology Licensing, Llc | Methods and systems for recognizing video stream hijacking on edge devices |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103384331A (en) | Video inter-frame forgery detection method based on light stream consistency | |
| Sitara et al. | Digital video tampering detection: An overview of passive techniques | |
| Lancini et al. | A robust video watermarking technique in the spatial domain | |
| Sowmya et al. | A survey on video forgery detection | |
| Ravi et al. | Compression noise based video forgery detection | |
| Al-Sanjary et al. | Detection of video forgery: A review of literature | |
| Su et al. | A practical design of digital video watermarking in H. 264/AVC for content authentication | |
| CN108650491B (en) | Video watermark detection method for monitoring system | |
| US8755598B2 (en) | System for embedding data | |
| CN102147912A (en) | Adaptive difference expansion-based reversible image watermarking method | |
| CN103561271A (en) | Video airspace tamper detection method for removing moving object shot by static camera lens | |
| CN101527850A (en) | Video self-recovery watermark method based on motion estimation | |
| Kingra et al. | Video inter-frame forgery detection approach for surveillance and mobile recorded videos | |
| Niu et al. | Video watermarking resistance to rotation, scaling, and translation | |
| KR101310140B1 (en) | Integrity checking Process of Image using Reversible Watermarking Process Tecjnique | |
| CN104270644A (en) | Video inter-frame tampering detecting method based on velocity field coincidence | |
| Sun et al. | Video watermarking synchronization based on profile statistics | |
| CN102413381A (en) | Video watermark based on optical flow method and digital holographic | |
| Habeeb et al. | A review: video tampering attacks and detection techniques | |
| KR20090101332A (en) | Watermarking and encryption technique for resolution improvement, forgery prevention and contents spread prevention in surveillance camera system | |
| Aminu Mustapha | Passive video forgery detection using frame correlation statistical features/Aminu Mustapha Bagiwa | |
| Bagiwa | Passive Video Forgery Detection Using Frame Correlation Statistical Features | |
| CN101635855B (en) | Method and device for embedding and blindly picking up video watermark | |
| Chennamma | A survey on video forgery detection | |
| Panthakkan et al. | Preserving the Integrity of Hyperspectral Satellite Imagery Using a Secure and Efficient Reversible Watermarking Technique for Ownership Protection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131106 |