CN104135597B - A kind of video jitter automatic testing method - Google Patents
A kind of video jitter automatic testing method Download PDFInfo
- Publication number
- CN104135597B CN104135597B CN201410318324.XA CN201410318324A CN104135597B CN 104135597 B CN104135597 B CN 104135597B CN 201410318324 A CN201410318324 A CN 201410318324A CN 104135597 B CN104135597 B CN 104135597B
- Authority
- CN
- China
- Prior art keywords
- video
- light stream
- jitter
- stream vector
- histogram
- 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.)
- Active
Links
Abstract
The present invention discloses a kind of video jitter automatic testing method, including：The first step, the selected characteristic point in video pictures, second step, the tracking and matching characteristic point in consecutive frame, 3rd step, characteristic point is removed to obtain interframe movement vector, the 4th step, extraction video jitter frequency, video jitter amplitude characteristic to carrying out abnormity point, 5th step, interframe light stream vector histogram feature, the 6th step are extracted, grader judges video jitter degree.The present invention does not need any priori when judging video with the presence or absence of shake, and has higher accuracy rate.
Description
Technical field
The present invention can be used for the numeral captured by mobile phone, DV, the first-class movement of monitoring camera or fixed equipment to regard
Frequently, automatic detection, specifically a kind of video jitter automatic testing method are carried out with the presence or absence of shake to digital video picture.
Background technology
The digital video of mobile terminal shooting is often accompanied by video jitter, and the presence of video jitter influences the vision matter of video
The subsequent treatment effect such as amount and target identification, therefore, it presently, there are various video digital image stabilization method and be used to reduce video jitter
Degree.But for an arbitrary digital video, it is necessary to first judge its video jitter degree, could correctly it select correctly
Video image stabilization method and method parameter.
On the one hand, video jitter detection method can be as the pretreatment module of video image stabilization method, for any video
Video jitter degree is first detected, corresponding digital image stabilization method and parameter are selected according to degree of jitter；On the other hand, video jitter detection side
Method can carry out shaking detection, to steady picture as the subsequent treatment module of video image stabilization method to the video after digital image stabilization method processing
The effect of method is evaluated.
At present, the shaking detection for digital video and evaluation are also without complete detection method and standard, purport of the present invention
Iing is proposed a kind of effective video shake automatic testing method.
The content of the invention
The present invention proposes a kind of video jitter automatic testing method, in the situation of no any input video priori
Under, it can be used for choosing digital image stabilization method species and regulation digital image stabilization method parameter, and available for the steady as effect of evaluation digital image stabilization method
Fruit.
To achieve the above object, the technical solution adopted by the present invention is：The present invention is found uniformly in video pictures first
The characteristic point of distribution, tracking characteristics point obtain sparse optical flow vector, then exclude to obtain interframe light stream by abnormal light stream vector
Vector.Interframe light stream vector is averaged to obtain the estimation of inter frame motion model, video jitter frequency is counted from motion model,
The class jitter feature of video jitter amplitude two, light stream vector histogram is obtained as a kind of jitter feature according to interframe light stream vector.
Above-mentioned three classes video jitter feature is finally input to the video jitter degree grader trained and obtains the shake of input video
Degree.
A kind of video jitter automatic testing method of the present invention, comprises the following steps：
The first step, choose characteristic point in present frame picture；
Second step, the tracking and matching characteristic point in consecutive frame, obtain interframe light stream vector；
3rd step, the interframe light stream vector obtained to second step carry out abnormity point exclusion；
4th step, the interframe light stream vector obtained according to the 3rd step are estimated inter frame motion model, then carried according to this model
Take out two video jitter frequency, video jitter amplitude features；
5th step, it is another as video jitter that the interframe light stream vector obtained according to the 3rd step counts light stream vector histogram
One feature；
6th step, video jitter frequency, video jitter amplitude and the light stream vector that the 4th step and the 5th step are extracted are straight
The square class jitter feature of figure three is input to the grader trained and obtains the video jitter degree.
Compared with prior art, the invention has the advantages that：
The present invention estimates video inter frame motion model using the thought of sparse optical flow calculating, it is not necessary to any video priori
Knowledge, most of video scene is applicable, and has higher accuracy rate.Can preferably it be selected according to the degree of jitter of video
Video image stabilization method and the parameter of digital image stabilization method are taken, is improved surely as the efficiency and effect of step.
Brief description of the drawings
The detailed description made by reading with reference to the following drawings to non-limiting example, further feature of the invention,
Objects and advantages will become more apparent upon：
Fig. 1 is a kind of video jitter automatic testing method overview flow chart proposed by the present invention.
Fig. 2 is the design sketch of feature point extraction in the present invention, and dot is selected characteristic point position wherein in figure.
Fig. 3 divides characteristic pattern during X and Y both directions statistics for light stream vector histogram in the present invention, wherein figure (a) is steady
Determine video X-direction light stream vector histogram, for figure (b) to stablize video Y-direction light stream vector histogram, figure (c) is slight jitter
Video X-direction light stream vector histogram, figure (d) is slight jitter video Y-direction light stream vector histogram, and figure (e) is acutely to tremble
Dynamic video X-direction light stream vector histogram, figure (f) are acutely shake video Y-direction light stream vector histogram.
Fig. 4 detects grader recall ratio -- precision ratio curve for video jitter in the present invention.
Fig. 5 lists intention for the present invention using case.
Embodiment
With reference to specific embodiment, the present invention is described in detail.Following examples will be helpful to the technology of this area
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill to this area
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made.These belong to the present invention
Protection domain.
As shown in figure 1, for the overview flow chart of one embodiment of the invention, specifically include：
The first step, choose characteristic point in current video frame；
Second step, the tracking and matching characteristic point in consecutive frame, obtain interframe light stream vector；
3rd step, the interframe light stream vector obtained to second step carry out abnormity point removal；, will during abnormal light stream vector removes
Phase and amplitude is considered as abnormal light stream vector with the light stream vector that most of light stream vector differs greatly, and has to be removed.
4th step, the interframe light stream vector obtained according to the 3rd step are estimated inter frame motion model, then carried according to this model
Take out two video jitter frequency, video jitter amplitude features；In the estimation of inter frame motion model, inter frame motion model is translation
Model, using interframe light stream vector average value as motion model estimation.
5th step, it is special as video jitter that the interframe light stream vector obtained according to the 3rd step counts light stream vector histogram
Sign；The statistics of light stream vector histogram, statistics be abnormity point exclude after light stream vector, light stream vector histogram with parallel connection
Mode is counted all normal interframe light stream vectors.
6th step, video jitter frequency, video jitter amplitude and the light stream vector that the 4th step and the 5th step are extracted are straight
The square class jitter feature of figure three is input to the grader trained, obtains the video jitter degree.In this step：Jitter level is classified
Device training set is the stable video of shooting and the shake video of artificial synthesized different jitter levels.Jitter level grader test set
It is video jitter subjective assessment that the label of video, which selectes standard,.Using SVMs as video jitter grade separation device.
Based on above-mentioned step, a kind of specific implementation details of video jitter automatic testing method are as follows：
Fig. 2 is feature point extraction schematic diagram in the present invention, and the chosen position of characteristic point should be evenly distributed in entirely as far as possible
Video pictures, characteristic point position are marked by dot.
When feature point tracking the match is successful obtain interframe light stream vector after, if formula (1) set up, then it is assumed that video pictures are in X
Exist on direction and once shake, if formula (2) is set up, then it is assumed that video pictures exist in the Y direction once to be shaken.
[P(x)|k+1-P(x)|k]·[P(x)|k-P(x)|k-1] ＜ 0 (1)
[P(y)|k+1-P(y)|k]·[P(y)|k-P(y)|k-1] ＜ 0 (2)
Left side Section 1 represents that kth frame is estimated to the interframe translational motion model of the frame of kth+1 in X-direction wherein in formula (1),
Section 2 represents that the frame of kth -1 estimates that left side Section 1 represents Y in formula (2) to the interframe translational motion model of kth frame in X-direction
Kth frame is estimated to the interframe translational motion model of the frame of kth+1 on direction, and Section 2 represents that the frame of kth -1 arrives kth frame in Y-direction
Interframe translational motion model is estimated.The frequency of video jitter can be obtained after video sequence by having traveled through.
Video jitter amplitude represents inter frame motion model (X positive directions, X opposite directions, Y positive directions or Y negative side in one direction
To) ultimate range of cumulative movement.Video jitter amplitude will make normalization operation, to adapt to different size of video pictures.
Fig. 3 is light stream vector histogram feature figure in the embodiment of the present invention, will be each when counting light stream vector histogram
Light stream vector projects to X and Y both directions and taken statistics again, illustrates the foundation of histogram by taking X-direction as an example first.Count X-direction
Histogram when, be subdivided into both direction, one be X positive direction (value is just), one is that (value is for X opposite direction
It is negative), the step-length of histogram is the 1% of video width, because the seldom part of interframe light stream vector displacement can be wide more than video
The 20% of degree, so histogram is divided into 40 grades, X both forward and reverse directions respectively have 20 grades.When light stream vector is more than in the displacement of X-direction
When 20%, then it is included among last rank, i.e., 1% to 19% each own step-length from length for video width
Unified more than 19% to calculate into a rank for 1% rank, the transverse axis scope of light stream vector histogram is from left to right successively
For：(- ∞, -19%], (- 19%, -18%] ..., (- 1%, 0], (0,1%], (1%, 2%] ..., (18%, 19%],
Totally 40 grades of (19% ,+∞).Similarly, the statistics with histogram for Y-direction is also same step-length and statistical method, unique different
Be that its step-length is video height 1%, rather than the 1% of video width.Finally make normalization behaviour to light stream vector histogram
Make, the height of histogram represents that the light stream vector in the section accounts for the ratio of total light stream vector number.
Fig. 3 depicts the light stream vector histogram of three videos, wherein, figure (a) is straight to stablize video X-direction light stream vector
Fang Tu, for figure (b) to stablize video Y-direction light stream vector histogram, figure (c) is slight jitter video X-direction light stream vector Nogata
Figure, figure (d) is slight jitter video Y-direction light stream vector histogram, and figure (e) is straight for acutely shake video X-direction light stream vector
Fang Tu, figure (f) are acutely shake video Y-direction light stream vector histogram.
Fig. 4 detects grader recall ratio -- precision ratio curve for video jitter in the present invention.
Fig. 5 lists intention for the present invention using case, embodies contribution of the present invention during steady picture, more perfect whole
The step of individual Video Stabilization.
From above-described embodiment as can be seen that the present invention does not need any priori to know when judging video with the presence or absence of shake
Know, and have higher accuracy rate.
The specific embodiment of the present invention is described above.It is to be appreciated that the invention is not limited in above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring the substantive content of the present invention.
Claims (7)
1. a kind of video jitter automatic testing method, it is characterised in that methods described comprises the following steps：
The first step, choose characteristic point in current video frame；
Second step, the tracking and matching characteristic point in consecutive frame, obtain interframe light stream vector；
3rd step, the interframe light stream vector obtained to second step carry out abnormity point removal；
4th step, the interframe light stream vector obtained according to the 3rd step are estimated inter frame motion model, then gone out according to this model extraction
Two video jitter frequency, video jitter amplitude features；
5th step, the interframe light stream vector obtained according to the 3rd step count light stream vector histogram as video jitter feature；
6th step, video jitter frequency, video jitter amplitude and the light stream vector histogram that the 4th step and the 5th step are extracted
Three class jitter features are input to the grader trained, obtain the video jitter degree；
When feature point tracking the match is successful obtain interframe light stream vector after, if formula (1) set up, then it is assumed that video pictures are in X-direction
Upper presence is once shaken, if formula (2) is set up, then it is assumed that video pictures exist in the Y direction once to be shaken：
[P(x)|k+1-P(x)|k]·[P(x)|k-P(x)|k-1] ＜ 0 (1)
[P(y)|k+1-P(y)|k]·[P(y)|k-P(y)|k-1] ＜ 0 (2)
Left side Section 1 represents that kth frame is estimated to the interframe translational motion model of the frame of kth+1 in X-direction wherein in formula (1), second
Item represents that the frame of kth -1 estimates that left side Section 1 represents Y-direction in formula (2) to the interframe translational motion model of kth frame in X-direction
Upper kth frame to the frame of kth+1 interframe translational motion model estimate, Section 2 represent Y-direction on the frame of kth -1 to kth frame interframe
Translational motion model is estimated；The frequency of video jitter can be obtained after video sequence by having traveled through；
Video jitter amplitude represents the ultimate range of inter frame motion model cumulative movement in one direction, and video jitter amplitude is made to return
One changes operation, to adapt to different size of video pictures；
When counting light stream vector histogram, each light stream vector is projected into X and Y both directions and taken statistics again, first with X side
Illustrate the foundation of histogram exemplified by, when counting the histogram of X-direction, be subdivided into both direction, one is that X positive direction (takes
Be worth for just), one be X opposite direction (value is negative), the step-length of histogram is the 1% of video width, because interframe light stream is sweared
The 20% of video width can be more than by measuring the seldom part of displacement, so histogram is divided into 40 grades, X both forward and reverse directions respectively have 20
Level；When light stream vector X-direction displacement be more than 20% when, then be included among last rank, i.e., from length be regarding
The rank that 1% to 19% each own step-length of frequency range degree is 1%, unified more than 19% to calculate into a rank, light stream arrow
The transverse axis scope of amount histogram is from left to right followed successively by：(- ∞, -19%], (- 19%, -18%] ..., (- 1%, 0], (0,
1%], (1%, 2%] ..., (18%, 19%], totally 40 grades of (19% ,+∞)；Similarly, it is also for the statistics with histogram of Y-direction
Same step-length and statistical method, it is unique the difference is that its step-length is the 1% of video height, rather than the 1% of video width；Most
Make normalization operation to light stream vector histogram afterwards, the height of histogram represents that the light stream vector in the section accounts for total light stream
The ratio of vector number.
2. a kind of video jitter automatic testing method according to claim 1, it is characterised in that in the 3rd step：Will
The light stream vector that phase and amplitude differs larger with most of light stream vector is considered as abnormal light stream vector, has to be removed.
3. a kind of video jitter automatic testing method according to claim 1, it is characterised in that in the 4th step：Frame
Between motion model be translation model, using the average value of interframe light stream vector as estimation.
4. a kind of video jitter automatic testing method according to claim 1, it is characterised in that in the 5th step：Light
Flow vector histogram is in parallel counted all normal interframe light stream vectors.
5. a kind of video jitter automatic testing method according to claim any one of 1-4, it is characterised in that the described 6th
In step：Using the stable video of shooting and the shake video of artificial synthesized different jitter levels as classifier training collection.
A kind of 6. video jitter automatic testing method according to claim 5, it is characterised in that the test of the grader
The selected standard of label for collecting video is video jitter subjective assessment.
7. a kind of video jitter automatic testing method according to claim 6, it is characterised in that the 6th step is using branch
Vector machine is held as grader.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410318324.XA CN104135597B (en) | 2014-07-04 | 2014-07-04 | A kind of video jitter automatic testing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410318324.XA CN104135597B (en) | 2014-07-04 | 2014-07-04 | A kind of video jitter automatic testing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104135597A CN104135597A (en) | 2014-11-05 |
| CN104135597B true CN104135597B (en) | 2017-12-15 |
Family
ID=51808112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410318324.XA Active CN104135597B (en) | 2014-07-04 | 2014-07-04 | A kind of video jitter automatic testing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104135597B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105681663B (en) * | 2016-02-26 | 2018-06-22 | 北京理工大学 | A kind of video jitter detection method based on interframe movement geometry flatness |
| CN105812788A (en) * | 2016-03-24 | 2016-07-27 | 北京理工大学 | Video stability quality assessment method based on interframe motion amplitude statistics |
| CN106210448B (en) * | 2016-07-22 | 2020-08-18 | 深圳市智慧城市建设运行管理有限公司 | Video image jitter elimination processing method |
| CN106713702A (en) * | 2017-01-19 | 2017-05-24 | 博康智能信息技术有限公司 | Method and apparatus of determining video image jitter and camera device jitter |
| CN108040225B (en) * | 2017-12-11 | 2020-02-07 | 上海瀚正信息科技股份有限公司 | Intelligent monitoring robot |
| CN109194878B (en) * | 2018-11-08 | 2021-02-19 | 深圳市闻耀电子科技有限公司 | Video image anti-shake method, device, equipment and storage medium |
| CN110062222B (en) * | 2019-03-15 | 2021-06-29 | 华为技术有限公司 | Video evaluation method, terminal, server and related products |
| CN110248048B (en) * | 2019-06-21 | 2021-11-09 | 苏宁云计算有限公司 | Video jitter detection method and device |
| CN110349177B (en) * | 2019-07-03 | 2021-08-03 | 广州多益网络股份有限公司 | Method and system for tracking key points of human face of continuous frame video stream |
| CN110401796B (en) * | 2019-07-05 | 2020-09-29 | 浙江大华技术股份有限公司 | Jitter compensation method and device of image acquisition device |
| CN110602487B (en) * | 2019-09-06 | 2021-04-20 | 高新兴科技集团股份有限公司 | Video image jitter detection method based on TSN (time delay network) |
| CN111372000B (en) * | 2020-03-17 | 2021-08-17 | Oppo广东移动通信有限公司 | Video anti-shake method and apparatus, electronic device, and computer-readable storage medium |
| CN113691733B (en) * | 2021-09-18 | 2023-04-18 | 北京百度网讯科技有限公司 | Video jitter detection method and device, electronic equipment and storage medium |
| CN113887547B (en) * | 2021-12-08 | 2022-03-08 | 北京世纪好未来教育科技有限公司 | Key point detection method and device and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103546726A (en) * | 2013-10-28 | 2014-01-29 | 东南大学 | Method for automatically discovering illegal land use |
| CN103763461A (en) * | 2013-12-31 | 2014-04-30 | 合一网络技术(北京)有限公司 | Video jitter detection method and device |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070171981A1 (en) * | 2006-01-25 | 2007-07-26 | Yingyong Qi | Projection based techniques and apparatus that generate motion vectors used for video stabilization and encoding |
| CN100530239C (en) * | 2007-01-25 | 2009-08-19 | 复旦大学 | Video stabilizing method based on matching and tracking of characteristic |
| US8451384B2 (en) * | 2010-07-08 | 2013-05-28 | Spinella Ip Holdings, Inc. | System and method for shot change detection in a video sequence |
| CN102202164B (en) * | 2011-05-20 | 2013-03-20 | 长安大学 | Motion-estimation-based road video stabilization method |
| CN102222349B (en) * | 2011-07-04 | 2013-01-23 | 江苏大学 | Prospect frame detecting method based on edge model |
-
2014
- 2014-07-04 CN CN201410318324.XA patent/CN104135597B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103546726A (en) * | 2013-10-28 | 2014-01-29 | 东南大学 | Method for automatically discovering illegal land use |
| CN103763461A (en) * | 2013-12-31 | 2014-04-30 | 合一网络技术(北京)有限公司 | Video jitter detection method and device |
Non-Patent Citations (5)
| Title |
|---|
| Shaking video synthesis for video stabilization performance assessment;Hui Qu, et al;《Visual Communications and Image Processing(VCIP),2013》;20131120;全文 * |
| VIdeo stabilization with L1-L2 optimization;Hui Qu, et al;《Image Processing(ICIP),2013 20th IEEE International Conference》;20130918;第29-33页 * |
| 基于方向直方图矢量的手势识别;江冬梅等;《信息技术与信息化》;20060415(第2期);第53-55页 * |
| 基于运动矢量的视频去抖动算法;宋利等;《上海交通大学学报》;20041230;第38卷;第63-65页 * |
| 稳定运动物体视频的特征方法;邱家涛等;《光学精密工程》;20121015;第20卷(第10期);第2300-2307页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104135597A (en) | 2014-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104135597B (en) | A kind of video jitter automatic testing method | |
| CN107133969B (en) | A kind of mobile platform moving target detecting method based on background back projection | |
| Shivakumara et al. | Efficient video text detection using edge features | |
| CN105681663B (en) | A kind of video jitter detection method based on interframe movement geometry flatness | |
| CN108280386B (en) | Monitoring scene detection method and device | |
| CN103226712B (en) | A kind of remnant object detection method based on finite state machine | |
| KR20140045854A (en) | Method and apparatus for monitoring video for estimating gradient of single object | |
| US11049256B2 (en) | Image processing apparatus, image processing method, and storage medium | |
| WO2014128688A1 (en) | Method, system and software module for foreground extraction | |
| CN107742307A (en) | Based on the transmission line galloping feature extraction and parameters analysis method for improving frame difference method | |
| CN102307274A (en) | Motion detection method based on edge detection and frame difference | |
| Mohan et al. | Video image processing for moving object detection and segmentation using background subtraction | |
| CN101976350A (en) | Grain storage pest detection and identification method based on video analytics and system thereof | |
| CN102855465B (en) | A kind of tracking of mobile object | |
| CN108364282B (en) | Image mosaic detection method and image mosaic detection system | |
| Gálai et al. | Feature selection for Lidar-based gait recognition | |
| Priya et al. | Edge strength extraction using orthogonal vectors for shot boundary detection | |
| CN102222349B (en) | Prospect frame detecting method based on edge model | |
| KR101537559B1 (en) | Device for detecting object, device for detecting object for vehicle and method thereof | |
| US20110085026A1 (en) | Detection method and detection system of moving object | |
| CN104378549A (en) | Snapshot method and device and terminal | |
| KR101214858B1 (en) | Moving object detecting apparatus and method using clustering | |
| CN114613006A (en) | Remote gesture recognition method and device | |
| CN106780544A (en) | The method and apparatus that display foreground is extracted | |
| KR101640527B1 (en) | Method and Apparatus for Monitoring Video for Estimating Size of Single Object |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |