CN106559605A - Digital video digital image stabilization method based on improved block matching algorithm - Google Patents
Digital video digital image stabilization method based on improved block matching algorithm Download PDFInfo
- Publication number
- CN106559605A CN106559605A CN201611032802.6A CN201611032802A CN106559605A CN 106559605 A CN106559605 A CN 106559605A CN 201611032802 A CN201611032802 A CN 201611032802A CN 106559605 A CN106559605 A CN 106559605A
- Authority
- CN
- China
- Prior art keywords
- frame
- block
- match block
- match
- present frame
- 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
- 238000000034 method Methods 0.000 title claims abstract description 28
- 230000006641 stabilisation Effects 0.000 title claims abstract description 19
- 238000011105 stabilization Methods 0.000 title claims abstract description 19
- 238000001914 filtration Methods 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000037433 frameshift Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/21—Circuitry for suppressing or minimising disturbance, e.g. moiré or halo
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20201—Motion blur correction
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Studio Devices (AREA)
Abstract
The present invention relates to digital video digital image stabilization method, cumulative errors are had for block matching algorithm in existing Video Stabilization, for motion blur situations such as process bad shortcoming, it is contemplated that proposing a kind of digital video digital image stabilization method based on improved block matching algorithm, step is as follows：(1) the abundant match block of selected characteristic；(2) KCF target tracking algorisms are run, judges picture change size；(3) result obtained according to (2nd) step, calculates the difference of Corresponding matching block top left co-ordinate in all match block top left co-ordinates and reference frame in present frame；(4) repeat step (2) and (3), until video terminates.Present invention is mainly applied to the steady picture of digital video.
Description
Technical field
The present invention relates to digital video digital image stabilization method, more particularly, it relates to a kind of based on improved block matching algorithm
Digital video digital image stabilization method.
Background technology
The steady picture of digital video, refers to and utilizes related algorithm, the original video that video capture device is obtained is processed, is removed
Go shake therein.The purpose of Video Stabilization, partly in order to making eye-observation comfortable, is conducive to artificial observation, differentiation etc.；
On the other hand the pretreatment stage of many other subsequent treatment, such as detection, tracking and compression are also served as.
The realization of typical Digital image stabilization includes motion estimation module and motion compensating module.In motion estimation module, pass
The motion estimation algorithm of system includes block matching algorithm, Gray Projection method and Feature Points Matching scheduling algorithm.Block matching algorithm has essence
The characteristics of spending high, therefore be one of the most frequently used algorithm in image stabilization system.But traditional block matching algorithm can be with video sequence
Row passage, matching result can slowly produce skew, additionally, the change (factor such as motion blur is caused) of match block can also make matching
As a result error is produced, the two factors cause the error of estimation slowly to add up.
The content of the invention
To overcome the deficiencies in the prior art, cumulative errors are had for block matching algorithm in existing Video Stabilization, for fortune
Situations such as dynamic model is pasted processes bad shortcoming, it is contemplated that proposing a kind of Digital image stabilization side based on improved block matching algorithm
Method.The technical solution used in the present invention is that, based on the digital video digital image stabilization method of improved block matching algorithm, step is as follows：
(1) in the first frame is simultaneously also reference frame, the diverse location in heart district domain in the picture, selected characteristic it is abundant
With block, meanwhile, core correlation filtering KCF (the Kernelized Correlation Filters) algorithm for follow-up tracking is extracted
Go out the feature of these match block, for tracking；
(2) for the second frame, match block is calculated in previous frame in the new position of present frame using block matching algorithm；Connect
Get off to run KCF target tracking algorisms, the new position that block matching algorithm is obtained before by the tracking area update of KCF being, then
Calculating is tracked in the new position, if the position that obtains of track algorithm and the same match block position deviation of reference frame compared with
Greatly, then present frame and reference frame difference are illustrated than larger, that is, picture is changed greatly；
(3) result obtained according to (2nd) step, in calculating present frame, all match block top left co-ordinates are right with reference frame
The difference of match block top left co-ordinate is answered, formula is as follows：
The coordinate figure in the present frame match block upper left corner is represented,Represent the seat in the reference frame match block upper left corner
Scale value,Represent that present frame and reference frame, for the difference of match block top left co-ordinate, are then averaging to these differences
Value, formula are as follows：
X directions and y directions present frame and reference frame are represented for the meansigma methodss of the difference of match block top left co-ordinate,
N represents the quantity of match block；Using it is steady as rear cutting by the way of, therefore cutting method is set for T pixel of each cutting up and down,
Work as differenceWithAbsolute value be respectively less than T pixels when, using directly full remuneration mode, will present frame in x directions and y side
To reverse movementWithDistance, then the image is respectively dismissed T pixels up and down, forms steady as picture；WhenOr's
When absolute value is more than T, illustrates that picture is changed greatly, then the match block of feature rich is extracted again in present frame, for next frame
Block- matching is searched for, meanwhile, present frame is set to into reference frame, compensation way is changed into a Contrary compensation T pixel, beyond part no longer
Compensation；
(4) repeat step (2) and (3), until video terminates.
Block matching algorithm calculate previous frame in match block in the new position of present frame, comprise the concrete steps that, if match block
Length a width of M and N, due to there is relative motion between sequence of frames of video, cause match block in a series of subsequent videos with it is front
The position of the match block of one frame shifts, it is assumed that on x, y direction, maximum offset is respectively dx and dy, then in video present frame
In select (M+2dx) * (N+2dy) region of search around match block, according to minimum total absolute error criterion in region of search
In carry out smallest match search, obtain best matching blocks, then calculate in previous frame matching result in match block and present frame frame
Coordinate difference, to each match block carry out matching search once, obtain the new position of each match block.
In step (1), picture centre region is defined as：Inside frame a quarter length and width, what selected characteristic was enriched
Match block, match block are typically of size of 16*16 pixels.
The characteristics of of the invention and beneficial effect are：
The present invention combines KCF track algorithms using block matching algorithm, can be right with point-device skew for calculating interframe
Contribute to steady picture, and obtain final stable image effect.
Description of the drawings：
Fig. 1：Target following frame diagram in shake video.
Fig. 2：Central area definition figure.
Fig. 3：Block search matching algorithm figure.
Specific embodiment
The present invention has cumulative errors for block matching algorithm in existing Video Stabilization, for processing situations such as motion blur
Bad shortcoming, proposes a kind of Digital image stabilization method based on improved block matching algorithm.The basic ideas of algorithm are exactly to introduce
KCF target tracking algorisms, i.e., after block matching algorithm operation, rerun KCF target tracking algorisms, obtains more accurate knot
Really.Next just consistent with common algorithm, motion filtering and motion compensation finally export the steady video as after.Due to the present invention
Mainly in the improvement of block matching algorithm, therefore the combination to block matching algorithm and KCF target followings describes in detail, other steps
Rapid such as motion filtering, is said using the simple method easily realized using common common method herein with motion compensation
It is bright.
Inventive algorithm general frame is as shown in figure 1, key step is as follows：
(1) in the first frame (while and reference frame), in the picture heart district domain (defined herein as：Apart from four points of frame
One of inside length and width, such as Fig. 2) diverse location, the abundant match block of selected characteristic, match block is typically of size of 16*16 (can be with
Suitably it is sized according to balance of the speed with precision).Meanwhile, the KCF algorithms for follow-up tracking extract these match block
Feature, for tracking.
(2) for the second frame, match block is calculated in previous frame in the new position of present frame using block matching algorithm.Block
Matching algorithm principle such as Fig. 3.If the length of match block a width of M and N.Due to there is relative motion between sequence of frames of video, cause with
Match block in latter series video is shifted with the position of the match block of former frame, it is assumed that maximum offset on x, y direction
Respectively dx and dy.Then (M+2dx) * (N+2dy) region of search is selected around match block in video present frame, according to most
Little total absolute error criterion carries out smallest match search in region of search, obtains best matching blocks, then calculates in previous frame
The coordinate difference of matching result in match block and present frame frame.Matching search is carried out to each match block once, each is obtained
The new position of match block.Next KCF target tracking algorisms are run.KCF algorithms are the track algorithms based on correlation filtering, the calculation
Method has the features such as tracking accuracy is high, speed is fast.The new position that block matching algorithm is obtained before by the tracking area update of KCF being
Put, calculating is then tracked in the new position, KCF track algorithms also can be recognized well to the slight change of target, Shandong
Rod is strong, therefore can obtain accurate position.If the position that track algorithm is obtained and the same match block position of reference frame
Deviation is larger, then illustrate present frame and reference frame difference than larger, that is, picture is changed greatly,
(3) result obtained according to (2nd) step, in calculating present frame, all match block top left co-ordinates are right with reference frame
Answer the difference of match block top left co-ordinate.Formula is as follows：
The coordinate figure in the present frame match block upper left corner is represented,Represent the seat in the reference frame match block upper left corner
Scale value,Represent present frame with reference frame for the difference of match block top left co-ordinate.Then these differences are averaging
Value, formula are as follows：
Represent x directions and y directions present frame with reference frame for the meansigma methodss of the difference of match block top left co-ordinate, N
Represent the quantity of match block.This method using surely as rear cutting by the way of, therefore cutting method is set for each cutting T up and down
Pixel (size of cutting pixel T can make corresponding change according to practical application).Work as differenceWithAbsolute value be respectively less than
During T pixels, using directly be fully compensated mode, will present frame x directions and y directions reverse movementWithDistance, then will
The image respectively dismisses T pixels up and down, forms steady as picture.WhenOrAbsolute value be more than T when, illustrate picture change
It is larger, then the match block of feature rich is extracted again in present frame, is searched for for next frame Block- matching, meanwhile, present frame is set
It is set to reference frame.Compensation way is changed into a Contrary compensation T pixel, is no longer compensate for beyond part.
(4) repeat step (2) and (3), until video terminates.
The present invention proposes a kind of Digital image stabilization method based on improved block matching algorithm, with reference to specific embodiment
The present invention is described in further detail.
Firstly, for the first frame, the in the picture a number of matching of diverse location selection of heart district domain (Fig. 2 is shown in definition)
Block, the size of match block are traditionally arranged to be 16*16 pixels, speed that the quantity and size of concrete match block can be as needed and
The balance of precision, is suitably adjusted.Meanwhile, KCF track algorithms are initialized, and allows KCF track algorithms to extract these respectively
The feature (extracting HOG features herein) of match block, with accurate tracking later.
For the second frame, the new of present frame match block place is calculated using block matching algorithm to each match block respectively
Position.Block matching algorithm principle such as Fig. 3.It is 16*16 pixels due to matching block size, and test video x, y direction peak excursion
Probably within 40 pixels, therefore (16+40*2) * (16+40*2) is selected around previous frame match block position in the current frame
Region, smallest match search is carried out in region of search according to minimum total absolute error criterion in the region, new is obtained
With block position.Next, the new position that block matching algorithm is obtained carries out KCF track algorithm calculating, match block is obtained accurate
Position.If the position of present frame match block is larger with the position difference of reference frame match block, a threshold value can be such as set,
X, y-coordinate difference are more than 20 pixels, then illustrate that picture is changed greatly, then match block is chosen in heart district domain again in the current frame, use
In the Block- matching of next frame.
According to result obtained in the previous step, corresponding with reference frame of all match block top left co-ordinates in present frame are calculated
Difference with block top left co-ordinate.Try to achieve the meansigma methodss of these differences WithX directions and y directions are represented respectively
Difference.This method using surely as rear cutting by the way of, this test cutting method is set to 20 pixels of each cutting up and down.When
DifferenceWithAbsolute value be respectively less than 20 pixel when, using directly full remuneration mode, will present frame in x directions and y directions
Reverse movement and distance, then the image is respectively dismissed T pixels up and down, forms steady as picture.WhenOrAbsolute value
During more than 20, illustrate that picture is changed greatly, then the match block of feature rich is extracted again in present frame, for next frame Block- matching
Search, meanwhile, present frame is set to into reference frame.Compensation way is changed into 20 pixel of a Contrary compensation, is no longer compensate for beyond part.
Ultimately produce the steady image as after.
Repeat previous step, until video terminates.
Jing is tested, and block matching algorithm combines KCF track algorithms, can be with point-device skew for calculating interframe, and this is right
Below steady as helping and very big.And it is last steady as effect is also very good.
Embodiments of the invention are the foregoing is only, the scope that the present invention relates to is not thereby limited, it is every using the present invention
Equivalent structure or equivalent flow conversion that description and accompanying drawing content are made, or directly or indirectly it is used in other related technologies
Field, is included within the spirit and scope of the claims of the present invention in the same manner.
Claims (3)
1. a kind of digital video digital image stabilization method based on improved block matching algorithm, is characterized in that, step is as follows：
(1) in the first frame is simultaneously also reference frame, the diverse location in heart district domain in the picture, the abundant matching of selected characteristic
Block, meanwhile, core correlation filtering KCF (the Kernelized Correlation Filters) algorithm for follow-up tracking is extracted
The feature of these match block, for tracking；
(2) for the second frame, match block is calculated in previous frame in the new position of present frame using block matching algorithm；Next
Operation KCF target tracking algorisms, the new position that block matching algorithm is obtained before by the tracking area update of KCF being, then at this
Calculating is tracked in new position, if the position that track algorithm is obtained is larger with the same match block position deviation of reference frame,
Present frame and reference frame difference are then illustrated than larger, that is, picture is changed greatly；
(3) result obtained according to (2nd) step, calculates corresponding with reference frame of all match block top left co-ordinates in present frame
Difference with block top left co-ordinate, formula are as follows：
The coordinate figure in the present frame match block upper left corner is represented,Represent the coordinate in the reference frame match block upper left corner
Value,Represent that present frame and reference frame, for the difference of match block top left co-ordinate, are then averaging to these differences
Value, formula are as follows：
X directions and y directions present frame and reference frame are represented for the meansigma methodss of the difference of match block top left co-ordinate, N is represented
The quantity of match block；Using it is steady as rear cutting by the way of, therefore cutting method is set for T pixel of each cutting up and down, is on duty
ValueWithAbsolute value be respectively less than T pixels when, using directly full remuneration mode, will present frame it is anti-in x directions and y directions
To movementWithDistance, then the image is respectively dismissed T pixels up and down, forms steady as picture；WhenOrIt is absolute
When value is more than T, illustrates that picture is changed greatly, then the match block of feature rich is extracted again in present frame, for next frame block
With search, meanwhile, present frame is set to into reference frame, compensation way is changed into a Contrary compensation T pixel, no longer mends beyond part
Repay；
(4) repeat step (2) and (3), until video terminates.
2. the digital video digital image stabilization method based on improved block matching algorithm as claimed in claim 1, is characterized in that, Block- matching
Algorithm calculate previous frame in match block in the new position of present frame, comprise the concrete steps that, if the length of match block a width of M and N, by
There is relative motion between sequence of frames of video, cause the position of the match block in a series of subsequent videos and the match block of former frame
Put and shift, it is assumed that maximum offset is respectively dx and dy on x, y direction, then select around match block in video present frame
One (M+2dx) * (N+2dy) region of search, carries out smallest match in region of search according to minimum total absolute error criterion and searches
Rope, obtains best matching blocks, then calculates the coordinate difference of matching result in match block and present frame frame in previous frame, to each
Match block carries out matching search once, obtains the new position of each match block.
3. the digital video digital image stabilization method based on improved block matching algorithm as claimed in claim 1, is characterized in that, step
(1), in, picture centre region is defined as：Inside frame a quarter length and width, the abundant match block of selected characteristic, match block
It is typically of size of 16*16 pixels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611032802.6A CN106559605A (en) | 2016-11-17 | 2016-11-17 | Digital video digital image stabilization method based on improved block matching algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611032802.6A CN106559605A (en) | 2016-11-17 | 2016-11-17 | Digital video digital image stabilization method based on improved block matching algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106559605A true CN106559605A (en) | 2017-04-05 |
Family
ID=58444902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611032802.6A Pending CN106559605A (en) | 2016-11-17 | 2016-11-17 | Digital video digital image stabilization method based on improved block matching algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106559605A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530322A (en) * | 2016-11-25 | 2017-03-22 | 天津大学 | Method for target tracking in jitter video |
| CN107749987A (en) * | 2017-09-30 | 2018-03-02 | 河海大学 | A kind of digital video digital image stabilization method based on block motion estimation |
| CN107977644A (en) * | 2017-12-18 | 2018-05-01 | 北京奇虎科技有限公司 | Image processing method and device, computing device based on image capture device |
| CN109087331A (en) * | 2018-08-02 | 2018-12-25 | 阿依瓦（北京）技术有限公司 | A kind of motion forecast method based on KCF algorithm |
| CN109544584A (en) * | 2018-11-30 | 2019-03-29 | 国网山东省电力公司信息通信公司 | It is a kind of to realize inspection surely as the method and system of precision measure |
| CN109712088A (en) * | 2018-12-14 | 2019-05-03 | 航天恒星科技有限公司 | A kind of remote sensing video satellite image processing method and system based on steady picture |
| CN110892354A (en) * | 2018-11-30 | 2020-03-17 | 深圳市大疆创新科技有限公司 | Image processing method and unmanned aerial vehicle |
| CN112150511A (en) * | 2020-11-02 | 2020-12-29 | 电子科技大学 | Target tracking algorithm based on combination of image matching and improved kernel correlation filter |
| CN114062265A (en) * | 2021-11-11 | 2022-02-18 | 易思维（杭州）科技有限公司 | Method for evaluating stability of supporting structure of visual system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004920A (en) * | 2010-11-12 | 2011-04-06 | 浙江工商大学 | Method for splitting and indexing surveillance videos |
-
2016
- 2016-11-17 CN CN201611032802.6A patent/CN106559605A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102004920A (en) * | 2010-11-12 | 2011-04-06 | 浙江工商大学 | Method for splitting and indexing surveillance videos |
Non-Patent Citations (3)
| Title |
|---|
| 宋李亚: "基于块匹配的数字视频稳像系统", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 张国栋: "基于电子稳像技术的视频稳像研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 赵菲: "视频稳像技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530322A (en) * | 2016-11-25 | 2017-03-22 | 天津大学 | Method for target tracking in jitter video |
| CN106530322B (en) * | 2016-11-25 | 2020-04-17 | 天津大学 | Method for tracking target in jittering video |
| CN107749987A (en) * | 2017-09-30 | 2018-03-02 | 河海大学 | A kind of digital video digital image stabilization method based on block motion estimation |
| CN107749987B (en) * | 2017-09-30 | 2020-09-18 | 河海大学 | Digital video image stabilization method based on block motion estimation |
| CN107977644A (en) * | 2017-12-18 | 2018-05-01 | 北京奇虎科技有限公司 | Image processing method and device, computing device based on image capture device |
| CN107977644B (en) * | 2017-12-18 | 2021-07-23 | 北京奇虎科技有限公司 | Image data processing method and device based on image acquisition equipment and computing equipment |
| CN109087331A (en) * | 2018-08-02 | 2018-12-25 | 阿依瓦（北京）技术有限公司 | A kind of motion forecast method based on KCF algorithm |
| CN109544584A (en) * | 2018-11-30 | 2019-03-29 | 国网山东省电力公司信息通信公司 | It is a kind of to realize inspection surely as the method and system of precision measure |
| CN110892354A (en) * | 2018-11-30 | 2020-03-17 | 深圳市大疆创新科技有限公司 | Image processing method and unmanned aerial vehicle |
| CN109712088A (en) * | 2018-12-14 | 2019-05-03 | 航天恒星科技有限公司 | A kind of remote sensing video satellite image processing method and system based on steady picture |
| CN112150511A (en) * | 2020-11-02 | 2020-12-29 | 电子科技大学 | Target tracking algorithm based on combination of image matching and improved kernel correlation filter |
| CN114062265A (en) * | 2021-11-11 | 2022-02-18 | 易思维（杭州）科技有限公司 | Method for evaluating stability of supporting structure of visual system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106559605A (en) | Digital video digital image stabilization method based on improved block matching algorithm | |
| CN109522850B (en) | Action similarity evaluation method based on small sample learning | |
| WO2018161775A1 (en) | Neural network model training method, device and storage medium for image processing | |
| CN101945284B (en) | Motion estimation device and method | |
| CN105261037B (en) | A kind of moving target detecting method of adaptive complex scene | |
| CN102741884B (en) | Moving body detecting device and moving body detection method | |
| Jiang et al. | Video object detection with locally-weighted deformable neighbors | |
| CN104219533B (en) | A kind of bi-directional motion estimation method and up-conversion method of video frame rate and system | |
| CN103824272B (en) | The face super-resolution reconstruction method heavily identified based on k nearest neighbor | |
| CN108171734B (en) | ORB feature extraction and matching method and device | |
| CN105872345A (en) | Full-frame electronic image stabilization method based on feature matching | |
| CN109191511A (en) | A kind of binocular solid matching process based on convolutional neural networks | |
| CN107633482A (en) | A kind of super resolution ratio reconstruction method based on sequence image | |
| CN116524062B (en) | Diffusion model-based 2D human body posture estimation method | |
| CN114663509B (en) | Self-supervision monocular vision odometer method guided by key point thermodynamic diagram | |
| CN106887010A (en) | Ground moving target detection method based on high-rise scene information | |
| CN103379255B (en) | Image processor and its processing method | |
| CN102088544B (en) | Fast image stabilization method of dynamic scene video with foreground object | |
| CN110363197A (en) | Based on the video area-of-interest exacting method for improving visual background extraction model | |
| CN101572820A (en) | Preprocessing method of video signal in detection process of moving target | |
| CN104182931B (en) | Super resolution method and device | |
| CN104751487A (en) | Method for detecting movement target based on colored RGB three-pane color-change frame difference | |
| CN111292357B (en) | Video inter-frame rapid motion estimation method based on correlation filtering | |
| CN110827324B (en) | Video target tracking method | |
| CN102917222B (en) | Mobile background video object extraction method based on self-adaptive hexagonal search and five-frame background alignment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170405 |
|
| WD01 | Invention patent application deemed withdrawn after publication |