CN103200392A - Vehicle-mounted video image stabilizing method based on viewing angle conversion - Google Patents
Vehicle-mounted video image stabilizing method based on viewing angle conversion Download PDFInfo
- Publication number
- CN103200392A CN103200392A CN2013101101908A CN201310110190A CN103200392A CN 103200392 A CN103200392 A CN 103200392A CN 2013101101908 A CN2013101101908 A CN 2013101101908A CN 201310110190 A CN201310110190 A CN 201310110190A CN 103200392 A CN103200392 A CN 103200392A
- Authority
- CN
- China
- Prior art keywords
- image
- carried out
- vehicle
- current dither
- method based
- 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
- Image Processing (AREA)
Abstract
The invention belongs to the technical field of video processing, and provides a vehicle-mounted video image stabilizing method based on viewing angle conversion. According to the method, the viewing angle conversion is firstly carried out before image stabilizing of a vehicle-mounted video is processed. The method includes that a rectangle is determined in an image without trapezoidal distortion, and according to size of change of a viewing angle, a corresponding trapezoid after the rectangle is distorted in an trapezoidal distortion image is determined; a transmission transformation matrix is calculated; viewing angle conversion of each frame of vehicle-mounted video images is carried out according to the transmission transformation matrix; a block to be matched, a matching criterion and a search strategy of a current dithering image are selected; motion estimation is carried out on the current dithering image; and an overall situation motion vector is calculated, and motion compensation is carried out. The vehicle-mounted video image stabilizing method not only reduces interframe offset of an image sequence caused by car body vibration, but also calibrates image trapezoidal distortion cause by camera lens angles, and is suitable for a vehicle-mounted intelligent video surveillance system.
Description
Technical field
Present invention relates in general to a kind of Vehicular video image stabilization method based on view transformation, relate more specifically to a kind of transmission transformation matrix that utilizes image is carried out keystone, reduce the Vehicular video image stabilization method based on view transformation of image sequence interframe skew afterwards by the piece estimation.
Background technology
Along with the continuous development of vehicle intelligent video monitoring, human detection and tracking technique on the bus receive increasing concern, how to improve human detection and matching precision, become the focus of people's research.How to proofread and correct the image trapezoidal distortion that the camera angle causes, how to eliminate the bigger skew between the sequence of video images that body oscillating causes, these all are to need the problem that solves.
Video surely comprises overall motion estimation and two parts of motion compensation as technology, and overall motion estimation is key wherein.At present typical a kind of overall motion estimation algorithm based on the piece coupling is interative least square method, and this method is got rid of the interference of local motion with a threshold value, determines the regional and final mask parameter of global motion by iteration.This method amount of calculation is too big, can not satisfy the requirement of vehicle-mounted monitoring real-time.Also having a kind of method is to avoid the interference of local motion target by four sub-pieces choosing the image peripheral regions, with the average of four sub-piece local motion vectors as global motion vector.Though this method is simple, can not adapt to the complex situations in the compartment fully, when the compartment was comparatively crowded, the passenger appeared at the lower left corner and the zone, the lower right corner of image possibly, therefore can reduce the accuracy of overall motion estimation.
And existing Vehicular video does not surely take into account the trapezoidal distortion that the camera angle causes as technology, can not guarantee the precision of human detection and coupling in the vehicle intelligent video monitoring system preferably.
Summary of the invention
The present invention is intended to overcome the above-mentioned deficiency on the prior art, provide a kind of simply, fast and effectively based on the Vehicular video image stabilization method of view transformation.
In order to achieve the above object, technical scheme of the present invention is:
A kind of Vehicular video image stabilization method based on view transformation is characterized in that, Vehicular video is being carried out surely carried out view transformation earlier before handling, and specifically comprises following steps:
1) in the image of no trapezoidal distortion, determines a rectangle, according to the visual angle change size, in the trapezoidal distortion image, determine after this rectangle distortion corresponding trapezoidal;
2) with rectangle and trapezoidal four summits as four groups of corresponding points, obtain the transmission transformation matrix;
3) according to the transmission transformation matrix each frame Vehicular video image is carried out view transformation;
4) choose to be matched of current dither image, matching criterior and search strategy;
5) current dither image is carried out estimation based on piece coupling;
6) ask the global motion vector of current dither image;
7) current dither image is carried out motion compensation.
As preferred implementation, in the step 1), the visual angle change size equals vehicle-mounted pick-up head optical axis and horizontal direction angle, visual angle change size
Wherein, a, b represent width and the height of bus interior respectively; In the step 4), choose two macro blocks in the upper left corner and the upper right corner of current dither image as to be matched; Choose SAD in the step 4) as the piece matching criterior, rhombus therapy is as search strategy; In the step 6), two to be matched local motion vector and the global motion vector of former frame image in the current dither image are averaged, as the global motion vector of current dither image.
The present invention has following advantage:
(1) the present invention utilizes four groups of corresponding points to calculate the transmission transformation matrix, and then the Vehicular video image is carried out view transformation, has proofreaied and correct the trapezoidal distortion of image, and is consuming time few under the prerequisite that guarantees picture quality, and higher execution efficient is arranged.
(2) the present invention choose current dither image two macro blocks in the upper left corner and the upper right corner as to be matched, the global motion vector of two to be matched local motion vector and former frame image is averaged, as the global motion vector of current dither image.So both improve steady picture speed greatly, guaranteed the accuracy of overall motion estimation again.
(3) the present invention chooses first two field picture of Vehicular video as the reference picture of all successive images, can effectively avoid reference picture to substitute caused accumulated error; And directly use global motion vector as motion compensation vector, algorithm speed is conducive to improve the whole efficiency of the steady picture of video than very fast.
(4) the present invention is simple, is easy to realize at DSP, is applicable to that camera is fixed and the vehicle intelligent video monitoring system of no-raster campaign, and is practical.
Description of drawings
Fig. 1 realizes the software flow block diagram of the inventive method.
Fig. 2 is the view transformation angle.
Fig. 3 is to be matched choosing.
Embodiment
At first referring to Fig. 1, technical scheme of the present invention is further specified below.
In the image of no trapezoidal distortion, determine a rectangle, according to the visual angle change size, in the trapezoidal distortion image, determine after this rectangle distortion corresponding trapezoidal, with rectangle and trapezoidal four summits as four groups of corresponding points, obtain the transmission transformation matrix, and then each frame video image carried out view transformation, obtain not having the sequence of video images of trapezoidal distortion.Choose two macro blocks in the upper left corner of current dither image and the upper right corner as to be matched, by current dither image being carried out estimation and the motion compensation based on the piece coupling, reduce the interframe skew of the image sequence that body oscillating causes.Concrete steps are as follows:
1) calculate the visual angle change size, i.e. the angle theta of vehicle-mounted pick-up head optical axis and horizontal direction, as shown in Figure 2.The position of S point expression camera, a represents box width, b represents height of wagon,
The angular field of view of expression camera, then
It is the arc tangent that the angular field of view of camera equals the ratio of box width and height.Get
Angular bisector as the camera optical axis direction, then the computing formula of θ is as the formula (1).
2) in the image of no trapezoidal distortion, determine a rectangle, according to the visual angle change size, in the trapezoidal distortion image, determine after this rectangle distortion corresponding trapezoidal.
3) with rectangle and trapezoidal four summits as four groups of corresponding points, calculate the transmission transformation matrix.
4) utilize the transmission transformation matrix that each frame video image is carried out view transformation, as the formula (2).Wherein, P (x, y) and P ' (x, y) represent before the view transformation respectively and view transformation after the point coordinates of image, M is 3 * 3 transmission transformation matrix.
5) two macro blocks in the upper left corner and the upper right corner of choosing current dither image are as to be matched, and the SAD(minimum is absolute difference always) as the piece matching criterior, rhombus therapy is as search strategy.To be matched choose as shown in Figure 3.
6) first two field picture of choosing Vehicular video is as the reference image, and current dither image is carried out estimation based on the piece coupling.
7) global motion vector of the current dither image of calculating, computing formula as the formula (3).Wherein
Be respectively two to be matched local motion vector in the current dither image,
It is respectively the global motion vector of current dither image and former frame image.
8) current dither image is carried out motion compensation, namely according to global motion vector, the pixel of current dither image along the reverse mobile equivalent of this motion vector, be offset to eliminate or to alleviate the interframe of shaking the sequence of video images that causes.
For making purpose of the present invention, implementation and advantage more clear, below concrete enforcement of the present invention is described in further detail.
1) general bus interior width and highly be respectively 2m and 2.5m according to formula (1), obtains the view transformation angle and is about 70 °.Wherein, a, b represent width and the height of bus interior respectively,
It is the arc tangent of the ratio of box width and height.
2) in the image of no trapezoidal distortion, determine a rectangle, according to the visual angle change size, in the trapezoidal distortion image, determine after this rectangle distortion corresponding trapezoidal.
3) with rectangle and trapezoidal four summits as four groups of corresponding points, calculate the transmission transformation matrix.
4) utilize the transmission transformation matrix that each frame video image is carried out view transformation, as the formula (2).Wherein, P (x, y) and P ' (x, y) represent before the view transformation respectively and view transformation after the point coordinates of image, M is 3 * 3 transmission transformation matrix.
5) gather the continuous dither image sequence of 320*240 with the frame rate of 25fps, each dither frame image all is divided into the sub-piece of 16*16, non-overlapping copies, two sub-pieces are chosen as to be matched in (24,24) and (296,24) position at current dither image.
6) choose first two field picture of Vehicular video as the reference image, in the matching window (size is 30*30) of reference picture, press in SAD matching criterior search and the current dither image to be matched immediate, obtain in the current dither image to be matched local motion vector.
7) two to be matched local motion vector and the global motion vector of former frame image in the current dither image are averaged, as the global motion vector of current dither image.Computing formula as the formula (3).Wherein
Be respectively two to be matched local motion vector in the current dither image,
It is respectively the global motion vector of current dither image and former frame image.
8) use global motion vector as motion compensation vector, current dither image is carried out motion compensation, the video image after obtaining shaking.
Claims (5)
1. the Vehicular video image stabilization method based on view transformation is characterized in that, Vehicular video is being carried out surely carried out view transformation earlier before handling, and comprises following steps:
1) in the image of no trapezoidal distortion, determines a rectangle, according to the visual angle change size, in the trapezoidal distortion image, determine after this rectangle distortion corresponding trapezoidal;
2) with rectangle and trapezoidal four summits as four groups of corresponding points, obtain the transmission transformation matrix;
3) according to the transmission transformation matrix each frame Vehicular video image is carried out view transformation;
4) choose to be matched of current dither image, matching criterior and search strategy;
5) current dither image is carried out estimation based on piece coupling;
6) ask the global motion vector of current dither image;
7) current dither image is carried out motion compensation.
2. the described Vehicular video image stabilization method based on view transformation of claim 1 is characterized in that, in the step 1), the visual angle change size equals vehicle-mounted pick-up head optical axis and horizontal direction angle, the visual angle change size, wherein, a, b represent width and the height of bus interior respectively.
3. the described Vehicular video image stabilization method based on view transformation of claim 1 is characterized in that, in the step 4), chooses two macro blocks in the upper left corner and the upper right corner of current dither image as to be matched.
4. the described Vehicular video image stabilization method based on view transformation of claim 1 is characterized in that can choose SAD in the step 4) as the piece matching criterior, rhombus therapy is as search strategy.
5. the described Vehicular video image stabilization method based on view transformation of claim 1, it is characterized in that, in the step 6), two to be matched local motion vector and the global motion vector of former frame image in the current dither image are averaged, as the global motion vector of current dither image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101101908A CN103200392A (en) | 2013-03-29 | 2013-03-29 | Vehicle-mounted video image stabilizing method based on viewing angle conversion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013101101908A CN103200392A (en) | 2013-03-29 | 2013-03-29 | Vehicle-mounted video image stabilizing method based on viewing angle conversion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103200392A true CN103200392A (en) | 2013-07-10 |
Family
ID=48722733
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013101101908A Pending CN103200392A (en) | 2013-03-29 | 2013-03-29 | Vehicle-mounted video image stabilizing method based on viewing angle conversion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103200392A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104469251A (en) * | 2013-09-23 | 2015-03-25 | 联想(北京)有限公司 | Image acquisition method and electronic equipment |
CN107852482A (en) * | 2015-08-05 | 2018-03-27 | 株式会社电装 | Calibrating installation, calibration method and calibration procedure |
CN107852462A (en) * | 2015-07-22 | 2018-03-27 | 索尼公司 | Camera model, solid-state imager, electronic equipment and image capture method |
CN114007054A (en) * | 2022-01-04 | 2022-02-01 | 宁波均联智行科技股份有限公司 | Method and device for correcting projection of vehicle-mounted screen picture |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101383899A (en) * | 2008-09-28 | 2009-03-11 | 北京航空航天大学 | Video image stabilizing method for space based platform hovering |
-
2013
- 2013-03-29 CN CN2013101101908A patent/CN103200392A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101383899A (en) * | 2008-09-28 | 2009-03-11 | 北京航空航天大学 | Video image stabilizing method for space based platform hovering |
Non-Patent Citations (3)
Title |
---|
A.ENGELSBERG AND G.SCHMIDT: "A COMPARATIVE REVIEW OF DIGITAL IMAGE STABILISING ALGORITHMS FOR MOBILE VIDEO COMMUNICATIONS", 《IEEE TRANSACTIONS ON CONSUMCR ELECTRONICS》, vol. 45, no. 3, 31 August 1999 (1999-08-31) * |
刘艳: "基于动态偏移场模型的视频稳定化技术研究", 《中国博士学位论文全文数据库信息科技辑》, no. 10, 15 October 2008 (2008-10-15) * |
张跃飞: "车载摄像机数字稳像技术研究", 《中国博士学位论文全文数据库信息科技辑》, no. 12, 15 December 2011 (2011-12-15), pages 29 - 47 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104469251A (en) * | 2013-09-23 | 2015-03-25 | 联想(北京)有限公司 | Image acquisition method and electronic equipment |
CN104469251B (en) * | 2013-09-23 | 2018-07-06 | 联想(北京)有限公司 | Image-pickup method and electronic equipment |
CN107852462A (en) * | 2015-07-22 | 2018-03-27 | 索尼公司 | Camera model, solid-state imager, electronic equipment and image capture method |
CN107852462B (en) * | 2015-07-22 | 2020-12-18 | 索尼公司 | Camera module, solid-state imaging element, electronic apparatus, and imaging method |
CN107852482A (en) * | 2015-08-05 | 2018-03-27 | 株式会社电装 | Calibrating installation, calibration method and calibration procedure |
CN114007054A (en) * | 2022-01-04 | 2022-02-01 | 宁波均联智行科技股份有限公司 | Method and device for correcting projection of vehicle-mounted screen picture |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103177439B (en) | A kind of automatic calibration method based on black and white lattice corners Matching | |
EP2546602B1 (en) | Stereo camera apparatus | |
US10432847B2 (en) | Signal processing apparatus and imaging apparatus | |
CN104144282A (en) | Fast digital image stabilization method applicable to space robot visual system | |
CN102799857B (en) | Video multi-vehicle outline detection method | |
US10148938B2 (en) | Vehicle-mounted image recognition device to set a stereoscopic-vision and monocular-vision image areas | |
CN102915525B (en) | Method for correcting barrel distortion of images for full-frame fish-eye lenses | |
CN103200392A (en) | Vehicle-mounted video image stabilizing method based on viewing angle conversion | |
CN103458261B (en) | Video scene variation detection method based on stereoscopic vision | |
CN110398979B (en) | Unmanned engineering operation equipment tracking method and device based on vision and attitude fusion | |
CN104512328A (en) | Automobile all-round looking image generation method and automobile all-round looking system | |
CN104660977B (en) | Eye view image generating means | |
CN109917359B (en) | Robust vehicle distance estimation method based on vehicle-mounted monocular vision | |
WO2022096027A1 (en) | Garage space tracking method and apparatus | |
EP3438603A1 (en) | Road surface displacement detection device and suspension control method | |
WO2019156072A1 (en) | Attitude estimating device | |
CN102855637A (en) | Covariance tracking method for fusing bilateral filtering in video images | |
CN103253194A (en) | Traveling vehicle auxiliary system | |
CN103002297A (en) | Method and device for generating dynamic depth values | |
US20210046870A1 (en) | Image processing apparatus, imaging apparatus, driving assistance apparatus, mobile body, and image processing method | |
JP5305750B2 (en) | Vehicle periphery display device and display method thereof | |
CN102096912A (en) | Method and device for processing image | |
CN111626227B (en) | Implementation method of binocular vision-based vehicle bottom perspective panoramic system | |
JP7303064B2 (en) | Image processing device and image processing method | |
CN105323447B (en) | More fish eye images processing methods, device and vehicle |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130710 |