CN202931463U - Characteristic block based video image stabilization device - Google Patents

Abstract

The utility model discloses a characteristic block based video image stabilization device, which comprises an image input module. The image input module is connected with a singlechip. An image display is connected to the singlechip. Electronic image stabilization is realized by image stabilization algorithm: firstly, offset of a tracking block is acquired from a source image; secondly, the image is subjected to separation process by the algorithm so as to determine a motion vector; thirdly, the offset is compensated through a motion compensation system; and finally, a stable image is outputted. The characteristic block based video image stabilization device can be widely used in consumer electronic products such as mobile phones, digital cameras, vidicons and the like, and simultaneously can also be widely applied in video image stabilization in a camera and monitoring system of other equipment. It shows through experimental results that the characteristic block based video image stabilization device has a good image stabilization effect.

Description

Video Stabilization device based on characteristic block

Technical field

The utility model belongs to digital image stabilization method and the device of electronic technology field, be specifically related to a kind of Video Stabilization device based on characteristic block, this device utilizes computer technology that real-time video image is processed, the video of one section float is processed the video image of stable output picture.

Background technology

Video Stabilization surely develops into present electronic steady image as the steady picture of, optics from early stage mechanical type, and the technology that surely looks like is towards more convenient, more succinct, more effective future development.Mechanical type similarly is surely to use dynamo-electric method, realizes image stabilization by instrument, this system operation inconvenience, and subsequent treatment is also more difficult.The steady picture of optics is the method for utilizing in some optics, realizes the compensation of image by some optical elements are set.Along with the progress of science and technology, the development of electronic computer technology, the appliance computer image processes to reach the technology of stable video image, and more convenient operation, process be more succinct, surely look like better effects if.

The electric video image stabilization system mainly comprises estimation, motion compensation and three modules of steady picture video output.Core technology is the design of algorithm, and Chinese scholars has been studied the algorithm of many relevant estimation, optical flow method is arranged, based on the method for aerial image vegetarian refreshments gray scale, based on the space characteristics method, based on methods such as bit plane couplings; Aspect motion compensation, have based on image and amplify, based on methods such as video reparations, but these methods still could not be accomplished good compromise on speed and effect, such as: aspect estimation, block-based method speed is fast, but be subjected to illumination effect very large, based on the characteristic point algorithm, have good anti-lighting effect, but too responsive to local motion.

Summary of the invention

The purpose of this utility model is, a kind of Video Stabilization device of following the trail of based on characteristic block is provided, and this device can reduce the high request to picture pick-up device, improves picture pick-up device to the ability of Video Stabilization under complex environment.

In order to realize above-mentioned task, the utility model is taked following technical solution:

A kind of Video Stabilization device of following the trail of based on characteristic block is characterized in that, comprises, and image input module, this image input module connects a single-chip microcomputer, is connected with image display panel on single-chip microcomputer.

Other characteristics of the present utility model are:

In described single-chip microcomputer, the motion estimation module, motion compensating module and the image reorganization module that have at least software form to realize.

Described image input module is mobile phone, digital camera or video camera.

The Video Stabilization device of following the trail of based on characteristic block of the present utility model, that utilizes single-chip microcomputer surely realizes electronic steady image as algorithm, at first obtain the side-play amount of following the trail of piece from source images, then utilize algorithm to carry out separating treatment to image and determine motion vector, then by motion compensating system, side-play amount is compensated the image of final stable output.Can be widely used in the consumable electronic products such as mobile phone, digital camera, video camera, the while also can be widely used in the shooting of other equipments and the Video Stabilization in surveillance, shows through experimental result, has surely to look like preferably effect.

Description of drawings

Fig. 1 is hardware block diagram of the present utility model;

Fig. 2 is based on the schematic diagram that characteristic block is followed the trail of steady picture;

Fig. 3 is the schematic diagram of estimation tracing algorithm;

Fig. 4 is based on the characteristic block search graph; Wherein, (a) being t-1 frame search figure, is (b) t frame search figure.

Fig. 5 is characteristic block search routine figure;

Fig. 6 is the movement compensating algorithm flow chart.

Below in conjunction with drawings and Examples, the utility model is described in further detail.

Embodiment

As shown in Figure 1, the present embodiment provides a kind of Video Stabilization device of following the trail of based on characteristic block, comprises image input module, and this image input module connects a single-chip microcomputer, is connected with image display panel on single-chip microcomputer.

In single-chip microcomputer, the motion estimation module, motion compensating module and the image reorganization module that have at least software form to realize.

In the present embodiment, image input module is mobile phone, digital camera or video camera.

The course of work of motion estimation module, motion compensating module and image reorganization module that software is realized is carried out as follows:

The first step is carried out piece to present frame and is cut apart;

Second step carries out tracing movement take the characteristic error minimum of every as matching criterior to every characteristic block and estimates;

In the 3rd step, add up described motion vector and obtain present frame with respect to the globe motion parameter of former frame;

In the 4th step, carry out motion compensation according to parameter.

Due to motion estimation module motion compensating module and the image reorganization module based on characteristic block that is realized by software being installed in single-chip microcomputer, wherein, at first this motion estimation module is divided into some with current frame image, and every is mated in specific hunting zone in reference frame.Motion compensating module adopts the criterion of mean square error, and this criterion has fast operation, and computational accuracy loudness is higher, has strengthened the steady picture ability of picture pick-up device.In addition, need to not judge based on the movement tendency of these kinematic parameters to each piece after acquisition piece kinematic parameter.after the judgement active block, need utilize the kinematic parameter of these pieces to estimate globe motion parameter, classical way is to adopt least square method to find the solution an overdetermined equation, but need to ask matrix inversion in solution procedure, the dimension of this matrix is generally all higher, therefore computing is more complicated, and might matrix to occur irreversible due to the computer numerical problem, cause to obtain globe motion parameter, estimate and utilize in the single-chip microcomputer of device of the present utility model the translation that each piece of present frame produces with respect to the former frame piece, then, adopt the method for average statistical to obtain global parameter.

as shown in Figure 2, at first single-chip microcomputer imports by Simulink and needs video to be processed, pass through estimation, calculate the motion vector between present frame and reference frame, through motion compensation, use the data of motion vector to obtain whether present frame being made progress translation or translation downwards, come each frame of Recovery image sequence " originally " appearance, remove undesired motion according to the result of estimation in other words conj.or perhaps, make the observer appear to have the sensation that is eliminated or weakens of rocking of video camera, then the image sequence after whole processing is recombinated, obtain the more stable new video of new image, export from image display panel at last.

Described motion estimation algorithm flow chart as shown in Figure 3, the motion of object is made of multiple element, comprises translation, rotation and distortion etc., the model of intactly describing object of which movement is complicated.So suppose that motion is to be made of translation, so only with two translation parameters dx of X and Y-direction (or level and vertical direction), dy characterizes kinematic parameter.By dx, dy consists of a motion vector D=(dx, dy) ^T

The pixel value of present frame coordinate points (i, j) is predicted as by former frame:

$\mathrm{MCP}(i,j,k)=\stackrel{\‾}{S}(i,j,k)=S(i+\mathrm{dx},j+\mathrm{dy},k-1)$

Here k represents present frame, and k-1 represents former frame, and S (i, j, k) represents original pixel value,

The pixel value of expression prediction.

Between current frame original image and predicted picture, error may be arranged, because this error is caused by the translational motion compensation prediction, therefore be called interframe displacement error DFD, be expressed as follows:

$\mathrm{DFD}(i,j,k)=S(i,j,k)-\stackrel{\‾}{S}(i,j,k)=S(i,j,k)-\stackrel{\‾}{S}(i+\mathrm{dx},j+\mathrm{dy},k-1)$

These results have been arranged, the coding of pixel in picture frame has been become coding to DFD field and motion vector D.

In the present embodiment, the translation criterion adopts algorithm mean square error fast and effectively, and this mean square error formula is:

$\mathrm{MSE}({\mathrm{\δ}}_x,{\mathrm{\δ}}_y)=\frac{1}{\mathrm{mn}}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j=1}{\overset{m}{\mathrm{\Σ}}}{[I_t(i,j)-I_{t-1}(i+{\mathrm{\δ}}_x,y+{\mathrm{\δ}}_y)]}^2$

In formula, (δ _x, δ _y) be displacement vector.If at a certain position MSE (δ _x, δ _y) reach minimum, the Optimum Matching point that will look for exactly of this point, I _t(i, j) and I _t-1(i+ δ _x, y+ δ _y) be respectively the gray value of present frame and previous frame.

Search procedure is illustrated in fig. 5 shown below, and after having determined matching criterior, will obtain best matching blocks by motion search, and the searching route of determining according to searching algorithm during motion search is sought the motion vector of best matching blocks and coupling.Seeking best matching blocks need to be at each pixel (n ₁, n ₂) optimize the translation algorithm criterion on all possible candidate's displacement vector.The present embodiment does not adopt so-called " thorough search " to realize, in order to reduce computation burden, restricted searching area is at-M usually ₁≤ d ₁≤ M ₁And-M ₂≤ d ₂≤ M ₂" search window " is centrally located in motion vector with on each estimated pixel, wherein M ₁And M ₂It is the integer of being scheduled to.

The digital image stabilization method of " based on characteristic block " realize thought as shown in Figure 4, in selected video image, a certain size zone as " characteristic block ", is exactly " optic centre " of location.Make every effort to reach this " optic centre " significantly rocking in video image of elimination." characteristic block " as search module, adopted the purpose based on the steady picture of full search rule realization character piece target of template.

Set up coordinate system take the video image upper left corner as the origin of coordinates, characteristic block is that a upper left corner coordinate is (x ₀, y ₀), length and width are respectively n, the rectangular area of m.The region of search is that upper left corner coordinate is (x ₁, y ₁), edge and characteristic block distance are respectively a, b, and length and width are respectively n+2a, the rectangular area of m+2b." based on characteristic block " be surely namely to draw the motion vector of present frame by " characteristic block " implemented fast and effectively steady picture in the region of search as algorithm.Picture frame carries out up and down according to this displacement parameter or the translation (motion compensation) of left and right.Then, then with present frame repeat this algorithm as the reference frame.

Shown in movement compensating algorithm flow process See Figure 6, the purpose of motion compensation keeps certain estimation to record to produce the stabilized image sequence exactly.The result of motion estimation process noted earlier can calculate the motion of two interframe, in order to produce a stable image sequence, must carry out calibration transfer from the present frame to the reference frame, and motion compensating module is estimated in conjunction with interframe movement, draws this integral transformation.Following affine transformation is used for proofreading and correct f _t-1Produced corresponding stable

${\stackrel{\‾}{f}}_{t-1}=R_{t-1}{f}_{t-1}+T_{t-1}$

The interframe conversion is rotated by interframe

And translation

Definition:

$f_{t-1}=R_t^i{f}_t+T_t^i$

Get in conjunction with above two formula:

${\stackrel{\‾}{f}}_t=R_{t-1}{R}_t^i{f}_t+R_{t-1}{T}_t^i+T_{t-1}$

So new rotation and translational movement are by providing with following formula:

$R_t=R_{t-1}{R}_t^i;$ $T_t=R_{t-1}{T}_t^i+T_{t-1}.$

Claims (2)

1. the Video Stabilization device based on characteristic block, is characterized in that, comprises image input module, and this image input module connects a single-chip microcomputer, is connected with image display panel on single-chip microcomputer.

2. the Video Stabilization device based on characteristic block as claimed in claim 1, is characterized in that, described image input module is mobile phone, digital camera or video camera.

Images