CN101729763A - Electronic image stabilizing method for digital videos - Google Patents

Abstract

The invention relates to an electronic image stabilizing method for digital videos, belonging to the technical field of digital image processing. The method comprises the following steps of: firstly, extracting characteristic points in two adjacent images in a video sequence by utilizing an SUSAN corner detection operator, matching the characteristic points by utilizing a relevant matching method and rejecting false matching point pairs; then selecting three pairs from the correct matching point pairs and solving for an affine transformation parameter between two images by utilizing a corresponding coordinate relation of the matching point pairs on a two-dimensional image; and finally, carrying out an affine inverse transformation on a next image according to an affine formula to obtain an image registered with a previous image in space so as to realize the electronic image stabilization of the digital videos. The electronic image stabilizing method of the digital videos can automatically eliminate jitters of the video images in real time, wherein the jitters comprise translations and rotating movements among images.

Description

Electronic image stabilizing method for digital videos

Technical field

The present invention relates to the digital image processing techniques field, relate in particular to a kind of electronic image stabilizing method for digital videos, specifically eliminate the method for the violent shaking phenomenon of video image that causes owing to DE Camera Shake automatically.

Background technology

Along with the arrival of digital Age and a large amount of appearances of digital product, digital camera has been widely used in the every field of military project, space flight and people's daily life.Digital camera is when being used, because artificial rock or the vibrations of vehicle equipment all can cause the shake and the rotation of capture video, this just makes that the observer is easy to generate visual fatigue.

At present, more common surely being based on the steady picture of machinery as the equipment majority, is the mechanical shaking of vehicle equipment is installed is realized that this equipment manufacturing cost height and low precision can not satisfy the demands by reducing or offset video camera.And common electronic steady image equipment majority can only be eliminated the translation shake of video and can not eliminate the rotation of video, and its video output also has bigger delay, does not reach real-time requirement.

Summary of the invention

The purpose of this invention is to provide a kind of electronic image stabilizing method for digital videos, it can eliminate the shake of video image automatically, in real time.

In order to achieve the above object, technical scheme of the present invention is as follows:

Electronic image stabilizing method for digital videos comprises the steps: at first, utilizes the characteristic point in adjacent two width of cloth images in the SUSAN Corner Detection operator extraction video sequence, utilizes the relevant matches method to mate these characteristic points, and the match point of rejecting mistake is right; Then, choose three pairs, utilize match point, try to achieve the affine transformation parameter between two width of cloth images the relation of the respective coordinates on two dimensional image from correct match point centering; At last, the back piece image is carried out affine inverse transformation, obtain and the preceding piece image image of registration spatially, thereby finish the method for digital video electronic steady image according to affine formula.

Video jitter is exactly the displacement that produces between adjacent two two field pictures in video sequence, and the key of electronic steady image of the present invention is exactly the shift transformation relation of finding out between this two width of cloth image, carry out inverse transformation according to this relation then, registration two width of cloth images again reach the purpose of steady picture.

The invention has the beneficial effects as follows: electronic image stabilizing method for digital videos of the present invention can be eliminated the shake of video image automatically, in real time, comprises the translation between image and rotatablely moves.

Embodiment

Below in conjunction with specific embodiment the present invention is done description in further detail:

Electronic image stabilizing method for digital videos of the present invention is a kind ofly to determine transformation relation between two width of cloth images based on Feature Points Matching and affine transformation, and then registration two width of cloth images are realized the method for steady picture again, and the specific implementation process of this method is as follows:

Suppose f ₁, f ₂Be the adjacent image of any two frames, wherein f in the video sequence ₁Be former frame image, i.e. reference picture; f ₂Be back one two field picture, image promptly subject to registration, then f ₁And f ₂Between registration just become f ₂Through spatial alternation and f ₁The process of registration.The transformation for mula that can describe this spatial alternation has a variety of, and six parameter affine transformations are the most accurately and effectively a kind of.This six parameters affine transformation is described below:

Suppose image f ₁Be reference picture, f ₂Be image subject to registration, and p (x, y) and q (x ', y ') be respectively image f ₁And f ₂In the coordinate of two corresponding points because f ₁And f ₂Between global motion show as translation and, can represent with linear transformation around the rotation of optical axis, so it is as follows to define six parameter affine models:

$\left[\begin{array}{c}{x}^{\′}\\ y^{\′}\end{array}\right]=\left[\begin{array}{cc}{a}_{11}& a_{12}\\ a_{21}& a_{22}\end{array}\right]\left[\begin{array}{c}x\\ y\end{array}\right]+\left[\begin{array}{c}{b}_1\\ b_2\end{array}\right]---\left(1\right)$

In the formula, B=(b ₁, b ²) ^TBe translation parameters, $A=\left[\begin{array}{cc}{a}_{11}& a_{12}\\ a_{21}& a_{22}\end{array}\right]$ Be parameter matrix, a _Ij, b _iBe real number.Only require a ₁₁, a ₁₂, a ₂₁, a ₂₂, b ₁, b ₂These six parameters just can be determined image f ₁And f ₂Between spatial alternation relation, according to this spatial alternation relation to image f ₂Carry out inverse transformation and can reach the purpose of steady picture.

According to the constraints of equation solution as can be known, want just known at least f of solving equation (1) ₁And f ₂In three couples of corresponding points p ₁, p ₂, p ₃And q ₁, q ₂, q ₃Below just introduce how to find out this three pairs of corresponding points:

(1) extraction of characteristic point

At first, at image f ₁Middle extract minutiae can utilize SUSAN Corner Detection operator to extract, can be with reference to ' " English-Chinese Image Engineering dictionary " by name, publishing house: publishing house of Tsing-Hua University, the books of author: Zhang Yujin ' for SUSAN Corner Detection operator.Concentrate on a certain regional area for fear of the characteristic point of extracting, earlier image is divided into the zonule of m * n five equilibrium, in each zone, only extract a characteristic point, so just obtain image f ₁Feature point set p _i, i=1,2 ..., m * n.

(2) Feature Points Matching

Feature Points Matching is exactly that the characteristic point that will be formed by the Same Scene spot projection on the different images is to finding out.Specifically, be exactly at image f ₂On find and image f ₁In characteristic point p _iCorresponding point.Concrete matching process is to utilize the relevant matches method, tentatively finds out p _iMatch point q _i, i=1,2 ..., m * n.The relevant matches method can be with reference to ' " Digital Image Processing " by name, publishing house: publishing house of Xian Electronics Science and Technology University, chief editor: the books of He Dongjian '.

(3) the rejecting erroneous point is right

In the matching process of characteristic point, can mate error unavoidably, it is right to produce wrong match point.As will accurately calculating six affine parameters, just necessary these wrong match points of rejecting are right, can not bring them into CALCULATION OF PARAMETERS.So it is right that the present invention utilizes the constant criterion of distance proportion to reject erroneous point, this criterion is as follows:

The feature point set P of given total N characteristic point and its matching characteristic point set Q.Arbitrfary point q among the Q _i(x ', y ') be corresponding points p among the P _i(x, the necessary and sufficient condition of correct match point y) is: p _i(x, y) other characteristic points p in the P _j(x, distance y) and q _i(x ', y) other match points q in the Q _jThe distance of (x ', y ') becomes equal proportion, promptly satisfies

$\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2}=k\sqrt{{(x_i^{\′}-x_j^{\′})}^2+{(y_i^{\′}-y_j^{\′})}^2}---\left(2\right)$

I in the formula, j=1,2 ..., N verifies coupling by distance criterion, and is right with deletion mistake coupling and local motor point.Usually, because correct match point logarithm is much larger than the some logarithm of mistake, thereby just can think correct coupling as long as satisfy condition to the distance of point over half.In actual applications, as long as the difference of distance, just can think that distance equates less than preset threshold.

Utilize this criterion, just can obtain image f ₁And f ₂In accurately the point of coupling to p _iAnd q _i,, can try to achieve a as long as therefrom choosing three pairs brings in the equation (1) ₁₁, a ₁₂, a ₂₁, a ₂₂, b ₁, b ₂These six parameters, and then just can determine image f ₁And f ₂Between spatial alternation relation, again according to this spatial alternation relation to image f ₂Carry out inverse transformation, can reach the purpose of the steady picture of video.

In sum, digital video image stabilization method of the present invention can be eliminated the shake of digital video image automatically, in real time, realizes the purpose of the steady picture of video.

Claims (1)

1. electronic image stabilizing method for digital videos, it is characterized in that this method comprises the steps: at first, utilizes the characteristic point in adjacent two width of cloth images in the SUSAN Corner Detection operator extraction video sequence, utilize the relevant matches method to mate these characteristic points, and the match point of rejecting mistake is right; Then, choose three pairs, utilize match point, try to achieve the affine transformation parameter between two width of cloth images the relation of the respective coordinates on two dimensional image from correct match point centering; At last, the back piece image is carried out affine inverse transformation, obtain and the preceding piece image image of registration spatially, thereby finish the method for digital video electronic steady image according to affine formula.