CN104301706B - A kind of synthetic method for strengthening bore hole stereoscopic display effect - Google Patents

Abstract

The present invention relates to a kind of synthetic method that can strengthen bore hole stereoscopic display effect, including:A. different Depth Information Acquistions are carried out according to different input datas；B. if binocular data, then by characteristic matching, set up energy function to obtain optimum parallax；C. if image+depth information, then extract parallax information and carry out the optimum parallax of filtering and noise reduction acquisition；D. depth information is split, divides and mark different regions；E. optimum parallax is estimated again to each piece of region；F. adjustment parallactic shift parameter obtains suitable deep space；G. multi-view image is synthesized according to drawing rule, is exported on bore hole TV.The present invention can effectively solve the problem that the not good display problem of bore hole TV stereoeffect.And the depth of field can be effectively synthesized under various resolution ratio and goes out all very strong multi-view image of screen sense.Solve the problems, such as that the multiple viewpoint amounts of calculation of synthesis are huge simultaneously, reduce and calculate the problem that invalid viewpoint is computed repeatedly in multi-view image.

Description

A kind of synthetic method for strengthening bore hole stereoscopic display effect

Technical field

A kind of the present invention relates to disparity computation, image segmentation and multiple views synthesis technical field, more particularly to enhancing bore holeThe synthetic method of stereo display effect.

Background technology

With the continuous development of television industries correlation technique, stereotelevision progresses into the visual field of people, stereotelevisionRealize it will being the great revolution in another field after DTV, it fully achieve will be from imaging to the major progress that images,Stereotelevision is an over all Integration system being fully established in digital technology, computer technology and network technology.

Stereotelevision is a kind of television system of the real space relationship that can simulate actual scenery.Also known as three-dimensional television(Threedimensionaltelevision,3DTV).Three-dimensional television image can not only give a kind of depth preception of spectators, Er QienengSeem sensation within reach outside screen is stretched in a kind of scenery of spectators.The difference of stereotelevision and existing common TVThat existing TV only transmits the two-dimensional image information of a viewpoint, and three-dimensional transmission of television be multiple views image information.PeopleSimultaneously watch the image of an object when object is watched with two eyes, the formation in left and right two of the object of different depthThe parallax of image is different, and brain then melts as going out different depth and distance according to the difference of these parallaxes, so as to produce three-dimensionalThree-dimensional sense organ.

Bore hole 3D TV is both a kind of leading-edge products of stereotelevision, is different from traditional hyperphoria with fixed eyeballs mirror stereotelevision, nakedEye 3D TVs are not needing any auxiliary equipment using two characteristics with parallax of people（Such as 3D eyes, helmet etc.）SituationUnder, you can the lifelike stereoscopic image with space and depth is obtained, is three-dimensional stereo display technique truly.But existingThe universal stereoeffect of some bore hole 3D TVs is not strong, and particularly at high resolutions, spectators' stereoeffect sense organ is not good.

Content of the invention

The purpose of the present invention is to overcome the deficiencies in the prior art, there is provided a kind of enhancing bore hole stereoscopic display effectSynthetic method.The method can significantly lift the stereo display effect of bore hole TV, and can under various resolution ratioMore preferably stereo image effect is brought to spectators.

The technical scheme is that：A kind of synthetic method for strengthening bore hole stereoscopic display effect, obtains including parallaxUnit is taken, deep space synthesis unit, multi-view image synthesis unit, method flow diagram are shown in Fig. 1.Wherein parallax acquiring unit canTo receive two kinds of data forms：Binocular image format and " video+depth "（Video+depth, VD）Form.

The parallax acquiring unit can obtain corresponding parallax information respectively according to different input picture forms.ParallaxAcquiring unit includes binocular parallax acquiring unit and VD parallax acquiring units.

The binocular parallax acquiring unit：View data according to input carries out component, is divided into left view and right view, andLeft and right view is calculated in conjunction with Stereo matching, obtain corresponding parallax information.

The VD parallaxes acquiring unit：View data according to input carries out component, isolates the depth in initial dataInformation, obtains corresponding parallax information in conjunction with pretreatment.

The deep space synthesis unit：The deep space corresponding to view data is reconstructed, is divided according to image segmentationGo out different regions, repartition the depth information of zones of different and be normalized.Spectators can adjust deep space with oneself,Received with reaching the third dimension for more preferably meeting spectators.

The multi-view image synthesis unit：Multi-view image is synthesized according to virtual viewpoint rendering method.

According to the technical scheme that the present invention is provided, the synthetic method of the enhancing bore hole stereoscopic display effect includes following stepSuddenly：1）Parallax acquiring unit obtains initial data image information, selects corresponding unit for different pattern of the inputs.IfInput data is binocular image, then carry out binocular parallax acquiring unit, if video+depth during input data, carries out VD and regardsDifference acquiring unit.

2）Specifically, binocular parallax acquiring unit separates left and right road view ImgL and ImgR, and ImgL and ImgR is being filteredAfter scaling arrange disparity range in [- α, α] calculate Matching power flow respectively obtain profile gradients and color gradient, α is mostBig permission parallax, wherein when pixel parallax angle, energy function is set up according to color gradient,, rootEnergy function is set up according to profile gradients, optimum parallax information is obtained by calculating minimum global energy function.

3）Specifically, VD parallaxes acquiring unit extracts the depth information that initial data includes in itself, needs in the unitFurther filtering and noise reduction is carried out to the depth information for extracting, accurate depth information has been obtained.Close according to virtual view is metInto specification, the parallax information that depth information is converted to original image.

4）Further, the deep space synthesis unit is processed to the parallax information that parallax acquiring unit is obtained.In the unit, need for parallax information to be converted to spatial depth information, map initial data.

5）Specifically, in conversion level spatial depth informationIn, need to carry out edge image to parallax informationSegmentation, determines optimal disparity plane parameter ω of each piece of cut zone.For the deep space for needing to generate N viewpoints,, wherein scale（N）For the relative deviation ratio of n-th parallax, spectators can be withBy adjusting scale（N）To obtain more preferably bore hole 3D impression.

6）Further, multi-view image synthesis unit synthesizes most according to virtual view composite formula according to deep spaceWhole multi-view image is simultaneously exported on bore hole tv display screen.

The technical problem to be solved：A kind of synthetic method for strengthening bore hole stereoscopic display effect is provided.The partyMethod can effectively solve the problem that the not good display problem of bore hole TV stereoeffect.And can be effectively synthesized under various resolution ratioThe depth of field and go out the screen all very strong multi-view image of sense.Solve the problems, such as that the multiple viewpoint amounts of calculation of synthesis are huge simultaneously, reduceCalculate the problem that invalid viewpoint is computed repeatedly in multi-view image.

Description of the drawings

The present invention is described in further detail with specific example below in conjunction with the accompanying drawings.

Flow chart of the accompanying drawing 1 for strengthening the synthetic method of bore hole stereoscopic display effect.

System block diagram of the accompanying drawing 2 for binocular parallax acquiring unit.

System block diagram of the accompanying drawing 3 for VD parallax acquiring units.

System block diagram of the accompanying drawing 4 for deep space synthesis unit.

Specific embodiment

With reference to the accompanying drawings and examples the present invention is described in further detail：

This method includes 101 parallax acquiring units, 102 deep space synthesis units and 103 multi-view images as shown in Figure 1Synthesis unit.

In Fig. 2,201 modules receive binocular image data, and binocular image is divided into left and right two-way image ImgL and ImgR.LeftRight two-way image will keep resolution ratio consistent, need to carry out image scaling to ImgL and ImgR Ru inconsistent, to ensure left and right figureAs resolution sizes.2 modular characters of Figure 20 coupling will carry out characteristic matching detection to left and right two-way image, in permitting for arrangingSIFT algorithms are executed in parallax [- α, α] perhaps（Scale Invariant Feature Transform, scale invariant feature becomeChange）, generate the characteristic vector based on SIFT.Characteristic matching module by set up gaussian pyramid come the yardstick to imageSpatial information,, wherein：Img（x,y）For the point on image ImgL or ImgR,It is that n is Gauss to the information after Img (x, y) gaussian filteringThe pyramidal number of plies, σ be the metric space factor, σ values more big can representative image general picture feature, σ is more little more can representative graphThe minutia of picture.Establish metric spaceAfterwards, the yardstick using two adjacent two-layers on same single order is emptyBetween the difference of function obtain DOG（Difference Of Gaussian）One layer of gaussian pyramid., wherein k be in gaussian pyramid two adjacent layers itBetween dimension scale, preferred this method is set to 0.5.One is obtained by the Local Extremum for detecting DOG spaces and estimates parallax βSet f, if in the module 203, energy function is set up according to color gradient,, according to profileGradient is setting up energy function.Energy letter based on parallax, wherein, pIt is the position of pixel with q, N represents the p in image, q coordinate relations, V is to weighThe correlation function of inconsistent degree, DIt is the function for weighing f and correct parallax error in classification.In 4 color gradient modules of Figure 20, Rs of the V by left and right two-way image, G, B pictureElement difference information is obtained.In 5 profile gradients modules of Figure 20, V is obtained by the shade of gray information gap of left and right two-way image.6 moulds of Figure 20In block, E（f）Minimum global energy distribution is exactly the optimum parallax distribution of original image.

In Fig. 3,301 modules receive the view data of video+depth form, and 302 modules extract the depth letter of VD formatted datasBreath.As various noises such as blocking effect etc. can be produced in video coding process, so 303 modules need to believe original depthBreath is filtered denoising, this method adopts adding window medium filtering, the situation of effective place to go and gaussian noise mixing.304 mouldsIn block by denoising after depth information smoothing processing, retain the parallax information that complete depth profile obtains optimum.

Deep space synthesis unit 102 needs to obtain deep space information aggregate to the process of optimum parallax.FirstFirst optimum parallax is split, method of the segmentation using edge segmentation, by detecting that the edge of parallax information marks off differenceParallax region.The parallax value in each region is sorted, and intermediate value is selected as optimal disparity plane parameter ω, to each segmentationAll parallaxes in cut zone are estimated again with optimal disparity plane parameter ω in region.For the depth for needing to generate N viewpointsSpace,, wherein Zone is each region sequence number,Relative deviation ratios of the scale (N) for n-th parallax.In Fig. 4, a-quadrant and B regions are the depth areas after different two segmentationDomain, the optimal disparity plane of the deep space of a-quadrant and B regions be respectively the space in DV (a) and DV (b), a-quadrant and B regions away fromFrom.By changing scale（N）, the space length in a-quadrant and B regions can be with proportional change.In Fig. 4, the Dis of A and B can be far and near with the final original image Scene of effective influence, and user is by changing scale（N）, can be withSelect suitable parameters to reach more preferable stereoeffect.

Multi-view image synthesis unit 103, willRegard according to the synthesis of multiple views composition rule with original imageDot image, and synthetic multi-view image is exported on bore hole TV, presenting to audience has the video image of stereoeffect.

Claims (1)

1. a kind of synthetic method for strengthening bore hole stereoscopic display effect, its feature specifically includes：

A. parallax acquiring unit obtains initial data image information, selects corresponding unit for different pattern of the inputs；

If b. input data is binocular stereo image, binocular parallax acquiring unit is selected, binocular stereo image is divided into a left sideRight two-way image simultaneously carries out characteristic matching detection allowing to execute in parallax to left and right two-way image, then by color gradient or wheelWide gradient obtains minimum global energy distribution building energy function；If input data is " video+depth " form, selectVD parallax acquiring units are selected, and extract the depth information of VD formatted datas and denoising and smooth place is filtered to original depth informationReason；

C. edge image segmentation is carried out to the parallax information that above-mentioned steps are obtained, by detecting that the edge of parallax information is marked off notSame parallax region, to each cut zone mark, the parallax value in each region is sorted, selects intermediate value as optimalDisparity plane parameter, to each cut zone with optimal disparity plane parameterAgain pre- to all parallaxes in cut zoneEstimate,For the relative offset parameter of parallax, the deep space composite formula of corresponding N number of viewpoint is；

D. can pass through to arrange in above-mentioned stepsTo adjust the space length between each parallax region so as to generating conjunctionSuitable deep space；

E. deep space information and original image are synthesized multi-view image according to multiple views composition rule, and will be synthetic manyVisual point image is exported on bore hole stereoscopic display equipment.