CN106910253A - Stereo-picture cloning process based on different cameral spacing - Google Patents

Stereo-picture cloning process based on different cameral spacing Download PDF

Info

Publication number
CN106910253A
CN106910253A CN201710097015.8A CN201710097015A CN106910253A CN 106910253 A CN106910253 A CN 106910253A CN 201710097015 A CN201710097015 A CN 201710097015A CN 106910253 A CN106910253 A CN 106910253A
Authority
CN
China
Prior art keywords
image
left view
source images
grid
represent
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.)
Granted
Application number
CN201710097015.8A
Other languages
Chinese (zh)
Other versions
CN106910253B (en
Inventor
侯春萍
刘琦
阎维青
陈华
王致芃
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201710097015.8A priority Critical patent/CN106910253B/en
Publication of CN106910253A publication Critical patent/CN106910253A/en
Application granted granted Critical
Publication of CN106910253B publication Critical patent/CN106910253B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating 3D models or images for computer graphics
    • G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformation in the plane of the image
    • G06T3/40—Scaling the whole image or part thereof
    • G06T3/4007—Interpolation-based scaling, e.g. bilinear interpolation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformation in the plane of the image
    • G06T3/40—Scaling the whole image or part thereof
    • G06T3/4038—Scaling the whole image or part thereof for image mosaicing, i.e. plane images composed of plane sub-images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/10—Processing, recording or transmission of stereoscopic or multi-view image signals
    • H04N13/106—Processing image signals
    • H04N13/128—Adjusting depth or disparity
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/20—Image signal generators
    • H04N13/271—Image signal generators wherein the generated image signals comprise depth maps or disparity maps

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Architecture (AREA)
  • Computer Graphics (AREA)
  • Computer Hardware Design (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)

Abstract

The invention belongs to technical field of image processing, to propose the stereo-picture clone algorithm for camera spacing when inconsistent.The technical solution adopted by the present invention is, stereo-picture cloning process based on different cameral spacing, image preprocessing is carried out first, post processing user draw profile to carry out follow-up distortion of the mesh, the shapes and sizes of stereo-picture cloning region are adjusted by adjusting parallax afterwards, finally the stereo-picture cloning region after adjustment is fused in target image.Present invention is mainly applied to image procossing occasion.

Description

Stereo-picture cloning process based on different cameral spacing
Technical field
The invention belongs to technical field of image processing, specifically, it is related to a kind of stereo-picture based on different cameral spacing Clone algorithm.
Background technology
With the increase of the 3D media such as 3D films and TV, 3D contents receive the extensive concern of user.Therefore, once with Family can easily touch 3D cameras and 3D display devices, just may desire to edit 3D figures as editor's 2D images Picture, however, the method for 2D picture edittings is applied directly into 3D rendering, tends not to produce the vision body of comfortable pleasure Test.
At present, had the algorithms of many maturations on 2D images clone, however 3D rendering clone be faced with it is many new Challenge：1) 3D rendering increased depth constraints compared to 2D images, therefore, in order to ensure the depth one in stereo-picture cloning procedure Cause property to the size and shape of cloning region, it is necessary to be adjusted.2) for comfortable visual experience, it is necessary to keep left view point diagram Corresponding relation between picture and right visual point image.
At present, existing some investigators are studied stereo-picture clone.Lo et al. proposes a kind of 3D and replicates With the method pasted, it is necessary to treat cloned object in Accurate Segmentation source images, and paste it into target image.Luo et al. Propose a kind of seamless three-dimensional image clone algorithm that deformation is perceived based on perspective, the algorithm need not to source images in treat gram Grand region carries out Accurate Segmentation, can process with smeared out boundary, be difficult the cloned object of segmentation.However, in stereo-picture gram In grand process, the studies above assumes that shooting source images are consistent with the camera spacing of the stereoscopic camera of photographic subjects image 's.However, in practice, the camera spacing for shooting the stereoscopic camera of source images and photographic subjects image is often inconsistent, therefore, phase Stereo-picture clone's problems demand when machine spacing is inconsistent is solved.
The content of the invention
To overcome the deficiencies in the prior art, the present invention is directed to propose the stereo-picture clone for camera spacing when inconsistent Algorithm.The technical solution adopted by the present invention is that the stereo-picture cloning process based on different cameral spacing carries out image pre- first Treatment, post processing user drafting profile to carry out follow-up distortion of the mesh, adjust solid by adjusting parallax afterwards , finally be fused to the stereo-picture cloning region after adjustment in target image by the shapes and sizes of image cloning region.
Image preprocessing comprises the concrete steps that, calculates the disparity map of source images and target image, while being source images and target Picture construction quadrilateral mesh, building mode is as follows：First with a uniform grid Ml=(Vl,El,Fl) represent source images a left side Visual point image, wherein VlRepresent the grid vertex of left view dot image, ElRepresent the Grid Edge of left view dot image, FlRepresent left view point The grid quadrangle of image, afterwards disparity map according to source images build the grid M of the right visual point image of source imagesr=(Vr,Er, Fr), wherein, VrRepresent the grid vertex of right visual point image, ErRepresent the Grid Edge of right visual point image, FrRepresent right visual point image Grid quadrangle.
Profile treatment comprises the concrete steps that, user's drafting is appointed with reference to the initial uniform grid of the left view dot image of source images Meaning profile is adjusted, and makes new profile be the minimum rectangle profile that profile is drawn comprising user, wherein the four of rectangular profile angle Point is the grid vertex in the initial uniform grid of the left view dot image of source images, is obtaining the left view dot image of source images After new profile, the new profile of the right visual point image of source images is obtained according to source images disparity map, profile enclosing region is source Image cloning region.
Adjustment parallax comes the i.e. iteration optimization deformation of shapes and sizes of iteration adjustment stereo-picture cloning region, specific steps It is,
1) parallax adjustment
Adjustment mode is as follows：Using the disparity map gradient fields of source images cloning region as the guide field G for solving following formula0, Force the parallax at the source images cloning region edge after adjustment identical with the parallax at target stereogram sticking area edge simultaneously, lead to Cross the disparity map for solving the source images cloning region after following minimum values for weighting Poisson equatioies are adjusted
WhereinIt is the weighting function formulated for source images cloning region, it is more smooth for punishing Region, DTTarget image disparity map is represented,The profile of source images cloning region is represented,Represent the source figure after adjustment As the gradient fields of the disparity map of cloning region；
2) the optimization deformation based on quadrilateral mesh
The relation of 2.1 object scene sizes and parallax
It is assumed that stereoscopic camera camera be arranged in parallel, according to geometrical relationship and similar triangles theorem, scene thing is derived Relation between body depth z and camera spacing b is：
Wherein, b represents stereoscopic camera spacing, and f represents camera focus, and z represents the depth of object scene, d=xL-xRRepresent Image parallactic, xLAnd xRRepresent Same Scene object in left view dot image and the corresponding abscissa of right visual point image；
According to geometrical relationship and similar triangles theorem, size and field of the object scene in stereo-picture plane are derived Relation between scape Object Depth is
Wherein, L represents the size of object scene, and x represents object scene size on the image plane, thus derives and appears on the scene Relation between size x and image parallactic d of the scenery body in stereo-picture plane is：
From above formula, object scene L the size x on stereo-picture and image parallactic d and stereoscopic camera spacing b have Close, and it is unrelated with stereoscopic camera focal length f；
For Same Scene object L1, it is respectively x on source images with the size on target image1And x2, parallax difference It is d1And d2, binocular solid camera spacing is respectively b1And b2, obtain following formula：
Can be obtained by above-mentioned two formula, scaling of the Same Scene object on source images and target image when camera spacing is inconsistent Relation is：
Thus, size of the object scene on stereo-picture is not only relevant with image parallactic, and with the phase of stereoscopic camera Machine spacing is relevant；
2.2 optimization deformations
Combined using the weight of perspective scaling item, line constraint item, the consistent item of parallax, vertical alignment item and position fixterm To instruct quadrilateral mesh to deform, wherein VlAnd VrRepresent the left view point image lattice summit before optimization deformation and right visual point image Grid vertex,WithRepresent the left view point image lattice summit after optimization deformation and right visual point image grid vertex；
(1) perspective scaling item definition perspective zoom factor is
Wherein,WithThe parallax after the adjustment of source images cloning region and the parallax before adjustment, b are represented respectivelySAnd bTPoint The stereoscopic camera spacing of source images and the stereoscopic camera spacing of photographic subjects image Biao Shi not shot；
For the arbitrary mess side of source images cloning region left view dot imageWherein, Vi lWithRepresent respectively excellent Change the ith and jth apex coordinate of the left view point image lattice before deformation, its zoom factor is defined as
Wherein,WithThe perspective on the ith and jth summit of the left view point image lattice before optimization deformation is represented respectively Zoom factor；
Perspective scaling item is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set；Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate；SijRepresent the zoom factor of Grid Edge；
(2) line constraint
To be with the length ratio before deformation after the Grid Edge definition deformation of source images left view dot image
Wherein, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation is represented respectively, WithThe ith and jth apex coordinate of the left view point image lattice after optimization deformation is represented respectively；
The length ratio of the right visual point image of source images is defined with identical method
Line constraint is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate；
(3) the consistent item of parallax
The horizontal coordinate component to be calculated for every groupIts parallax should keep one with the parallax after adjustment Cause, orderRepresent the parallax after adjustment.The consistent item of parallax is defined as
Wherein,Left view dot image horizontal coordinate to be calculated is represented,Represent that right visual point image level to be calculated is sat Mark；
(4) vertical alignment
The vertical coordinate component to be calculated for every groupIt should be made to level off to 0 after adjusting, so as to eliminate vertical Look at the influence of difference straight, vertical alignment is defined as
Wherein,Left view dot image vertical coordinate to be calculated is represented,Represent right visual point image vertical seat to be calculated Mark；
(5) position fixterm
For the center for preventing source images cloning region changes, introduce position fixterm and fix source images clone area The center of the left grid in domain, orderThe center of initial left grid is represented, wherein, VlRepresent an initial left side Visual point image mesh vertex coordinates, Vi lRepresent i-th apex coordinate of initial left view point image lattice.Position fixterm definition For
Wherein,The left view point image lattice apex coordinate after optimization deformation is represented,Represent the left view after optimization deformation I-th apex coordinate of dot image grid；
Optimize deformation gross energy be defined as above-mentioned five energy terms weight add and：
Φ=ωsΦs+ωlΦl+ωdΦd+ωvΦv+ωdΦd
Wherein, Φs、Φl、Φd、ΦvAnd ΦdPerspective scaling energy, line constraint energy, the enable of parallax one are represented respectively Amount, vertical alignment energy and position fixed energies.ωs、ωl、ωd、ωvAnd ωdIt is the corresponding weight coefficient of above-mentioned energy；
Optimization warp mesh is obtained by the minimum value for solving the energy function.
Image co-registration is comprised the concrete steps that, after the optimization warp mesh for obtaining source images cloning region, is inserted by bilinearity It is worth the source images cloning region after being deformed, the source images after deformation is cloned by area using the method for plane graph cut afterwards The left view dot image and right visual point image in domain are fused to the left view dot image and right visual point image of target image to obtain most respectively Whole stereo-picture fusion results.
The features of the present invention and beneficial effect are：
Stereo-picture clone algorithm based on different cameral spacing proposed by the present invention, realizes the solid for shooting source images Stereo-picture clone in the case that camera spacing is different with the stereoscopic camera spacing of photographic subjects image, proposes compared to forefathers Algorithm realizes more accurate stereo-picture clone.
Brief description of the drawings：
Fig. 1 gives profile modification schematic diagram.
A source images left view point image outline in figure, the right visual point image profile of b source images.
Fig. 2 gives the relation of object scene size and parallax.
A top views in figure, b right views
Fig. 3 present invention and the contrast for not considering camera spacer conditions.
A source images left view dot images；
B target image left view dot images；
C does not consider the left view dot image of the stereo clone image obtained during camera spacing；
D considers the left view dot image of the stereo clone image obtained during camera spacing；
The right visual point image of e source images；
The right visual point image of f target images；
G does not consider the right visual point image of the stereo clone image obtained during camera spacing；
H considers the right visual point image of the stereo clone image obtained during camera spacing.
Specific embodiment
At present, existing many stereo-picture clone algorithms, existing algorithm assumes that shooting source images and photographic subjects image Stereoscopic camera spacing it is consistent, in fact, camera spacing has very big influence to stereo-picture clone.
The present invention proposes a kind of stereo-picture clone algorithm based on different cameral spacing.Image is carried out first to locate in advance Reason, post processing user drafting profile to carry out follow-up distortion of the mesh, adjust stereogram by adjusting parallax afterwards As the shapes and sizes of cloning region, finally the stereo-picture cloning region after adjustment is fused in target image.The present invention Concrete technical scheme be divided into the following steps：
1. image preprocessing
In image pre-processing phase, it is necessary to calculate the disparity map of source images and target image, at the same need for source images and Target image builds quadrilateral mesh, and building mode is as follows：First with a uniform grid Ml=(Vl,El,Fl) represent source images Left view point diagram, wherein VlRepresent the grid vertex of left view dot image, ElRepresent the Grid Edge of left view dot image, FlRepresent left view The grid quadrangle of dot image, afterwards disparity map according to source images build the grid M of the right viewpoint figure of source imagesr=(Vr,Er, Fr), wherein VrRepresent the grid vertex of right visual point image, ErRepresent the Grid Edge of right visual point image, FrRepresent right visual point image Grid quadrangle.
2. profile treatment
Operated for the ease of user, it is allowed to which user is any profile of left view point Image Rendering of source images.Afterwards, it is Iteration adjustment is optimized, the initial uniform grid of the left view point diagram that can combine source images enter to any profile that user draws Row adjustment.New profile is set to be the minimum rectangle profile that profile is drawn comprising user, wherein the four of rectangular profile angle point is source Grid vertex in the initial uniform grid of the left view dot image of image.Obtain source images left view dot image new profile it Afterwards, the new profile of the right visual point image of source images is obtained according to source images disparity map.Profile modification schematic diagram is as shown in Figure 1.Wheel Wide enclosing region is source images cloning region.
3. iteration optimization deformation
3.1 parallaxes are adjusted
When stereo-picture is cloned, source images and target image are often shot by different stereoscopic cameras, therefore, source images Often there is certain difference in cloning region parallax and target image parallax, in order to avoid parallax occurs in the stereo-picture after clone Jumping phenomenon.Need to be adjusted the parallax of source images cloning region, adjustment mode is as follows：Using source images cloning region Disparity map gradient fields are used as the guide field G for solving following formula0, while force adjustment after source images cloning region edge parallax and The parallax at target stereogram sticking area edge is identical, by solve it is following weighting Poisson equatioies minimum values adjusted after The disparity map of source images cloning region
WhereinIt is the weighting function formulated for source images cloning region, it is more smooth for punishing Region, DTTarget image disparity map is represented,The profile of source images cloning region is represented,Represent the source figure after adjustment As the gradient fields of the disparity map of cloning region.
3.2 optimizations based on quadrilateral mesh deform
3.2.1 the relation of object scene size and parallax
In general, with object to beholder distance increase, object can taper into.Stereo-picture has depth Information, the size of object is influenceed by perceived depth in stereo-picture, and the perceived depth of stereo-picture is subject to image parallactic Influence.Therefore, the size of object depends on the parallax of object in stereo-picture.Fig. 2 gives and closed between article size and parallax The schematic diagram of system, it is herein assumed that stereoscopic camera camera be arranged in parallel, this set is the generally setting of stereoscopic camera.
Geometrical relationship and similar triangles theorem in Fig. 1 (a), can derive object scene depth z and camera Relation between spacing b is：
Wherein, b represents stereoscopic camera spacing, and f represents camera focus, and z represents the depth of object scene, d=xL-xRRepresent Image parallactic, xLAnd xRRepresent Same Scene object in left view dot image and the corresponding abscissa of right visual point image.
Geometrical relationship and similar triangles theorem in Fig. 1 (b), can derive that object scene is flat in stereo-picture The relation between size and object scene depth on face is
Wherein, L represents the size of object scene, and x represents object scene size on the image plane.
It is possible thereby to derive that the relation between size x and image parallactic d of the object scene in stereo-picture plane is
From above formula, object scene L the size x on stereo-picture and image parallactic d and stereoscopic camera spacing b have Close, and it is unrelated with stereoscopic camera focal length f.
For Same Scene object L1, it is respectively x on source images with the size on target image1And x2, parallax difference It is d1And d2, binocular solid camera spacing is respectively b1And b2, following formula can be obtained：
Can be obtained by above-mentioned two formula, scaling of the Same Scene object on source images and target image when camera spacing is inconsistent Relation is：
It follows that size of the object scene on stereo-picture is not only relevant with image parallactic, and and stereoscopic camera Camera spacing it is relevant.
3.2.2 optimization deformation
In order to realize that the stereo-picture under different stereoscopic camera spacing is cloned, the present invention scales item, straight line about using perspective The weight of the consistent item of Shu Xiang, parallax, vertical alignment item and position fixterm combines to instruct quadrilateral mesh to deform, wherein VlWith VrThe left view point image lattice summit before optimization deformation and right visual point image grid vertex are represented,WithAfter representing optimization deformation Left view point image lattice summit and right visual point image grid vertex.
(1) perspective scaling item
Introduced according to 3.1.1, size of the object scene on stereo-picture is relevant with image parallactic and stereoscopic camera spacing. Zoom factor is had an X-rayed in present invention definition
Wherein,WithThe parallax after the adjustment of source images cloning region and the parallax before adjustment, b are represented respectivelySAnd bTPoint The stereoscopic camera spacing of source images and the stereoscopic camera spacing of photographic subjects image, V Biao Shi not shoti lBefore representing optimization deformation I-th apex coordinate of left view point image lattice.
For the arbitrary mess side of source images cloning region left view dot imageWherein, Vi lWithRepresent respectively excellent Change the ith and jth apex coordinate of the left view point image lattice before deformation, its zoom factor is defined as
Wherein,WithThe perspective on the ith and jth summit of the left view point image lattice before optimization deformation is represented respectively Zoom factor.
Perspective scaling item is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set；Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate；SijRepresent the zoom factor of Grid Edge.
(2) straight line scaling item
During deformation is optimized, to avoid source images cloned object recurring structure distortion, Grid Edge is forced along original The direction come zooms in and out, and is with the length ratio before deformation after the Grid Edge definition deformation for being source images left view dot image
Wherein, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation is represented respectively, WithThe ith and jth apex coordinate of the left view point image lattice after optimization deformation is represented respectively.
The length ratio of the right visual point image of source images is defined with identical method
Line constraint is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate.
(3) the consistent item of parallax
The horizontal coordinate component to be calculated for every groupIts parallax should keep one with the parallax after adjustment Cause, orderRepresent the parallax after adjustment.The consistent item of parallax is defined as
Wherein,Left view dot image horizontal coordinate to be calculated is represented,Represent that right visual point image level to be calculated is sat Mark.
(4) vertical alignment
The vertical coordinate component to be calculated for every groupIt should be made to level off to 0 after adjusting, so as to eliminate vertical Look at the influence of difference straight, vertical alignment is defined as
Wherein,Left view dot image vertical coordinate to be calculated is represented,Represent right visual point image vertical seat to be calculated Mark.
(5) position fixterm
For the center for preventing source images cloning region changes, introduce position fixterm and fix source images clone area The center of the left grid in domain, orderRepresent the center of initial left grid, wherein VlRepresent an initial left side Visual point image mesh vertex coordinates, Vi lRepresent i-th apex coordinate of initial left view point image lattice.Position fixterm definition For
Wherein,The left view point image lattice apex coordinate after optimization deformation is represented,Represent the left view after optimization deformation I-th apex coordinate of dot image grid.
Optimize deformation gross energy be defined as above-mentioned five energy terms weight add and：
Φ=ωsΦs+ωlΦl+ωdΦd+ωvΦv+ωdΦd
Wherein, Φs、Φl、Φd、ΦvAnd ΦdPerspective scaling energy, line constraint energy, the enable of parallax one are represented respectively Amount, vertical alignment energy and position fixed energies.ωs、ωl、ωd、ωvAnd ωdIt is the corresponding weight coefficient of above-mentioned energy.
Optimization warp mesh is obtained by the minimum value for solving the energy function.
4. image co-registration
After the optimization warp mesh for obtaining source images cloning region, the source images after bilinear interpolation is deformed Cloning region, is afterwards regarded the left view dot image of the source images cloning region after deformation and the right side using the method for plane graph cut Dot image is fused to the left view dot image and right visual point image of target image and merges knot obtaining final stereo-picture respectively Really.
To make the object, technical solutions and advantages of the present invention clearer, of the invention specific implementation is illustrated below Journey.
The present invention is emulated based on MATLAB platforms, and concrete technical scheme is divided into following steps：
1. image preprocessing
In image pre-processing phase, it is necessary to calculate the disparity map of source images and target image, at the same need for source images and Target image builds quadrilateral mesh, and building mode is as follows：First with a uniform grid Ml=(Vl,El,Fl) represent source images Left view dot image, wherein VlRepresent the grid vertex of left view dot image, ElRepresent the Grid Edge of left view dot image, FlRepresent left The grid quadrangle of visual point image, afterwards disparity map according to source images build the grid M of the right visual point image of source imagesr=(Vr, Er,Fr), wherein, VrRepresent the grid vertex of right visual point image, ErRepresent the Grid Edge of right visual point image, FrRepresent right viewpoint figure The grid quadrangle of picture.
2. profile treatment
Operated for the ease of user, it is allowed to which user draws any profile for the left view point diagram of source images.Afterwards, in order to Iteration adjustment is optimized, the initial uniform grid of the left view point diagram that can combine source images is carried out to any profile that user draws Adjustment.New profile is set to be the minimum rectangle profile that any profile is drawn comprising user, wherein the four of rectangular profile angle point is Grid vertex in the initial uniform grid of the left view point diagram of source images.Obtain source images left view point diagram new profile it Afterwards, the new profile of the right viewpoint figure of source images is obtained according to source images disparity map.Profile modification schematic diagram is as shown in Figure 1.Profile Enclosing region is source images cloning region.
3. iteration optimization deformation
3.1 parallaxes are adjusted
When stereo-picture is cloned, source images and target image are often shot by different stereoscopic cameras, therefore, source images Often there is certain difference in cloning region parallax and target image parallax, in order to avoid parallax occurs in the stereo-picture after clone Jumping phenomenon.Need to be adjusted the parallax of source images cloning region, adjustment mode is as follows：Using source images cloning region Disparity map gradient fields are used as the guide field G for solving following formula0, while force adjustment after source images cloning region edge parallax and The parallax at target stereogram sticking area edge is identical, by solve it is following weighting Poisson equatioies minimum values adjusted after The disparity map of source images cloning region
WhereinIt is the weighting function formulated for source images cloning region, it is more smooth for punishing Region, DTTarget image disparity map is represented,The profile of source images cloning region is represented,Represent the source figure after adjustment As the gradient fields of the disparity map of cloning region.
3.2 optimizations based on quadrilateral mesh deform
3.2.1 the relation of object scene size and parallax
In general, with object to beholder distance increase, object can taper into.Stereo-picture has depth Information, the size of object is influenceed by perceived depth in stereo-picture, and the perceived depth of stereo-picture is subject to image parallactic Influence.Therefore, the size of object depends on the parallax of object in stereo-picture.Fig. 2 gives and closed between article size and parallax The schematic diagram of system, it is herein assumed that stereoscopic camera camera be arranged in parallel, this set is the generally setting of stereoscopic camera.
Geometrical relationship and similar triangles theorem in Fig. 1 (a), can derive object scene depth z and camera Relation between spacing b is：
Wherein, b represents stereoscopic camera spacing, and f represents camera focus, and z represents the depth of object scene, d=xL-xRRepresent Image parallactic, xLAnd xRRepresent Same Scene object in left view dot image and the corresponding abscissa of right visual point image.
Geometrical relationship and similar triangles theorem in Fig. 1 (b), can derive that object scene is flat in stereo-picture The relation between size and object scene depth on face is
Wherein, L represents the size of object scene, and x represents object scene size on the image plane.
It is possible thereby to derive that the relation between size x and image parallactic d of the object scene in stereo-picture plane is
From above formula, object scene L the size x on stereo-picture and image parallactic d and stereoscopic camera spacing b have Close, and it is unrelated with stereoscopic camera focal length f.
For Same Scene object L1, it is respectively x on source images with the size on target image1And x2, parallax difference It is d1And d2, binocular solid camera spacing is respectively b1And b2, following formula can be obtained：
Can be obtained by above-mentioned two formula, scaling of the Same Scene object on source images and target image when camera spacing is inconsistent Relation is：
It follows that size of the object scene on stereo-picture is not only relevant with image parallactic, and and stereoscopic camera Camera spacing it is relevant.
3.2.2 optimization deformation
In order to realize that the stereo-picture under different stereoscopic camera spacing is cloned, the present invention scales item, straight line about using perspective The weight of the consistent item of Shu Xiang, parallax, vertical alignment item and position fixterm combines to instruct quadrilateral mesh to deform, wherein VlWith VrThe left view point image lattice summit before optimization deformation and right visual point image grid vertex are represented,WithAfter representing optimization deformation Left view point image lattice summit and right visual point image grid vertex.
(1) perspective scaling item
Introduced according to 3.1.1, between the camera of size and image parallactic and stereoscopic camera of the object scene on stereo-picture Away from relevant.Zoom factor is had an X-rayed in present invention definition
Wherein,WithThe parallax after the adjustment of source images cloning region and the parallax before adjustment, b are represented respectivelySAnd bTPoint The stereoscopic camera spacing of source images and the stereoscopic camera spacing of photographic subjects image, V Biao Shi not shoti lBefore representing optimization deformation I-th apex coordinate of left view point image lattice.
For the arbitrary mess side of source images cloning region left view dot imageWherein, Vi lWithRepresent respectively excellent Change the ith and jth apex coordinate of the left view point image lattice before deformation, its zoom factor is defined as
Wherein,WithThe perspective on the ith and jth summit of the left view point image lattice before optimization deformation is represented respectively Zoom factor.
Perspective scaling item is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set；Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate；SijRepresent the zoom factor of Grid Edge.
(2) line constraint
During deformation is optimized, to avoid source images cloned object recurring structure distortion, Grid Edge is forced along original The direction come zooms in and out, and is with the length ratio before deformation after the Grid Edge definition deformation for being source images left view dot image
Wherein, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation is represented respectively, WithThe ith and jth apex coordinate of the left view point image lattice after optimization deformation is represented respectively.
The length ratio of the right visual point image of source images is defined with identical method
Line constraint is defined as
Wherein, ElAnd ErSource images cloning region left view point image lattice side and right visual point image Grid Edge are represented respectively Set, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRespectively Represent the ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithAfter representing optimization deformation respectively The ith and jth apex coordinate of left view point image lattice,WithThe right visual point image grid after optimization deformation is represented respectively Ith and jth apex coordinate.
(3) the consistent item of parallax
The horizontal coordinate component to be calculated for every groupIts parallax should be consistent with the parallax after adjustment, orderRepresent the parallax after adjustment.The consistent item of parallax is defined as
Wherein,Left view dot image horizontal coordinate to be calculated is represented,Represent that right visual point image level to be calculated is sat Mark.
(4) vertical alignment
The vertical coordinate component to be calculated for every groupIt should be made to level off to 0 after adjusting, so as to eliminate vertical Look at the influence of difference straight, vertical alignment is defined as
Wherein,Left view dot image vertical coordinate to be calculated is represented,Represent right visual point image vertical seat to be calculated Mark.
(5) position fixterm
For the center for preventing source images cloning region changes, introduce position fixterm and fix source images clone area The center of the left grid in domain, orderThe center of initial left grid is represented, wherein, VlRepresent an initial left side Visual point image mesh vertex coordinates, Vi lRepresent i-th apex coordinate of initial left view point image lattice.Position fixterm definition For
Wherein,The left view point image lattice apex coordinate after optimization deformation is represented,Represent the left view after optimization deformation I-th apex coordinate of dot image grid.
Optimize deformation gross energy be defined as above-mentioned five energy terms weight add and：
Φ=ωsΦs+ωlΦl+ωdΦd+ωvΦv+ωdΦd
Wherein, Φs、Φl、Φd、ΦvAnd ΦdPerspective scaling energy, line constraint energy, the enable of parallax one are represented respectively Amount, vertical alignment energy and position fixed energies.ωs、ωl、ωd、ωvAnd ωdIt is the corresponding weight coefficient of above-mentioned energy.
Optimization warp mesh is obtained by the minimum value for solving the energy function.The weight coefficient for using herein is respectively ωs=50, ωl=50, ωd=50, ωv=100 and ωp=1.Gross energy Φ is a double optimization function, and is based on Matlab R2013b programmings obtain last solution.
4. image co-registration
After the optimization warp mesh for obtaining source images cloning region, the source images after bilinear interpolation is deformed Cloning region, is afterwards regarded the left view dot image of the source images cloning region after deformation and the right side using the method for plane graph cut Dot image is fused to the left view dot image and right visual point image of target image and merges knot obtaining final stereo-picture respectively Really.
5. test and its effect assessment
In order to check the performance of stereo-picture clone algorithm proposed by the present invention, result of the present invention to stereo-picture clone Shown.The source stereo-picture of input is shot by JVC Everio 3D high-definition cameras, and camera spacing is 35mm, target Stereo-picture is shot by FUJIFILM REAL 3D W1 cameras, and camera spacing is 77mm.
Stereo-picture cloning process Comparative result when Fig. 3 illustrates consideration camera spacing and do not consider camera spacing. In Fig. 3, (a) and (e) is respectively the left view dot image and right visual point image of source images；B () and (f) represents the left view of target image Dot image and right visual point image；C () and (g) represented when not considering camera spacing, that is, assume to shoot source images and target image The left view dot image and right visual point image of the stereo clone image that camera spacing is obtained when consistent；D () and (g) represents and considers camera The left view dot image and right visual point image of the stereo clone image obtained during spacing.Existing algorithm assumes that shooting source images and bat The stereoscopic camera spacing for taking the photograph target image is consistent, however, in this case, stereo-picture clone's result of acquisition is often not Accurately.When camera spacing is not considered, i.e., only consider influence of the parallax to source images cloning region shapes and sizes, now, Source images cloning region is exaggerated, such as shown in (c) and (g).However, it is contemplated that during camera spacing, source images cloning region reduces, such as Figure (d) and (g).Therefore, algorithm proposed by the present invention can realize more accurate scaling.

Claims (5)

1. a kind of stereo-picture cloning process based on different cameral spacing, it is characterized in that, image preprocessing is carried out first, afterwards The profile that treatment user draws adjusts stereo-picture clone area to carry out follow-up distortion of the mesh, afterwards by adjusting parallax , finally be fused to the stereo-picture cloning region after adjustment in target image by the shapes and sizes in domain.
2. the stereo-picture cloning process of different cameral spacing is based on as claimed in claim 1, it is characterized in that, image preprocessing Comprise the concrete steps that, calculate the disparity map of source images and target image, while be that source images and target image build quadrilateral mesh, Building mode is as follows：First with a uniform grid Ml=(Vl,El,Fl) represent source images left view dot image, wherein VlRepresent The grid vertex of left view dot image, ElRepresent the Grid Edge of left view dot image, FlThe grid quadrangle of left view dot image is represented, it The disparity map according to source images builds the grid M of the right visual point image of source images afterwardsr=(Vr,Er,Fr), wherein, VrRepresent right viewpoint The grid vertex of image, ErRepresent the Grid Edge of right visual point image, FrRepresent the grid quadrangle of right visual point image.
3. the stereo-picture cloning process of different cameral spacing is based on as claimed in claim 1, it is characterized in that, profile treatment tool Body step is that any profile that user draws is adjusted with reference to the initial uniform grid of the left view dot image of source images, is made New profile is the minimum rectangle profile that profile is drawn comprising user, and wherein the four of rectangular profile angle point is the left view of source images Grid vertex in the initial uniform grid of dot image, after the new profile of left view dot image of source images is obtained, according to source Image parallactic figure obtains the new profile of the right visual point image of source images, and profile enclosing region is source images cloning region.
4. the stereo-picture cloning process of different cameral spacing is based on as claimed in claim 1, it is characterized in that, adjust parallax The shapes and sizes of iteration adjustment stereo-picture cloning region are iteration optimization deformation, are comprised the concrete steps that,
1) parallax adjustment
Adjustment mode is as follows：Using the disparity map gradient fields of source images cloning region as the guide field G for solving following formula0, while compeling Make the parallax at the source images cloning region edge after adjustment identical with the parallax at target stereogram sticking area edge, by solving It is following weighting Poisson equatioies minimum values adjusted after source images cloning region disparity map
s u b j e c t t o D ^ S c | ∂ Ω 0 L = D T | ∂ Ω 0 L
WhereinIt is the weighting function formulated for source images cloning region, for punishing more smooth area Domain, DTTarget image disparity map is represented,The profile of source images cloning region is represented,Represent the source images gram after adjustment The gradient fields of the disparity map in grand region；
2) the optimization deformation based on quadrilateral mesh
The relation of 2.1 object scene sizes and parallax
It is assumed that stereoscopic camera camera be arranged in parallel, according to geometrical relationship and similar triangles theorem, object scene depth is derived Degree z and camera spacing b between relation be：
z = ( b - d ) f d
Wherein, b represents stereoscopic camera spacing, and f represents camera focus, and z represents the depth of object scene, d=xL-xRRepresent image Parallax, xLAnd xRRepresent Same Scene object in left view dot image and the corresponding abscissa of right visual point image；
According to geometrical relationship and similar triangles theorem, size of the object scene in stereo-picture plane and scene thing are derived Relation between body depth is
x = f L f + z
Wherein, L represents the size of object scene, and x represents object scene size on the image plane, thus derives scene thing Relation between size x and image parallactic d of the body in stereo-picture plane is：
x = L d b
From above formula, size xs of the object scene L on stereo-picture is relevant with image parallactic d and stereoscopic camera spacing b, and It is unrelated with stereoscopic camera focal length f；
For Same Scene object L1, it is respectively x on source images with the size on target image1And x2, parallax is respectively d1 And d2, binocular solid camera spacing is respectively b1And b2, obtain following formula：
x 1 = L 1 d 1 b 1
x 2 = L 1 d 2 b 2
Can be obtained by above-mentioned two formula, scaling relation of the Same Scene object on source images and target image when camera spacing is inconsistent For：
x 2 x 1 = d 2 b 1 d 1 b 2
Thus, size of the object scene on stereo-picture is not only relevant with image parallactic, and between the camera of stereoscopic camera Away from relevant；
2.2 optimization deformations
Referred to using the weight combination of perspective scaling item, line constraint item, the consistent item of parallax, vertical alignment item and position fixterm Lead quadrilateral mesh deformation, wherein VlAnd VrRepresent the left view point image lattice summit before optimization deformation and right visual point image grid Summit,WithRepresent the left view point image lattice summit after optimization deformation and right visual point image grid vertex；
(1) perspective scaling item definition perspective zoom factor is
S V i = D ^ S c ( V i l ) b S D S c ( V i l ) b T
Wherein,WithThe parallax after the adjustment of source images cloning region and the parallax before adjustment, b are represented respectivelySAnd bTDifference table Show the stereoscopic camera spacing of the stereoscopic camera spacing and photographic subjects image that shoot source images, Vi lRepresent the left view before optimization deformation I-th apex coordinate of dot image grid；
For the arbitrary mess side of source images cloning region left view dot imageWherein, Vi lWithRepresent that optimization becomes respectively The ith and jth apex coordinate of the left view point image lattice before shape, its zoom factor is defined as
S i j = 1 2 ( S V i l + S V j l )
Wherein,WithRespectively represent optimization deformation before left view point image lattice ith and jth summit perspective scaling because Son；
Perspective scaling item is defined as
Φ s = Σ ( i , j ) ∈ E l ( | | V ^ i l - V ^ j l | | - S i j | | V i l - V j l | | ) 2 + Σ ( i , j ) ∈ E r ( | | V ^ i r - V ^ j r | | - S i j | | V i r - V j r | | ) 2
Wherein, ElAnd ErThe set of source images cloning region left view point image lattice side and right visual point image Grid Edge is represented respectively； Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRepresent respectively The ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithThe left view after optimization deformation is represented respectively The ith and jth apex coordinate of dot image grid,WithI-th of the right visual point image grid after optimization deformation is represented respectively With j-th apex coordinate；SijRepresent the zoom factor of Grid Edge；
(2) line constraint
To be with the length ratio before deformation after the Grid Edge definition deformation of source images left view dot image
L i j l = | | V ^ i l - V ^ j l | | | | V i l - V j l | |
Wherein, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation is represented respectively,With The ith and jth apex coordinate of the left view point image lattice after optimization deformation is represented respectively；
The length ratio of the right visual point image of source images is defined with identical method
Line constraint is defined as
Φ l = Σ ( i , j ) ∈ E l | | ( V ^ i l - V ^ j l ) - L i j l ( V i l - V j l ) | | 2 + Σ ( i , j ) ∈ E r | | ( V ^ i r - V ^ j r ) - L i j r ( V i r - V j r ) | | 2
Wherein, ElAnd ErThe set of source images cloning region left view point image lattice side and right visual point image Grid Edge is represented respectively, Vi lWithThe ith and jth apex coordinate of the left view point image lattice before optimization deformation, V are represented respectivelyi rWithRepresent respectively The ith and jth apex coordinate of the right visual point image grid before optimization deformation；WithThe left view after optimization deformation is represented respectively The ith and jth apex coordinate of dot image grid,WithI-th of the right visual point image grid after optimization deformation is represented respectively With j-th apex coordinate；
(3) the consistent item of parallax
The horizontal coordinate component to be calculated for every groupIts parallax should be consistent with the parallax after adjustment, orderRepresent the parallax after adjustment.The consistent item of parallax is defined as
Φ d = Σ ( V ^ x i l , V ^ x i r ) ∈ V ^ x | | ( V ^ x i l - V ^ x i r ) - d x i | | 2
Wherein,Left view dot image horizontal coordinate to be calculated is represented,Represent right visual point image horizontal coordinate to be calculated；
(4) vertical alignment
The vertical coordinate component to be calculated for every groupIt should be made to level off to 0 after adjusting, so as to eliminate vertically regard Poor influence, vertical alignment is defined as
Φ v = Σ ( V ^ y i l , V ^ y i r ) ∈ V ^ y | | V ^ y i l - V ^ y i r | | 2
Wherein,Left view dot image vertical coordinate to be calculated is represented,Represent right visual point image vertical coordinate to be calculated；
(5) position fixterm
For the center for preventing source images cloning region changes, introduce position fixterm and fix a source images cloning region left side The center of grid, orderThe center of initial left grid is represented, wherein, VlRepresent initial left view point Image lattice apex coordinate, Vi lRepresent i-th apex coordinate of initial left view point image lattice.Position fixterm is defined as
Φ p = | | C l - 1 | V ^ l | Σ V ^ i l ∈ V ^ L V ^ i l | | 2
Wherein,The left view point image lattice apex coordinate after optimization deformation is represented,Represent the left view point diagram after optimization deformation As i-th apex coordinate of grid；
Optimize deformation gross energy be defined as above-mentioned five energy terms weight add and：
Φ=ωsΦs+ωlΦl+ωdΦd+ωvΦv+ωdΦd
Wherein, Φs、Φl、Φd、ΦvAnd ΦdRespectively represent perspective scaling energy, line constraint energy, the consistent energy of parallax, hang down Straight alignment energy and position fixed energies.ωs、ωl、ωd、ωvAnd ωdIt is the corresponding weight coefficient of above-mentioned energy；
Optimization warp mesh is obtained by the minimum value for solving the energy function.
5. the stereo-picture cloning process of different cameral spacing is based on as claimed in claim 1, it is characterized in that, image co-registration tool Body step is, after the optimization warp mesh for obtaining source images cloning region, the source figure after bilinear interpolation is deformed As cloning region, afterwards using the method for plane graph cut by the left view dot image of the source images cloning region after deformation and the right side Visual point image is fused to the left view dot image and right visual point image of target image and merges knot obtaining final stereo-picture respectively Really.
CN201710097015.8A 2017-02-22 2017-02-22 Stereo image cloning method based on different camera distances Active CN106910253B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710097015.8A CN106910253B (en) 2017-02-22 2017-02-22 Stereo image cloning method based on different camera distances

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710097015.8A CN106910253B (en) 2017-02-22 2017-02-22 Stereo image cloning method based on different camera distances

Publications (2)

Publication Number Publication Date
CN106910253A true CN106910253A (en) 2017-06-30
CN106910253B CN106910253B (en) 2020-02-18

Family

ID=59208811

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710097015.8A Active CN106910253B (en) 2017-02-22 2017-02-22 Stereo image cloning method based on different camera distances

Country Status (1)

Country Link
CN (1) CN106910253B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107767339A (en) * 2017-10-12 2018-03-06 深圳市未来媒体技术研究院 A kind of binocular stereo image joining method
CN111292367A (en) * 2020-02-18 2020-06-16 青岛联合创智科技有限公司 Binocular camera depth map generation method with variable baseline

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1129813A (en) * 1995-08-11 1996-08-28 刘江海 Treatment of negative films of stereo photographs and pictures and products thereof
US20070273610A1 (en) * 2006-05-26 2007-11-29 Itt Manufacturing Enterprises, Inc. System and method to display maintenance and operational instructions of an apparatus using augmented reality
CN101375315A (en) * 2006-01-27 2009-02-25 图象公司 Methods and systems for digitally re-mastering of 2D and 3D motion pictures for exhibition with enhanced visual quality
CN101479765A (en) * 2006-06-23 2009-07-08 图象公司 Methods and systems for converting 2D motion pictures for stereoscopic 3D exhibition
CN102074014A (en) * 2011-02-23 2011-05-25 山东大学 Stereo matching method by utilizing graph theory-based image segmentation algorithm
CN104519343A (en) * 2013-09-26 2015-04-15 西克股份公司 3D camera in accordance with stereoscopic principle and method of detecting depth maps
CN105701812A (en) * 2016-01-12 2016-06-22 南京工程学院 Visual identification system suitable for cotton picking robot
CN105959595A (en) * 2016-05-27 2016-09-21 西安宏源视讯设备有限责任公司 Virtuality to reality autonomous response method for virtuality and reality real-time interaction

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1129813A (en) * 1995-08-11 1996-08-28 刘江海 Treatment of negative films of stereo photographs and pictures and products thereof
CN101375315A (en) * 2006-01-27 2009-02-25 图象公司 Methods and systems for digitally re-mastering of 2D and 3D motion pictures for exhibition with enhanced visual quality
US20070273610A1 (en) * 2006-05-26 2007-11-29 Itt Manufacturing Enterprises, Inc. System and method to display maintenance and operational instructions of an apparatus using augmented reality
CN101479765A (en) * 2006-06-23 2009-07-08 图象公司 Methods and systems for converting 2D motion pictures for stereoscopic 3D exhibition
CN102074014A (en) * 2011-02-23 2011-05-25 山东大学 Stereo matching method by utilizing graph theory-based image segmentation algorithm
CN104519343A (en) * 2013-09-26 2015-04-15 西克股份公司 3D camera in accordance with stereoscopic principle and method of detecting depth maps
CN105701812A (en) * 2016-01-12 2016-06-22 南京工程学院 Visual identification system suitable for cotton picking robot
CN105959595A (en) * 2016-05-27 2016-09-21 西安宏源视讯设备有限责任公司 Virtuality to reality autonomous response method for virtuality and reality real-time interaction

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
JIN WOO YOO 等: "Content-Driven Retargeting of Stereoscopic Images", 《IEEE SIGNAL PROCESSING LETTERS》 *
王志远: "立体图像克隆和立体显示串扰的研究", 《中国博士学位论文全文数据库 信息科技辑》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107767339A (en) * 2017-10-12 2018-03-06 深圳市未来媒体技术研究院 A kind of binocular stereo image joining method
CN111292367A (en) * 2020-02-18 2020-06-16 青岛联合创智科技有限公司 Binocular camera depth map generation method with variable baseline
CN111292367B (en) * 2020-02-18 2023-04-07 青岛联合创智科技有限公司 Binocular camera depth map generation method with variable baseline

Also Published As

Publication number Publication date
CN106910253B (en) 2020-02-18

Similar Documents

Publication Publication Date Title
US7983477B2 (en) Method and apparatus for generating a stereoscopic image
CN101902657B (en) Method for generating virtual multi-viewpoint images based on depth image layering
US9443338B2 (en) Techniques for producing baseline stereo parameters for stereoscopic computer animation
CN105678687A (en) Stereo image stitching method based on content of images
CN111047709B (en) Binocular vision naked eye 3D image generation method
TW201243763A (en) Method for 3D video content generation
US20130170737A1 (en) Stereoscopic image converting apparatus and stereoscopic image displaying apparatus
CN106327454A (en) Method and device of synthesizing image
CN103761734A (en) Binocular stereoscopic video scene fusion method for keeping time domain consistency
CN106548494A (en) A kind of video image depth extraction method based on scene Sample Storehouse
CN104794683A (en) Video connecting method based on planar scanning around gradient joint regions
CN102368826A (en) Real time adaptive generation method from double-viewpoint video to multi-viewpoint video
CN101754042A (en) Image reconstruction method and image reconstruction system
CN106447718B (en) A kind of 2D turns 3D depth estimation method
CN104093013A (en) Method for automatically regulating image parallax in stereoscopic vision three-dimensional visualization system
CN106204717B (en) A kind of stereo-picture quick three-dimensional reconstructing method and device
CN102695070B (en) Depth consistency fusion processing method for stereo image
CN104751508B (en) The full-automatic of new view is quickly generated and complementing method in the making of 3D three-dimensional films
CN105979241B (en) A kind of quick inverse transform method of cylinder three-dimensional panoramic video
CN106910253A (en) Stereo-picture cloning process based on different cameral spacing
CN105072433B (en) Depth perception mapping method applied to head track virtual reality system
CN104159099A (en) Method of setting binocular stereoscopic camera in 3D stereoscopic video production
CN104796624A (en) Method for editing and propagating light fields
Knorr et al. An image-based rendering (ibr) approach for realistic stereo view synthesis of tv broadcast based on structure from motion
CN103686118A (en) Image depth adjustment method and device

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
GR01 Patent grant
GR01 Patent grant