CN105376543B - The parallax picture capturing method and system of a kind of 3D rendering - Google Patents
The parallax picture capturing method and system of a kind of 3D rendering Download PDFInfo
- Publication number
- CN105376543B CN105376543B CN201410383267.3A CN201410383267A CN105376543B CN 105376543 B CN105376543 B CN 105376543B CN 201410383267 A CN201410383267 A CN 201410383267A CN 105376543 B CN105376543 B CN 105376543B
- Authority
- CN
- China
- Prior art keywords
- parallax
- rendering
- disparity map
- cost matrix
- block
- 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.)
- Active
Links
Abstract
The invention discloses the parallax picture capturing method and system of a kind of 3D rendering, gray proces are carried out by left image respectively to 3D rendering and right image, corresponding half-tone information is obtained；Block- matching processing is carried out to the half-tone information of the left image and right image, obtains including the three-dimensional matrice of disparity map information；Cost matrix processing is carried out to each row of the three-dimensional matrice, the corresponding parallax of every a line cost matrix is obtained by viterbi algorithm；Parallax according to every a line forms the disparity map of the 3D rendering；The maximum likelihood disparity map minimum with true disparity map gap is found in all possible disparity maps using viterbi algorithm, so that obtained disparity map is more accurate, the optimum solution of disparity map is sought by viterbi algorithm, greatly reduce amount of calculation, so that obtaining disparity map can be realized by hardware, bring and greatly facilitate.
Description
Technical field
The present invention relates to the parallax picture capturing method and system of 3D display fields, more particularly to a kind of 3D rendering.
Background technology
In general, 3D rendering is generally made up of two-way image, corresponding to a certain specific viewing angle.According to this two-way
The parallax information that video is included, can generate the image corresponding to new viewing angle, and this is the crucial institute of bore hole 3D TVs
.Existing that disparity map is calculated by input picture, the first step is typically to carry out Block- matching, followed by excellent using different methods
Change the result of Block- matching, to obtain accurate result.But the Global Algorithm of existing optimization Block- matching result is excessively complicated,
Realize there is huge difficulty with hardware, rather than Global Algorithm can not then provide gratifying result, therefore existing acquisition
Disparity map accuracy rate is low in the method for disparity map, and it is greatly inconvenient to bring.
Therefore, prior art has yet to be improved and developed.
The content of the invention
The technical problem to be solved in the present invention is parallax picture capturing method and system there is provided a kind of 3D rendering, it is intended to
Solve the disparity map hardware realization difficulty that existing parallax picture capturing method is obtained, the problem of accuracy rate is low.
The technical proposal for solving the technical problem of the invention is as follows：
A kind of parallax picture capturing method of 3D rendering, wherein, comprise the following steps：
A, left image respectively to 3D rendering and right image carry out gray proces, obtain corresponding half-tone information；
B, the half-tone information to the left image and right image carry out Block- matching processing, obtain including the three of disparity map information
Tie up matrix；
C, each row to the three-dimensional matrice carry out cost matrix processing, and every a line cost is obtained by viterbi algorithm
The corresponding parallax of matrix；
D, the parallax according to per a line form the disparity map of the 3D rendering.
The parallax picture capturing method of described 3D rendering, wherein, the step A is specifically included：
A1, the left image and right image be subjected to gray proces, be specially：I=R+G+B, wherein, R, G, B is the left side
Component value in image and right image before each pixel conversion；I be each pixel change after gray value, and I take R, G and
The most-significant byte of B sums.
The parallax picture capturing method of described 3D rendering, wherein, the step B is specifically included：
B1, definitionWithThe respectively half-tone information of left image and right image, wherein, with the upper left corner
Point is origin, pointFor the coordinate both horizontally and vertically of image；
B2, the half-tone information progress Block- matching processing by left image and right image described in Block- matching function pair, described piece
Adaptation function is：, wherein, h
For an integer, its span is between 1 to 4, and integer d is defined on intervalOn,With
It is integer；
B3, by left image d pixel of horizontal displacement to the right, pass through the Block- matching function and calculate the left image after displacement
With right image centered on point (x, y), (2h+ 1) * (2h+1) it is individual it is point-to-point between difference sum, so as to obtain corresponding
Include the three-dimensional matrice of disparity map information.
The parallax picture capturing method of described 3D rendering, wherein, the step C is specifically included：
C1, each row to the three-dimensional matrice carry out cost matrix processing, obtain the corresponding cost matrix of every a line；
Disparity constraint condition and optimization object function that C2, basis are pre-set, construct hidden Markov model, to each
The corresponding cost matrix of row is optimized；By viterbi algorithm find out it is each exercise the optimization object function value it is minimum when pair
The parallax answered.
The parallax picture capturing method of described 3D rendering, wherein, the step C1 is specifically included：
C11, each row to the three-dimensional matrice carry out cost matrix processing；The three-dimensional matrice is X*Y* (dmax-dmin
+ 1) three-dimensional matrice, wherein, X and Y be respectively 3D rendering both vertically and horizontally on pixel quantity；Pass through described piece
Adaptation function calculates the cost matrix J of m rowsmFor, wherein, y ∈ [1, Y], m
∈ [1, X], K=dmax-dmin+ 1。
The parallax picture capturing method of described 3D rendering, wherein, the step C2 is specifically included：
C21, m rows parallax is set as D={ D1..., DY, wherein, dmin≤Dy≤dmax, 1≤y≤Y；The disparity constraint
Condition is, i.e., the difference of the parallax adjacent with a line is no more than a certain positive integer N；
C22, definition cost function P (x) are met：0≤P(1)≤P(2)≤……≤P(N)；Wherein, as x ＞ N, P (x)
For ∞；
C23, the optimization object function of definition m rows are：, its
In, Jm(Dy；Y) it is the cost matrix JmIn DyThe element of row y row；
C24, the corresponding cost matrix of every a line is optimized, being found out by viterbi algorithm makes Em（D）Minimum regards
Poor D draws as the parallax of the m rows。
The parallax picture capturing method of described 3D rendering, wherein, the step D includes：
D1, the parallax matrix arranged according to the corresponding X rows Y of the corresponding parallax composition of every a line of the three-dimensional matrice；
D2, the parallax matrix are the corresponding disparity map of the 3D rendering.
A kind of disparity map of 3D rendering obtains system, wherein, including：
Gray scale acquisition module, carries out gray proces for left image respectively to 3D rendering and right image, obtains corresponding
Half-tone information；
Block- matching processing module, carries out Block- matching processing for the half-tone information to the left image and right image, obtains
Include the three-dimensional matrice of disparity map information；
Cost matrix disparity computation module, carries out cost matrix processing for each row to the three-dimensional matrice, passes through
Viterbi algorithm obtains the corresponding parallax of every a line cost matrix；
Disparity map generation module, forms the disparity map of the 3D rendering for the parallax according to per a line.
The disparity map of described 3D rendering obtains system, wherein, the Block- matching processing module includes：
Block- matching computing unit, for definingWithRespectively the gray scale of left image and right image is believed
Breath, wherein, the point using the upper left corner is origin, pointFor the coordinate both horizontally and vertically of image；Pass through Block- matching function
Block- matching processing is carried out to the half-tone information of the left image and right image, the Block- matching function is：
,
Wherein, h is an integer, and its span is between 1 to 4, and integer d is defined on intervalOn,WithIt is integer；
Three-dimensional matrice generation unit, for by left image d pixel of horizontal displacement to the right, passing through the Block- matching function
Calculate displacement after left image and right image centered on point (x, y), (2h+ 1) * (2h+1) it is individual it is point-to-point between difference it
With so as to obtain the corresponding three-dimensional matrice for including disparity map information.
The disparity map of described 3D rendering obtains system, wherein, the cost matrix disparity computation module includes：
Cost matrix acquiring unit, carries out cost matrix processing for each row to the three-dimensional matrice, obtains each
The corresponding cost matrix of row；
Disparity computation unit, for according to the disparity constraint condition and optimization object function pre-set, constructing hidden Ma Er
Can husband's model, cost matrix corresponding to every a line optimizes；Each enforcement optimization mesh is found out by viterbi algorithm
Corresponding parallax when offer of tender numerical value is minimum.
The parallax picture capturing method and system of a kind of 3D rendering provided by the present invention, efficiently solve existing parallax
The disparity map hardware that picture capturing method is obtained realizes difficult, the problem of accuracy rate is low, by left image respectively to 3D rendering and
Right image carries out gray proces, obtains corresponding half-tone information；Block is carried out to the half-tone information of the left image and right image
With processing, obtain including the three-dimensional matrice of disparity map information；Cost matrix processing is carried out to each row of the three-dimensional matrice, led to
Cross viterbi algorithm and obtain the corresponding parallax of every a line cost matrix；Parallax according to every a line forms the 3D rendering
Disparity map；By the Optimal Construction after Block- matching into hidden markov process, using viterbi algorithm in all possible parallaxes
The disparity map of maximum likelihood is found in figure, maximum likelihood disparity map and real disparity map gap are minimum so that obtained parallax
Figure is more accurate, and seeks the optimum solution of disparity map by viterbi algorithm, greatly reduces amount of calculation, can be realized and obtained by hardware
Disparity map is taken, brings and greatly facilitates.
Brief description of the drawings
The flow chart of the parallax picture capturing method preferred embodiment for the 3D rendering that Fig. 1 provides for the present invention.
The schematic diagram of viterbi algorithm in the parallax picture capturing method for the 3D rendering that Fig. 2 provides for the present invention.
The disparity map for the 3D rendering that Fig. 3 provides for the present invention obtains the structured flowchart of system preferred embodiment.
Embodiment
The present invention provides the parallax picture capturing method and system of a kind of 3D rendering, to make the purpose of the present invention, technical scheme
And advantage is clearer, clear and definite, the present invention is described in more detail for the embodiment that develops simultaneously referring to the drawings.It should be appreciated that this
The specific embodiment of place description is not intended to limit the present invention only to explain the present invention.
Referring to Fig. 1, the flow chart of the parallax picture capturing method preferred embodiment for the 3D rendering that Fig. 1 provides for the present invention,
As illustrated, the described method comprises the following steps：
Step S100, left image respectively to 3D rendering and right image carry out gray proces, obtain corresponding half-tone information；
Step S200, the half-tone information to the left image and right image carry out Block- matching processing, obtain including disparity map
The three-dimensional matrice of information；
Step S300, each row to the three-dimensional matrice carry out cost matrix processing, are obtained often by viterbi algorithm
The corresponding parallax of a line cost matrix；
Step S400, the parallax according to per a line form the disparity map of the 3D rendering.
Above-mentioned steps are described in detail with reference to specific embodiment.
In the step s 100, respectively to 3D rendering left image and right image carries out gray proces, obtains corresponding gray scale
Information.Specifically, it is exactly that gray proces are carried out to the left image and right image, is specially：I=R+G+B, wherein, R, G, B
For the component value before each pixel conversion in the left image and right image；I is the gray value after each pixel is changed, and
I takes the most-significant byte of R, G and B sum.Due to the RGB color signal that the TV signal of input is 24 bits.The gray scale produced by RGB
Signal definition is： I=0.2989R+0.5870G+0.1140B.For reduction computational complexity, the present invention is believed using 8 bit gradations
Breath.It is defined as by the RGB grey scale signals produced：I=R+G+B, and I takes the most-significant byte of RGB sums.
In step s 200, Block- matching processing is carried out to the half-tone information of the left image and right image, obtained comprising regarding
The three-dimensional matrice of poor figure information.Specifically, left image and the corresponding half-tone information of right image can obtain by step S100, definesWithThe respectively half-tone information of left image and right image, wherein, the point using the upper left corner is origin, pointFor the coordinate both horizontally and vertically of image.Carry out Block- matching function SAD (the summation of of Block- matching processing
Absolute difference) it is defined as：,
Wherein, h is an integer, and its span is between 1 to 4, and integer d is defined on intervalOn,With
It is integer.The Block- matching function is meant left image d pixel of horizontal displacement to the right, calculates the left figure after displacement
With right figure it is point-to-point between grey value difference, by the Block- matching function calculate left image and the right image after displacement with
Centered on point (x, y), (2h+ 1) * (2h+1) it is individual it is point-to-point between difference sum so that obtain it is corresponding include disparity map
The three-dimensional matrice J of information.The three-dimensional matrice J is X*Y* (dmax-dmin+ 1) three-dimensional matrice, wherein, X and Y are respectively 3D figures
As the pixel quantity on both vertically and horizontally, that is to say, that the three-dimensional matrice J, its preceding two peacekeepings input picture
In the same size, the size of the third dimension is。
In step S300, cost matrix processing is carried out to each row of the three-dimensional matrice, passes through viterbi algorithm
(Viterbi Algorithm) obtains the corresponding parallax of every a line cost matrix.In particular it is necessary in explanation, left figure
Point (x, y) parallax is that D is meant that:Xth row y row pixel corresponds to the pixel that the xth row y+D of right figure are arranged in left figure.
The step S300 is specifically included：S310, each row to the three-dimensional matrice carry out cost matrix processing, obtain
Per the corresponding cost matrix of a line；
Disparity constraint condition and optimization object function that S320, basis are pre-set, construct hidden Markov model
(Hidden Markove Model), cost matrix corresponding to every a line is optimized；Found out by viterbi algorithm each
Corresponding parallax when exercising the optimization object function value minimum.
In practical application, first carrying out cost matrix processing to each row of the three-dimensional matrice；The three-dimensional matrice is
X*Y*(dmax-dmin+ 1) three-dimensional matrice, wherein, X and Y be respectively 3D rendering both vertically and horizontally on pixel count
Amount；The cost matrix J of m rows is calculated by the Block- matching functionmFor, its
In, y ∈ [1, Y], m ∈ [1, X], K=dmax-dmin+ 1.Y is the label of pixel from left to right, such as the 31st pixel,
Then y is 31, and its span is 1 to Y.D is dminWith dmaxBetween all integer.Corresponding m value arrives X for 1.
Further according to the disparity constraint condition and optimization object function pre-set, specifically, the step S320 is specific
Including：
S321, m rows parallax is set as D={ D1..., DY, wherein, dmin≤Dy≤dmax, 1≤y≤Y；The parallax is about
Beam condition is, i.e., the difference of the parallax adjacent with a line is no more than a certain positive integer N；
S322, definition cost function P (x) are met：0≤P(1)≤P(2)≤……≤P(N)；Wherein, as x ＞ N, P
(x) it is ∞；
S323, the optimization object function of definition m rows are：, its
In, Jm(Dy；Y) it is the cost matrix JmIn DyThe element of row y row；
S324, the corresponding cost matrix of every a line is optimized, being found out by viterbi algorithm makes Em（D）Minimum regards
Poor D draws as the parallax of the m rows.That is the target optimized is to look for making Em（D）Minimum regards
Poor D is the parallax of the m rows.By above-mentioned constraints and optimization object function, hidden markov process is constructed, so
Afterwards by viterbi algorithm can find out it is each exercise the optimization object function value it is minimum when corresponding parallax.
In practical application, the viterbi algorithm is specific as follows：
First the implication of the symbol to will be detailed below occurring carries out concentration explanation.D R= dmin - dmax+ 1；c = N + 1.a
← b refers to assign a by b.For The row vector of row element composition；For Row element group
Into column vector.For RowRow are to theThe row vector of column element composition.For's
TheArrangeRow is to theThe column vector of row element composition.
Algorithm flow is as follows：
S1、；S2、；S3, construction andAn equal amount of full null matrix；S4, constructionMatrix A.
The step S4 is specially：S41、；S42, for 1≤j≤N, calculateWith
；S43, the calculating of matrix A can be represented with following matrix：
；
S44, reconstruct vector；S45、；S46、；
S47、；S48、.So as to obtain matrix A (A
Element be located between two parallel lines, the number on matrix is row label).
S5, the minimum value and its corresponding rower for calculating each row of A；S5 is specifically included：S51、All row are most
Small value；S52、Rower corresponding to the minimum value of all row.
S6、；
S7、
If S8,, output；Otherwise, return to S4；
S9, definition are a length ofVector；
S10、；
S11、；
S12、；
S13、；
If S14,,, return to S12；
S15, the m obtained disparity map of optimization of passing through are dmapm←
Above step is repeated to all rows, the disparity map dmap of entire image is just can obtain.
Further, above-mentioned algorithm is described below, referring to Fig. 2, the disparity map for the 3D rendering that Fig. 2 provides for the present invention
The schematic diagram of viterbi algorithm in acquisition methods.Laterally numeral represents the horizontal coordinate of image, i.e. y.Longitudinal direction numeral represents parallax,
Here disparity range is -1 to 2, i.e. dmin=-1, dmax= 2.Circle in figure is referred to as state, with vertical and horizontal numeral to coming
It is identified, such as S (0；3) the 3rd circle of the 2nd row, its physical significance is that the parallax of second pixel point is 0 state.On
Stating algorithm can be expressed as finding the process of survivor path in Fig. 2 grid.
Referring to Fig. 2, step is as follows：
H1, generalIt is filled into from top to bottom in first row circle, withRepresent this group of number.
H2, the 2nd four states arranged are shifted by four states in prostatitis, are represented with the line for having arrow, these
Line is referred to as path.
H3, the weight for calculating four paths for entering each stateThe weight of state S (d, 2) four paths is entered
Calculated by following formula：。
H4, the state computation all to the 2nd row, and choose respective minimum value to update, ;Note
Record each obtains minimum value, i.e. this state is which state transfer by previous column, shown in FIG as reality
Line, referred to as survivor path.
H5, to the right advance one are arranged, and repeat H3- H4, until the right column of grid chart, is arranged in this diagram for the 4th.
H6, to select numerical value minimumCorresponding state;Since the state, institute is recorded from right to left along solid line
The state of experience assumes in this figureMinimum, obtained state diagram is：。
H7, output parallax 0,0,1,0.Thus intuitively it is demonstrated by the process of viterbi algorithm.
In step S400, the parallax according to every a line forms the disparity map of the 3D rendering.Specifically, in step
Obtained in rapid S300 after the corresponding parallax of every a line, it is corresponding just to constitute the 3D rendering according to the parallax that correspondence is obtained per a line
Disparity map.That is, constituting the parallax matrix that corresponding X rows Y is arranged according to the corresponding parallax of every a line of the three-dimensional matrice；
The parallax matrix is the corresponding disparity map of the 3D rendering.From step S300, m rows parallax is D={ D1..., DY
, and m ∈ [1, X], X and Y be respectively 3D rendering both vertically and horizontally on pixel quantity；That is, by calculating
The corresponding parallax of every a line of three-dimensional matrice, so that just can obtain a parallax matrix, i.e., the corresponding disparity map of described 3D rendering.
The parallax picture capturing method for the 3D rendering that the present invention is provided, is parallax between consecutive points according to the physical characteristic of image
Gap within a certain range, parallax is arranged from left to right according to the order of pixel, parallax possible to each pixel
Depending on the parallax of its leftmost pixel point, thus a Markov chain is constituted, by the Optimal Construction after Block- matching into hidden Ma Er
Can husband's process, using viterbi algorithm found in all possible disparity maps real disparity map gap it is minimum it is maximum seemingly
Right disparity map, that is, the optimum solution of disparity map is sought using viterbi algorithm, it is implemented as so that amount of calculation is reduced into hardware
Possible degree so that disparity map result is more accurate, brings to user and greatly facilitates.
Parallax picture capturing method based on above-mentioned 3D rendering, system is obtained present invention also offers a kind of disparity map of 3D rendering
System, as shown in figure 3, the system includes：
Gray scale acquisition module 10, carries out gray proces for left image respectively to 3D rendering and right image, obtains correspondence
Half-tone information；Specifically as described in step S100；
Block- matching processing module 20, carries out Block- matching processing for the half-tone information to the left image and right image, obtains
To the three-dimensional matrice for including disparity map information；Specifically as described in step S200；
Cost matrix disparity computation module 30, carries out cost matrix processing for each row to the three-dimensional matrice, leads to
Cross viterbi algorithm and obtain the corresponding parallax of every a line cost matrix；Specifically as described in step S300；
Disparity map generation module 40, forms the disparity map of the 3D rendering for the parallax according to per a line；Specifically
As described in step S400.
Further, the Block- matching processing module 20 includes：
Block- matching computing unit, for definingWithRespectively the gray scale of left image and right image is believed
Breath, wherein, the point using the upper left corner is origin, pointFor the coordinate both horizontally and vertically of image；Pass through Block- matching function
Block- matching processing is carried out to the half-tone information of the left image and right image, the Block- matching function is：, wherein, h is an integer, its
Span is between 1 to 4, and integer d is defined on intervalOn,WithIt is integer；
Three-dimensional matrice generation unit, for by left image d pixel of horizontal displacement to the right, passing through the Block- matching function
Calculate displacement after left image and right image centered on point (x, y), (2h+ 1) * (2h+1) it is individual it is point-to-point between difference it
With so as to obtain the corresponding three-dimensional matrice for including disparity map information.
Further, the cost matrix disparity computation module 30 includes：
Cost matrix acquiring unit, carries out cost matrix processing for each row to the three-dimensional matrice, obtains each
The corresponding cost matrix of row；
Disparity computation unit, for according to the disparity constraint condition and optimization object function pre-set, constructing hidden Ma Er
Can husband's model, cost matrix corresponding to every a line optimizes；Each enforcement optimization mesh is found out by viterbi algorithm
Corresponding parallax when offer of tender numerical value is minimum.
In summary, the parallax picture capturing method and system for a kind of 3D rendering that the present invention is provided, by scheming respectively to 3D
The left image and right image of picture carry out gray proces, obtain corresponding half-tone information；To the left image and the gray scale of right image
Information carries out Block- matching processing, obtains including the three-dimensional matrice of disparity map information；In generation, is carried out to each row of the three-dimensional matrice
Valency matrix disposal, the corresponding parallax of every a line cost matrix is obtained by viterbi algorithm；Parallax shape according to every a line
Into the disparity map of the 3D rendering；By the Optimal Construction after Block- matching into hidden markov process, using viterbi algorithm in institute
The disparity map of maximum likelihood is found in the possible disparity map having, maximum likelihood disparity map and real disparity map gap are minimum,
So that obtained disparity map is more accurate, and seeks the optimum solution of disparity map by viterbi algorithm, amount of calculation is greatly reduced, can
Realized by hardware and obtain disparity map, realize that simply cost is relatively low.
It should be appreciated that the application of the present invention is not limited to above-mentioned citing, for those of ordinary skills, can
To be improved or converted according to the above description, all these modifications and variations should all belong to the guarantor of appended claims of the present invention
Protect scope.
Claims (7)
1. the parallax picture capturing method of a kind of 3D rendering, it is characterised in that comprise the following steps：
A, left image respectively to 3D rendering and right image carry out gray proces, obtain corresponding half-tone information；
B, the half-tone information to the left image and right image carry out Block- matching processing, obtain including the three-dimensional square of disparity map information
Battle array；
C, each row to the three-dimensional matrice carry out cost matrix processing, and every a line cost matrix is obtained by viterbi algorithm
Corresponding parallax；
D, the parallax according to per a line form the disparity map of the 3D rendering；
The step A is specifically included：
A1, the left image and right image be subjected to gray proces, be specially：I=R+G+B, wherein, R, G, B is the left figure
Component value in picture and right image before each pixel conversion；I is the gray value after each pixel is changed, and I takes R, G and B
The most-significant byte of sum；
The step C is specifically included：
C1, each row to the three-dimensional matrice carry out cost matrix processing, obtain the corresponding cost matrix of every a line；
Disparity constraint condition and optimization object function that C2, basis are pre-set, construct hidden Markov model, to every a line pair
The cost matrix answered is optimized；By viterbi algorithm find out it is each exercise the optimization object function value it is minimum when it is corresponding
Parallax.
2. the parallax picture capturing method of 3D rendering according to claim 1, it is characterised in that the step B is specifically included：
B1, definition IL(x, y) and IR(x, y) is respectively the half-tone information of left image and right image, wherein, using the point in the upper left corner as
Origin, point (x, y) is the coordinate both horizontally and vertically of image；
B2, the half-tone information progress Block- matching processing by left image and right image described in Block- matching function pair, the Block- matching
Function is：Wherein, h is one
Integer, its span is between 1 to 4, and integer d is defined on interval [dmin,dmax] on, dminAnd dmaxIt is integer；
B3, by left image d pixel of horizontal displacement to the right, pass through the Block- matching function calculate the left image after displacement with it is right
Image is centered on point (x, y), (2h+1) * (2h+1) it is individual it is point-to-point between difference sum, so as to obtain corresponding comprising regarding
The three-dimensional matrice of poor figure information.
3. the parallax picture capturing method of 3D rendering according to claim 2, it is characterised in that the step C1 is specifically wrapped
Include：
C11, each row to the three-dimensional matrice carry out cost matrix processing；The three-dimensional matrice is X*Y* (dmax-dmin+1)
Three-dimensional matrice, wherein, X and Y be respectively 3D rendering both vertically and horizontally on pixel quantity；Pass through the Block- matching
Function calculates the cost matrix J of m rowsmForWherein, y ∈ [1, Y], m ∈
[1, X], K=dmax-dmin+1。
4. the parallax picture capturing method of 3D rendering according to claim 3, it is characterised in that the step C2 is specifically wrapped
Include：
C21, m rows parallax is set as D={ D1..., DY, wherein, dmin≤Dy≤dmax, 1≤y≤Y；The disparity constraint condition
For | Dy-Dy+1|≤N, the i.e. difference of the parallax adjacent with a line are no more than a certain positive integer N；
C22, definition cost function P (x) are met：0≤P(1)≤P(2)≤……≤P(N)；Wherein, as x ＞ N, P (x) is
∞；
C23, the optimization object function of definition m rows are：Wherein, Jm
(Dy；Y) it is the cost matrix JmIn DyThe element of row y row；
C24, the corresponding cost matrix of every a line is optimized, being found out by viterbi algorithm makes Em(D) minimum parallax D makees
For the parallax of the m rows, that is, draw
5. the parallax picture capturing method of 3D rendering according to claim 4, it is characterised in that the step D includes：
D1, the parallax matrix arranged according to the corresponding X rows Y of the corresponding parallax composition of every a line of the three-dimensional matrice；
D2, the parallax matrix are the corresponding disparity map of the 3D rendering.
6. a kind of disparity map of 3D rendering obtains system, it is characterised in that including：
Gray scale acquisition module, carries out gray proces for left image respectively to 3D rendering and right image, obtains corresponding gray scale
Information；
Block- matching processing module, Block- matching processing is carried out for the half-tone information to the left image and right image, comprising
The three-dimensional matrice of disparity map information；
Cost matrix disparity computation module, carries out cost matrix processing for each row to the three-dimensional matrice, passes through Wei Te
The corresponding parallax of every a line cost matrix is obtained than algorithm；
Disparity map generation module, forms the disparity map of the 3D rendering for the parallax according to per a line；
The gray scale acquisition module is additionally operable to the left image and right image carrying out gray proces, is specially：I=R+G+B, its
In, R, G, B is the component value before each pixel conversion in the left image and right image；I is after each pixel is changed
Gray value, and I takes the most-significant byte of R, G and B sum；
The cost matrix disparity computation module includes：
Cost matrix acquiring unit, carries out cost matrix processing for each row to the three-dimensional matrice, obtains every a line pair
The cost matrix answered；
Disparity computation unit, for according to the disparity constraint condition and optimization object function pre-set, constructing hidden Markov
Model, cost matrix corresponding to every a line is optimized；Each enforcement optimization aim letter is found out by viterbi algorithm
Corresponding parallax when numerical value is minimum.
7. the disparity map of 3D rendering according to claim 6 obtains system, it is characterised in that the Block- matching processing module
Including：
Block- matching computing unit, for defining IL(x, y) and IR(x, y) is respectively the half-tone information of left image and right image, its
In, the point using the upper left corner is origin, and point (x, y) is the coordinate both horizontally and vertically of image；Pass through Block- matching function pair institute
The half-tone information for stating left image and right image carries out Block- matching processing, and the Block- matching function is：Wherein, h is an integer, its
Span is between 1 to 4, and integer d is defined on interval [dmin,dmax] on, dminAnd dmaxIt is integer；
Three-dimensional matrice generation unit, for by left image d pixel of horizontal displacement to the right, being calculated by the Block- matching function
Left image and right image after displacement are centered on point (x, y), (2h+1) * (2h+1) it is individual it is point-to-point between difference sum, from
And obtain the corresponding three-dimensional matrice for including disparity map information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410383267.3A CN105376543B (en) | 2014-08-06 | 2014-08-06 | The parallax picture capturing method and system of a kind of 3D rendering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410383267.3A CN105376543B (en) | 2014-08-06 | 2014-08-06 | The parallax picture capturing method and system of a kind of 3D rendering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105376543A CN105376543A (en) | 2016-03-02 |
| CN105376543B true CN105376543B (en) | 2017-08-08 |
Family
ID=55378292
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410383267.3A Active CN105376543B (en) | 2014-08-06 | 2014-08-06 | The parallax picture capturing method and system of a kind of 3D rendering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105376543B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114390271B (en) * | 2020-10-19 | 2023-08-18 | 苏州佳世达光电有限公司 | System and method for distinguishing continuous image sequence |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102111637A (en) * | 2011-03-29 | 2011-06-29 | 清华大学 | Stereoscopic video depth map generation method and device |
| KR101896941B1 (en) * | 2012-06-25 | 2018-09-12 | 엘지이노텍 주식회사 | Stereo matching apparatus and its method |
| KR20140095838A (en) * | 2013-01-25 | 2014-08-04 | 한국전자통신연구원 | Apparatus and method for recovering three dimensional information |
| CN103136750B (en) * | 2013-01-30 | 2015-08-19 | 广西科技大学 | The Stereo matching optimization method of binocular vision system |
| CN103714543A (en) * | 2013-12-26 | 2014-04-09 | 南京理工大学 | Simple tree dynamic programming binocular and stereo matching method based on invariant moment spatial information |
-
2014
- 2014-08-06 CN CN201410383267.3A patent/CN105376543B/en active Active
Non-Patent Citations (3)
| Title |
|---|
| Hidden Markov Measure Fields for Disparity Estimation;Edgar arce 等;《Proceedings of the Fourth Mexican International Conference on Computer Science, 2003》;20030923;全文 * |
| Stereo Matching with Reliable Disparity Propagation;Xun Sun等;《3D Imaging, Modeling, Processing, Visualization and Transmission (3DIMPVT), 2011 International Conference on》;20110718;全文 * |
| 就双目立体视觉的视差估计算法研究;陈瑞芳;《中国优秀硕士学位论文全文数据库》;20120831;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105376543A (en) | 2016-03-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3540637B1 (en) | Neural network model training method, device and storage medium for image processing | |
| WO2020001168A1 (en) | Three-dimensional reconstruction method, apparatus, and device, and storage medium | |
| CN106101535B (en) | A kind of video stabilizing method based on part and mass motion disparity compensation | |
| CN108335322A (en) | Depth estimation method and device, electronic equipment, program and medium | |
| JP7026222B2 (en) | Image generation network training and image processing methods, equipment, electronics, and media | |
| US20160156898A1 (en) | Parallel scaling engine for multi-view 3DTV display and method thereof | |
| TWI545523B (en) | Image distortion correction method and image distortion correction device using the same | |
| CN108280858B (en) | Linear global camera motion parameter estimation method in multi-view reconstruction | |
| KR100897542B1 (en) | Method and Device for Rectifying Image in Synthesizing Arbitary View Image | |
| CN108416751A (en) | A kind of new viewpoint image combining method assisting full resolution network based on depth | |
| CN107483911A (en) | A kind of signal processing method and system based on more mesh imaging sensors | |
| CN104809719A (en) | Virtual view synthesis method based on homographic matrix partition | |
| CN112070821A (en) | Low-power-consumption stereo matching system and method for acquiring depth information | |
| CN108921942A (en) | The method and device of 2D transformation of ownership 3D is carried out to image | |
| CN106997579A (en) | The method and apparatus of image mosaic | |
| CN103067671A (en) | Method and device of image display | |
| CN105279729A (en) | Three-dimensional information encryption method based on integral imaging and multi-chaotic system | |
| CN105376543B (en) | The parallax picture capturing method and system of a kind of 3D rendering | |
| Fan et al. | RS-DPSNet: Deep plane sweep network for rolling shutter stereo images | |
| CN104184981A (en) | Low-resolution display method and device based on downsampling | |
| CN105611269B (en) | Real-time disparity computation system based on FPGA | |
| CN105578168B (en) | The parallax drawing generating method and system of a kind of 3D rendering | |
| CN112541972A (en) | Viewpoint image processing method and related equipment | |
| CN104902260B (en) | The acquisition methods and system of a kind of image parallactic | |
| US20180158194A1 (en) | Determining Optical Flow |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C41 | Transfer of patent application or patent right or utility model | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20160620 Address after: 518052 Guangdong city of Shenzhen province Nanshan District Xili liuxiandong Zhongshan Road No. 1001 TCL Science Park Research Building D4 Building 8 layer B1 unit 802-1. Applicant after: SHENZHEN TCL HIGH-TECH DEVELOPMENT CO., LTD. Address before: 516006 Guangdong province Huizhou Zhongkai hi tech Development Zone No. nineteen District Applicant before: TCL Group Co., Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |