CN103139554B - A kind of 3 D video Rate-distortion optimization method and optimization device - Google Patents
A kind of 3 D video Rate-distortion optimization method and optimization device Download PDFInfo
- Publication number
- CN103139554B CN103139554B CN201110374120.4A CN201110374120A CN103139554B CN 103139554 B CN103139554 B CN 103139554B CN 201110374120 A CN201110374120 A CN 201110374120A CN 103139554 B CN103139554 B CN 103139554B
- Authority
- CN
- China
- Prior art keywords
- distortion
- rate
- information
- coding
- texture
- 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 a kind of 3 D video Rate-distortion optimization method to include: (1) uses texture information and side information combined calculation unicode rate;(2) texture information and side information combined calculation Combine distortion are used;(3) unicode rate is utilized to calculate associating rate distortion costs with Combine distortion;Described associating rate distortion costs is used for texture information and the combined coding of side information or the coding for side information.The invention also discloses a kind of 3 D video rate-distortion optimization device.Texture information in 3 D video and side information are joined together to carry out rate-distortion optimization by the present invention, texture information and the distortion performance potentiality in side information are fully excavated, improve the distortion performance of 3 D video compression, there is wide application and adaptability.
Description
Technical field
The present invention relates to the Encoder Optimization technology in a kind of coding and decoding video field, the particularly compressed encoding of 3 D video
Rate-distortion optimization technology in technology.
Background technology
At present, joint video team (JVT, Joint Video Team) the H.264/AVC video compression standard generation formulated
The state-of-the-art video image compression technology of table.Because which employs motion prediction based on rate-distortion optimization and model selection skill
Art, and the video coding and decoding technology of other a series of advanced persons, the most H.264/AVC standard and other coding and decoding videos of same time
Standard is compared outstanding distortion performance.Due to the excellent properties of rate-distortion optimization algorithm, its have become as one important
Encoder Optimization technology, be widely applied in coding and decoding video field.
The method of conventional rate-distortion optimization is for calculating and optimizing Lagrange rate distortion costs function method.Select in mode
On rate-distortion optimization as a example by, optimize process for search current coded unit, such as macro block, pattern, make following glug bright
Day function is minimum:
JT(k)=D (T, k)+λ R (T, k)
In above-mentioned function, J represents Lagrangian cost value;
T represents the elementary area of present encoding, such as current macro;
K is the coding mode of current coded unit, such as 8x8 pattern in frame;
D represents the distortion or its estimated value that current coded unit T k in mode brings when encoding;
R represents code check required when encoding for current coded unit T k in mode or its estimated value;
λ represents Lagrange multiplier, and in H.264/AVC standard, this value is determined by quantization parameter Qp.
Pattern k obtained through said process is in rate distortion sense optimum pattern, selects this pattern to make in coding
For current coded unit coding mode in final code stream.That is, by the pattern that rate distortion costs is minimum, as rate-distortion optimization
Result, be used for encoding.In some cases, optimization model is also likely to be the pattern that rate distortion costs is maximum.
Since the TV invention forties in 20th century, experienced by black-and-white television, color television and the numeral developed is high
Definition television three phases, TV tech the most gradually develops to giant-screen, colorization, fine definition, multimedia direction.Three-dimensional
TV (3DTV) has depth perception and telepresenc, it will makes spectators obtain third dimension to greatest extent and is subject to, thus will become and
One new developing direction.Research to the compression coding technology of the 3 D video being applied to three-dimensional television is significant.
Texture information is i.e. the two-dimensional image sequence that currently used video camera obtains.Obtain texture information be one by three-dimensional
Real world projects to the process of camera image plane.During this, the information of the third dimension is dropped so that only rely on stricture of vagina
Reason information can not obtain all information of corresponding three-dimensional scenic.
Three-dimensional scenic is expressed, it is necessary to introduce the side information corresponding with texture information, to represent for perfect
Three-dimensional information.It is, in general, that the resolution of each frame of side information is consistent with texture information or lower, and use gray-scale map
Form, the gray value of each pixel characterizes the information such as degree of depth of this point in respective texture information.Utilize texture information and
Corresponding side information and necessary camera parameters, can complete representation three-dimensional scenic.The most common supplementary letter
Breath form is the depth information will expressed with gray value after the far and near information of corresponding point is direct or transformed on texture information.
Human eye produces third dimension and is based primarily upon binocular parallax (binocular parallax) and motion parallax (motion
Parallax) two aspects.If using a device to make right and left eyes be respectively seen different anaglyphs, then will in brain
Produce accurate three-dimensional body, and the location that this object is in the scene, here it is have the third dimension of the degree of depth.In order to reach this
One effect, needs to utilize one or more pairs of texture and side information to generate corresponding single eye images respectively.
Under three-dimensional television applied environment, except the texture information via transmission, user side need to utilize benefit toward contact
Fill information combined with texture information and synthesize the more virtual visual point image single eye images as confession spectators' viewing, to obtain preferably
Three-dimensional sense is subject to.In three-dimensional television field, by different as being referred to as the projection of the three-dimensional scenic obtained under different cameras angle
Visual point image;Here, video camera can be that reality exists, it is also possible to is virtual, the viewpoint under virtual video camera angle
Image is referred to as virtual visual point image.For weighing the quality of 3 D video compressed encoding comprehensively, be considered as and assess may with
Family end is decoded or the quality of each visual point image that is synthesized into.
On the other hand, be required for carrying out encoding after acquisition for the texture information of three-dimensional television and side information so as
Channel.Fairly simple method is that texture information and side information are respectively with the existing coding techniques such as H.264/AVC
Encode.In coding, texture information is carried out respectively with the rate-distortion optimization of side information, only considered coded views
Image fault, and do not account for the virtual view distortion of synthesis;Only individually consider the coding code of texture information and side information
Rate, and do not account for the optimization problem under the unicode rate of the two.
Therefore, the information consolidation of texture information in 3 D video and side information is used, both are combined
Rate-distortion optimization, find under the rate distortion optimum meaning of virtual view distortion and texture and side information unicode rate texture and
The optimization model of side information, can improve 3 d video encoding efficiency and final three-dimensional video-frequency viewing quality.
In existing 3 D video compressed encoding, the problem existing for rate-distortion optimization technology is as follows:
In three-dimensional television system, the texture information of 3 D video and side information have coordinated the table of three-dimensional scene information
Reaching, only the two cooperating, is mutually matched, and could obtain the 3 D video display effect of optimum.But, existing rate distortion is excellent
Change technology only accounts for texture information and side information respective rate distortion optimal cost when encoding 3 D video, and does not considers
Texture information combines calculating and the optimization of rate distortion optimal cost with side information.
3 D video, in addition to transmission viewpoint, also needs to comprehensively utilize texture information at user side and obtains one with side information
The virtual view video image information of series, shows preferably 3-D effect for three-dimensional display apparatus.Existing rate-distortion optimization
Technology only accounts for transmitting the distortion of viewpoint when encoding 3 D video, and does not accounts for texture and complementary sequence regards in transmission
The size of the distortion that the distortion of point may cause on more multiple views.Therefore in the angle that whole 3 D video is applied, existing
Rate-distortion optimization method do not account for applying to the new composition of distortion during 3 d video encoding, be unable to this should
It is issued to the optimal performance in rate distortion sense with environment.
In view of feature and the applied environment of 3 D video, the rate distortion considering texture information and side information is special
Property, calculate and comprehensively weigh the encoder bit rate of texture information and side information and coded views that coding causes and virtual view
Comprehensive distortion, and carry out comprehensive rate-distortion optimization on this basis, it is possible to obtain more fully rate more excellent than existing method
Distortion coding efficiency.
Summary of the invention
An object of the present disclosure is to provide a kind of Rate-distortion optimization method for 3 D video.
A kind of 3 D video Rate-distortion optimization method, including: (1) uses texture information and the associating of side information combined calculation
Code check;(2) texture information and side information combined calculation Combine distortion are used;(3) unicode rate is utilized to calculate with Combine distortion
Associating rate distortion costs;Described associating rate distortion costs for texture information and the combined coding of side information or is believed for supplementing
The coding of breath.
As preferably, described associating rate distortion costs refers to use connection for the combined coding of texture information and side information
Close rate distortion costs and combine the coding mode determining texture information and side information;Determine in coding mode combining, by texture
The combination of the N number of pattern in information coding mode and M pattern in side information coding mode is selected associating rate distortion generation
The mode combinations that valency is optimum;Wherein: N is integer, 2 <=N <=K, K is the number of texture information coding mode;M is integer, 1
<=M <=L, L is the number of side information coding mode.
As preferably, described associating rate distortion costs refers to use connection for the combined coding of texture information and side information
Close rate distortion costs and combine the optimization quantized value determining texture information and side information conversion coefficient;Conversion coefficient is determined combining
Optimization quantized value in, by NL quantized value in texture information quantization of transform coefficients value and side information quantization of transform coefficients value
In ML quantized value combination in select the quantized value combination that associating rate distortion costs is optimum;Wherein: NL is integer, 2 <=
NL <=KL, KL is the number of texture information quantization of transform coefficients value to be selected;ML is integer, and 1 <=ML <=LL, LL is supplementary
The number of information conversion coefficient quantization value to be selected.
As preferably, described unicode rate RunionThe computational methods of (T, G) are Runion(T, G)=aTR(T)+aGR(G);Its
In: Runion(T, G) is unicode rate;T is texture information current coded unit;G is side information current coded unit;R (T) is
Following one of both: (1) texture information current coded unit encoder bit rate;(2) texture information current coded unit encoder bit rate
Estimated value;R (G) is following one of both: (1) side information current coded unit encoder bit rate;(2) side information present encoding
In cell encoding rate estimation value;aTFor real number, for texture information current coded unit encoder bit rate weighted value;aGFor real number, for
Side information current coded unit encoder bit rate weighted value.
As preferably, described Combine distortion DunionThe computational methods of (T, G) areIts
In: Dunion(T, G) is Combine distortion;T is texture information current coded unit;G is side information current coded unit;Di(T,
G) it is distortion corresponding to target view i;biFor real number, for target view i weight;P=1,2 ... for target view number;I=
1,2 ... P is target view label.
As preferably, the distortion D that described target view i is correspondingiThe computational methods of (T, G) are one of following four method:
(1) the distortion D that described target view i is correspondingiThe computational methods of (T, G) are Di(T, G)=E (Wi(T, G), Vi(T ', G '));Its
In: Di(T, G) is the distortion of target view i;T is texture information current coded unit;G is side information current coded unit;
T ' is that texture information current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;G ' is supplementary
Information current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;Wi(x, y) and Vi(x y) is profit
The method obtaining viewpoint i respective image information with side information y with texture x;(x, y) for calculating the method for difference between x and y for E;
(2) the distortion D that described target view i is correspondingiThe computational methods of (T, G) are Di(T, G)=F (T, T ', H (G, G '));Wherein: Di(T,
G) it is the distortion of target view i;T is texture information current coded unit;G is side information current coded unit;T ' is texture letter
Breath current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;G ' is side information present encoding
Unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;(x, y) for calculating the method for difference between x and y for H;
(x, y, z) for utilizing side information coding distortion z to rebuild the method for difference between texture y after calculating texture x and coding for F;(3) institute
State current coded unit and include the most described T of respective image of multiple viewpoint to be encoded by multiple TjComposition, described G is by multiple Gj
Composition, the distortion D that described target view i is correspondingiThe computational methods of (T, G) areIts
In: Di(T, G) is the distortion that target view i is corresponding;TjFor the corresponding unit to be encoded of texture information in viewpoint j to be encoded;GjFor treating
The corresponding unit to be encoded of side information in coded views j;DI, u(Tj, Gj) for utilizing viewpoint j to be encoded to generate corresponding target view i
Distortion corresponding to pixel u;cI, j, uFor real number, viewpoint j to be encoded during for calculating the Combine distortion of pixel u of target view i
Weight;I=1,2 ... P is target view label;P=1,2 ... for target view number;J=1,2 ... Q is to be encoded regarding
Piont mark;Q=1,2 ... for viewpoint number to be encoded;U=1,2 ... V is respective pixel piont mark;V=1,2 ... for right
Answer pixel number;(4) the distortion D that described target view i is correspondingiThe computational methods of (T, G) areWherein: Di(T, G) is the distortion of target view i;Tt is texture information present encoding
Unit and the most each coding unit;G is side information current coded unit;T ' be texture information current coded unit with
The reconstruction coding unit obtained after present encoding parameter coding, decoding;G ' is that side information current coded unit is with present encoding
The reconstruction coding unit obtained after parameter coding, decoding;N is pixel n in current coded unit;S is in current coded unit
Sum of all pixels;(x, y) for calculating the method for difference between x and y for H;(x, y) for the method calculating sensitive factor for I.
As preferably, described side information is depth information.
As preferably, Combine distortion DunionThe computational methods of (T, G) areWherein:
Dunion(T, G) is Combine distortion;T is texture information current coded unit;G is side information current coded unit;Di(T, G) is
The distortion that target view i is corresponding;biFor real number, for target view i weight;P=1,2 ... for target view number;I=1,
2 ... P is target view label.
As preferably, the distortion D that described target view i is correspondingiThe computational methods of (T, G) are one of following four method:
(1) the distortion D that described target view i is correspondingiThe computational methods of (T, G) are Di(T, G)=E (Wi(T, G), Wi(T ', G '));Its
In: Di(T, G) is the distortion of target view i;T is texture information current coded unit;G is side information current coded unit;
T ' is that texture information current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;G ' is supplementary
Information current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;Wi(x, y) for utilizing texture x
The method obtaining viewpoint i respective image information with side information y;(x, y) for calculating the method for difference between x and y for E;(2) described mesh
The distortion D that mark viewpoint i is correspondingiThe computational methods of (T, G) are Di(T, G)=F (T, T ', H (G, G '));Wherein: Di(T, G) is mesh
The distortion of mark viewpoint i;T is texture information current coded unit;G is side information current coded unit;T ' is that texture information is worked as
Front coding unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;G ' is side information present encoding list
Unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;(x, y) for calculating the function of difference between x and y for H;
(x, y z) rebuild the function of difference between texture y for combining after side information coding distortion z calculates texture x and coding to F;(3) institute
State current coded unit and include the most described T of respective image of multiple viewpoint to be encoded by multiple TjComposition, described G is by multiple Gj
Composition, the distortion D that described target view i is correspondingiThe computational methods of (T, G) areWherein:
Di(T, G) is the distortion that target view i is corresponding;TjFor the corresponding unit to be encoded of texture information in viewpoint j to be encoded;GjFor waiting to compile
The corresponding unit to be encoded of side information in code viewpoint j;DI, u(Tj, Gj) for utilizing viewpoint j to be encoded to generate the picture of corresponding target view i
The distortion that element u is corresponding;cI, j, uFor real number, the weight of viewpoint j to be encoded during for calculating the Combine distortion of pixel u of target view i;
I=1,2 ... P is target view label;P=1,2 ... for target view number;J=1,2 ... Q is viewpoint label to be encoded;
Q=1,2 ... for viewpoint number to be encoded;U=1,2 ... V is respective pixel piont mark;V=1,2 ... for corresponding pixel points
Number;(4) the distortion D that described target view i is correspondingiThe computational methods of (T, G) are
Wherein: Di(T, G) is the distortion of target view i;Tt is texture information current coded unit and the most each coding unit;
G is side information current coded unit;T ' is that texture information current coded unit is to obtain after present encoding parameter coding, decoding
Reconstruction coding unit;G ' is that side information current coded unit is compiled with the reconstruction obtained after present encoding parameter coding, decoding
Code unit;N is pixel n in current coded unit;S is sum of all pixels in current coded unit;H (x, y) for calculate x Yu y it
Between the function of difference;(x, y) for calculating the function of sensitive factor for I.
The present invention the second purpose is to provide a kind of 3 D video rate-distortion optimization device.
This device is by encoding and rebuilding module, unicode rate computing module, Combine distortion computing module, combine rate distortion generation
Valency computing module, poll optimize module composition;Described coding needs texture information current coded unit with supplementary with rebuilding module
Information current coded unit or one of both are as input, and are connected with unicode rate computing module, Combine distortion computing module,
Described unicode rate computing module is connected with combining rate distortion costs computing module, and described Combine distortion computing module needs shooting
Machine parameter is as input, and is connected with combining rate distortion costs computing module, and associating rate distortion costs computing module is excellent with poll
Changing module to be connected, poll optimizes module and is connected with rebuilding module with coding, and exports optimized parameter;Described coding and reconstruction module
For carrying out texture information and side information or the coding of one of both according to the coding parameter of input, and obtain rebuilding image,
Described unicode rate computing module is used for calculating texture information and side information combined coding code check, and described Combine distortion calculates mould
Block is for calculating the Combine distortion of texture information and side information, and described associating rate distortion costs computing module is used for utilizing associating
Code check calculates associating rate distortion costs with Combine distortion, and poll optimizes module and is used for judging rate distortion costs and making optimization certainly
Plan, when optimal conditions not yet meets, the setting of adjusting and optimizing parameter, when optimal conditions meets, by optimum coding parameter
Output.
Present invention beneficial effect compared with prior art:
1, the texture information in 3 D video and side information join together to carry out rate-distortion optimization innovatively, use
The unicode rate of texture information and side information is as the code check in rate-distortion optimization;Use distortion weighing and the work of target view
For the distortion in rate-distortion optimization.
2, by associating rate-distortion optimization, texture information and the distortion performance potentiality in side information have fully been excavated,
Improve the distortion performance of 3 D video compression.
3, by position and its weight of in present invention target view are adjusted flexibly, can obtain regarding for different three-dimensionals
Frequently the rate-distortion optimization result under applied environment, therefore the present invention has wide application and adaptability.
Accompanying drawing explanation
Rate-distortion optimization method for 3 D video a kind of to the present invention with embodiment is made with device below in conjunction with the accompanying drawings
Further illustrate.
Fig. 1 is a kind of 3 D video rate-distortion optimization device embodiment schematic diagram of the present invention;
Fig. 2 is a kind of 3 D video rate-distortion optimization device embodiment schematic diagram of the present invention;
Fig. 3 is a kind of 3 D video rate-distortion optimization device embodiment schematic diagram of the present invention;
Detailed description of the invention
The solution that the present invention be directed to the problem existing for 3 D video rate distortion and propose.
The embodiment of a kind of typical utilization 3 D video of the present invention rate-distortion optimization device is as it is shown in figure 1, by compiling
Code and reconstruction module, unicode rate computing module, Combine distortion computing module, to combine rate distortion costs computing module, poll excellent
Change module composition.Described coding is connected with unicode rate computing module, Combine distortion computing module with rebuilding module, unicode rate
Computing module with combine rate distortion costs computing module be connected, Combine distortion computing module need camera parameters as input,
And be connected with combining rate distortion costs computing module, associating rate distortion costs computing module optimizes module with poll and is connected, poll
Optimize and be connected with rebuilding module with coding.Described coding is used for carrying out texture information according to the coding parameter of setting with rebuilding module
With the Video coding of side information, described unicode rate computing module is used for calculating texture information and side information combined coding code
Rate, described Combine distortion computing module, for calculating the Combine distortion of texture information and side information, described combines rate distortion generation
Valency computing module is used for utilizing unicode rate to calculate associating rate distortion costs with Combine distortion, and poll optimizes module and is used for judging rate
Distortion cost also makes Optimal Decision-making, when optimal conditions not yet meets, and the setting of adjusting and optimizing parameter, meet in optimal conditions
Time, by optimum coding parameter output.In actual applications, the phase that the coding parameter of regulation and control and value thereof, coding module use
Pass technology, unicode rate and the computational methods of Combine distortion, rate distortion requirement the concrete grammar such as setting can be according to reality
Situation sweetly disposition.
Introduce the embodiment of several utilization a kind of 3 D video Rate-distortion optimization method of the present invention in detail below:
Example one:
This example describes a kind of 3 D video Rate-distortion optimization method, and coding unit is image block in this instance;Side information
For depth information.First the current coded unit to texture information coding uses independent rate-distortion optimization algorithm from texture information
The K kind pattern of coding selects the N kind pattern (wherein K > N) of optimum, then uses associating Rate-distortion optimization method to compile from current
The pattern of all permissions of the texture information coding mode of this N kind optimum of code unit and the depth information of current coded unit
Various combinations are chosen texture information coding mode and the combination of encoding depth information pattern of optimum, and by the pattern of this optimum
United application on the coding of the texture of current coded unit and the depth information of current coded unit;Above-mentioned rate of combining is lost
In true optimization, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is the stricture of vagina of current coded unit
Reason information is with code check during current examination pattern-coding, and R (G) is that the depth information of current coded unit is compiled with current institute die trial formula
Code check during code;In the associating rate-distortion optimization of this example, target view number is set to 1, and the computational methods of Combine distortion are
Dunion(T, G)=D1(T, G)=E (W1(T, G), W1(T ', G ')), wherein Dunion(T, G) is Combine distortion, D1(T, G) is target
The distortion that viewpoint is corresponding, T is texture information current block, and G is depth information sequence current block, and T ' is that texture information current block is to work as
The reconstructed block obtained after premode coding, decoding, G ' is that depth information sequence current block is to obtain after present mode coding, decoding
Reconstructed block, W1(x, y) for utilizing texture x and depth information y to obtain the virtual regarding generation side of target view respective image information
Method, (x, y) for calculating the function of total square error between x and y for E;The computational methods combining rate distortion costs in this example are J=
Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter.
Example two:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image macro in this instance;Mend
The information of filling is parallax information.Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use texture
The rate-distortion optimization algorithm of information consolidation uncoded parallax information information obtains optimum from the K kind pattern that texture information encodes
N kind pattern (wherein K >=N), use parallax information independence rate-distortion optimization algorithm from the L kind pattern that parallax information encodes
Obtaining optimum M kind pattern (wherein L > M), wherein texture information rate-distortion optimization algorithm is in texture information independence rate distortion
On the basis of optimized algorithm, the distortion to diverse location carries out different weightings and obtains, position include interior of articles, background,
Object boundaries etc., the judgement of position utilizes analysis parallax information to obtain, optimum in texture N kind optimization model and parallax M kind
The combination of pattern is used associating rate-distortion optimization choose the mode combinations of optimum, and optimum mode combinations is applied to texture
And on the coding of parallax current macro;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=
aTR(T)+aGR (G), wherein R (T) is the texture information current macro estimated value with present mode encoder bit rate, and R (G) is parallax
Information sequence current macro is with present mode encoder bit rate, aT, aGFor real number, selected by texture information and parallax information respectively
During several optimization model, calculated distortion meansigma methods determines;In associating rate-distortion optimization, target view number is set to 2,
The computational methods of Combine distortion are Wherein Dunion
(T, G) is Combine distortion, Di(T, G) is the distortion that target view i is corresponding, and T is texture information current macro, and G is parallax information
Sequence current macro, T ' is texture information current macro with the reconstruction macro block obtained after present mode coding, decoding, and G ' is parallax
Information sequence current macro is with the reconstruction macro block obtained after present mode coding, decoding, Wi(x, y) for utilizing texture x and parallax
Information y obtains the virtual regarding generation method, V of target view i respective image informationi(x, y) for utilizing texture x and parallax information y
Obtaining the simplification projecting method of target view i respective image information, (x, y) for calculating the letter of total absolute value error between x and y for E
Number;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is according to coded quantization parameter
The global parameter determined.
Example three:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;Use in optimization model selects
Texture information is combined the rate-distortion optimization algorithm of uncoded depth information information and is obtained from the K kind pattern that texture information encodes
Optimum N kind pattern (wherein K >=N), the rate-distortion optimization using depth information to combine uncoded texture information information is calculated
Method obtains the M kind pattern (wherein L > M) of optimum, wherein texture information rate-distortion optimization from the L kind pattern of encoding depth information
Distortion D=D in algorithm1(T, G)=E (W1(T, G), W1(T ', G)), D1(T, G) is the distortion of target view 1 correspondence, and T is stricture of vagina
Reason information current block, G is depth information sequence current block, and T ' is that texture information current block is to obtain after present mode coding, decoding
The reconstructed block arrived, W1(x, y) virtual the regarding for utilizing texture x and depth information y to obtain target view 1 respective image information generates
Method, (x, y) for calculating the function of total square error between x and y for E;Distortion D=D in depth information rate-distortion optimization algorithm1
(T, G)=E (W1(T, G '), W1(T, G)), D1(T, G) is the distortion of target view 1 correspondence, and T is texture information current block, and G is
Depth information sequence current block, T ' is that texture information current block is with the reconstructed block obtained after present mode coding, decoding, W1(x,
Y) for utilizing texture x and depth information y to obtain the virtual regarding generation method of target view 1 respective image information, (x, y) for meter for E
Calculate the function of total square error between x and y;Associating is used in the combination of texture N kind optimization model and degree of depth M kind optimization model
Rate-distortion optimization chooses the mode combinations of optimum, and optimum mode combinations is applied on the coding of texture and degree of depth current block;
In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=aTR(T)+aGR (G), wherein R (T) is texture letter
Breath current block is with the estimated value of present mode encoder bit rate, and R (G) is that depth information sequence current block is with present mode encoder bit rate
Estimated value, aT, aGFor real number, determined by the quantization parameter of texture information and depth information respectively;In associating rate-distortion optimization, mesh
Mark viewpoint number is set to 3, and the computational methods of Combine distortion are
Wherein Dunion(T, G) is Combine distortion, Di(T, G) is the distortion that target view i is corresponding, and T is texture information current block, and G is deep
Degree information sequence current block, T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, and G ' is the degree of depth
Information sequence current block is with the reconstructed block obtained after present mode coding, decoding, Wi(x, y) and Vi(x y) respectively utilizes texture
X and depth information y obtains different virtual regarding generation method of two kinds of target view i respective image information, and (x, y) for calculating x for E
And the function of average every pixel square error, b between yiFor real number, for target view i weight, target view i regard with coding
Between point, parallax size determines;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is root
Code check and distortion dynamics adjustment according to encoded sequence.
Example four:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains N kind pattern (the wherein K of optimum from the K kind pattern that texture information encodes
>=N), use associating rate-distortion optimization to choose the mould of optimum in the combination of texture N kind optimization model and all patterns of the degree of depth
Formula combines, and optimum mode combinations is applied on the coding of texture and degree of depth current block;In associating rate-distortion optimization,
The computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is that texture information current block is compiled with present mode
Code code check, R (G) is that depth information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view
Number is set to 1, and the computational methods of Combine distortion are Runion(T, G)=D1(T, G)=F1(T, T ', H (G, G ')), wherein: D1(T,
G) being the distortion of target view, T is that texture information works as block;G is side information current block;T ' is texture information current coded unit
With the reconstructed block obtained after present mode coding, decoding;G ' is that side information current coded unit is with present mode coding, decoding
After the reconstructed block that obtains;H (x, y) for calculate x and between the function of absolute value of each two corresponding point difference;F1(x, y z) are
Rebuild after calculating texture x and coding in conjunction with depth coding distortion z square error between texture y and function, concrete grammar is for depending on
Big brief summary and camera parameters according to z, it is thus achieved that the project migration in target view that distortion z causes, recycles this side-play amount and looks for
Knot and the position of corresponding point of degree of depth distortion calculate error between x with y;The computational methods of associating rate distortion costs are J
=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter.
Example five:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is the figure of two viewpoints in this instance
As block;) side information that uses is depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model
In selection, the rate-distortion optimization algorithm of texture information independence is used to obtain two viewpoints from the K kind pattern that texture information encodes
The N kind pattern (wherein K >=N) that current block is optimum, in the combination of texture N kind optimization model and the combination of all mode combinations of the degree of depth
The middle mode combinations using associating rate-distortion optimization to choose optimum, and optimum mode combinations is applied to texture and the degree of depth two
On the coding of individual viewpoint current block;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (T1)+
R(T2)+R(G1)+R(G2), wherein R (T1) it is that the texture information current block of viewpoint 1 to be encoded is with present mode encoder bit rate, R
(G1) it is that the depth information sequence current block of viewpoint 1 to be encoded is with present mode encoder bit rate;R(T2) it is the stricture of vagina of viewpoint 2 to be encoded
Reason information current block is with present mode encoder bit rate, R (G2) it is that the depth information sequence current block of viewpoint 2 to be encoded is to work as front mould
Formula encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, and the computational methods of Combine distortion are Wherein: Dunion(T, G)
For Combine distortion, D1(T, G) is the distortion that target view is corresponding, TjFor the texture information current block of viewpoint j to be encoded, GjFor treating
The depth information sequence current block of coded views j, Tj' be viewpoint j to be encoded texture information current block with present mode coding,
The reconstructed block obtained after decoding, Gj' be viewpoint j to be encoded depth information sequence current block with present mode coding, decoding after
The reconstructed block obtained, W1(x, y) for utilizing texture x and depth information y to obtain the virtual regarding raw of target view respective image information
One-tenth method, Eu(x, y) for the function of the square error between pixel u of calculating x and y, cI, j, uThe mistake caused for viewpoint j to be encoded
The very weight when calculating target view corresponding blocks pixel u, its computational methods are, when pixel u virtual regarding generation method W1(x,
Y) interpolation can be need not from viewpoint j to be encoded and directly obtain in, and can not directly obtain from another viewpoint, then c1, j, u=1, if
Interpolation can not be need not from viewpoint j to be encoded and directly obtain, can directly obtain from another viewpoint, then c1, j, u=0, if being unsatisfactory for
Any of the above-described, thenWherein hjFor the parallax range between viewpoint j to be encoded and target view;Associating rate distortion
The computational methods of cost are J=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter.
Example six:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains N kind pattern (the wherein K of optimum from the K kind pattern that texture information encodes
>=N), use associating rate-distortion optimization to choose the mould of optimum in the combination of texture N kind optimization model and all patterns of the degree of depth
Formula combines, and optimum mode combinations is applied on the coding of texture and degree of depth current block;In associating rate-distortion optimization,
The computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is that depth information sequence current block is compiled with present mode
Code code check;In associating rate-distortion optimization, target view number is set to 1, and the computational methods of Combine distortion areWherein Dunion(T, G) is Combine distortion, D1(T, G) is that target regards
The distortion that point is corresponding, T is texture information current block and current block left side block and the right block, and G is depth information sequence current block,
T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, G ' be depth information sequence current block with
The reconstructed block that obtains after present mode coding, decoding, (x, y) for calculating the function of sensitive factor, the acting as of this function utilizes I
X, y and surrounding pixel thereof are calculated the change complexity of this image-region, and image change complexity is the highest, and this value is more
Greatly, I (T, Tn')=(Tn′-TN, l])2+(Tn′-TN, r)2, wherein Tn' for pixel u in reconstructed block, Tn.lFor T in original imagen′
One, the left side of respective pixel pixel, if Tn' on left picture boundary, then Tn.l=Tn, Tn.rFor T in original imagen' corresponding
One, the right of pixel pixel, if Tn' on image right margin, then Tn.l=Tn, (x, y) for calculating difference between x and y for H
Function, H (Gn, Gn')=| Gn′-Gn|;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), its
Middle λ is a fixing global parameter.
Example seven:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains N kind pattern (the wherein K of optimum from the K kind pattern that texture information encodes
>=N), use associating rate-distortion optimization to choose the mould of optimum in the combination of texture N kind optimization model and all patterns of the degree of depth
Formula combines, and optimum mode combinations is applied on the coding of texture and degree of depth current block;In associating rate-distortion optimization,
The computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is that depth information sequence current block is compiled with present mode
Code code check;In associating rate-distortion optimization, target view number is set to 1, and the computational methods of Combine distortion areWherein Dunion(T, G) is Combine distortion, D1(T, G) is that target regards
The distortion that point is corresponding, T is texture information current block and current block left side block and the right block, and G is depth information sequence current block,
T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, G ' be depth information sequence current block with
The reconstructed block that obtains after present mode coding, decoding, (x, y) for calculating the function of sensitive factor, the acting as of this function utilizes I
X, y and surrounding pixel thereof are calculated the change complexity of this image-region, and image change complexity is the highest, and this value is more
Greatly, Wherein Tn' for pixel n in reconstructed block, Tn.lFor original
T in imagenOne, the left side of ' respective pixel pixel, if Tn' on left picture boundary, then Tn.l=Tn, Tn.rFor original image
Middle TnOne, the right of ' respective pixel pixel, if Tn' on image right margin, then Tn.l=Tn, Tn.l' for Tn' respective pixel
One, left side pixel, if Tn' on left picture boundary, then Tn.l'=Tn', Tn.r' for TnOne, the right of ' respective pixel picture
Element, if Tn' on current block right margin, then Tn.r'=Tn', (x, y) for calculating the function H (G of difference between x and y for Hn, Gn′)
=(Gn′-Gn)2;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is one fixing
Global parameter.
Example eight:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;The optimization quantized value using associating rate-distortion optimization to complete conversion coefficient selects;In conversion
During the optimization quantized value of coefficient selects, the KL kind of the rate-distortion optimization algorithm texture information coding of texture information independence is used to become
The quantized value changing coefficient obtains the quantized value (wherein KL >=NL) of optimum NL kind conversion coefficient, at texture NL kind optimal transformation
The combination of the quantized value of the quantized value of coefficient and all conversion coefficients of the degree of depth is used associating rate-distortion optimization choose the mould of optimum
Formula combines, and by the quantized value united application of optimum conversion coefficient on the coding of texture and degree of depth current block;In associating
In rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is that texture information is current
Block is with the current quantisation value encoder bit rate of conversion coefficient, and R (G) is the depth information sequence current block current quantisation with conversion coefficient
Value encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, and the computational methods of Combine distortion are Dunion(T, G)
=D1(T, G)=E (W1(T, G), W1(T ', G ')), wherein Dunion(T, G) is Combine distortion, D1(T, G) is that target view is corresponding
Distortion, T is texture information current block, and G is depth information sequence current block, and T ' is that texture information current block is with Current Transform system
The reconstructed block obtained after quantification value coding, decoding, G ' is that depth information sequence current block is compiled with current transform coefficient quantized value
The reconstructed block obtained after code, decoding, W1(x, y) for utilizing texture x and depth information y to obtain target view respective image information
Virtual regarding generation method, (x, y) for calculating the function of total square error between x and y for E;The computational methods of associating rate distortion costs
For J=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter.
Example nine:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, associating
The computational methods of distortion are Dunion(T, G)=D1(T, G)=E (W1(T, G), W1(T ', G ')), wherein Dunion(T, G) is associating
Distortion, D1(T, G) is the distortion that target view is corresponding, and T is texture information current block, and G is depth information sequence current block, and T ' is
Texture information current block is with the reconstructed block obtained after present mode coding, decoding, and G ' is that depth information sequence current block is with currently
The reconstructed block obtained after pattern-coding, decoding, W1(x, y) for utilizing texture x and depth information y to obtain target view respective image
Information virtual regarding generation method, (x, y) for calculating the function of total square error between x and y for E;The meter of associating rate distortion costs
Calculation method is J=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter.
Example ten:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 2, associating
The computational methods of distortion are Wherein Dunion(T,
G) it is Combine distortion, Di(T, G) is the distortion that target view i is corresponding, and T is texture information current macro, and G is parallax information sequence
Current macro, T ' is texture information current macro with the reconstruction macro block obtained after present mode coding, decoding, and G ' is parallax information
Sequence current macro is with the reconstruction macro block obtained after present mode coding, decoding, Wi(x, y) for utilizing texture x and parallax information y
Obtain the virtual regarding generation method, V of target view i respective image informationi(x, y) for utilizing texture x and parallax information y to obtain mesh
The simplification projecting method of mark viewpoint i respective image information, (x, y) for calculating the function of total absolute value error between x and y for E;Associating
The computational methods of rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is according to coded quantization parameter determination
Global parameter.
Example 11:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 3, associating
The computational methods of distortion are Wherein Dunion(T, G) is for combining mistake
Very, Di(T, G) is the distortion that target view i is corresponding, and T is texture information current block, and G is depth information sequence current block, and T ' is
Texture information current block is with the reconstructed block obtained after present mode coding, decoding, and G ' is that depth information sequence current block is with currently
The reconstructed block obtained after pattern-coding, decoding, Wi(x, y) and Vi(x y) respectively utilizes texture x and depth information y to obtain target
Different virtual regarding generation method of two kinds of viewpoint i respective image information, (x y) puts down for calculating average every pixel between x and y E
The function of side's error, biFor real number, for target view i weight, parallax size between target view i and coded views determine;Connection
The computational methods closing rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), the wherein code of encoded sequence according to λ
Rate adjusts with distortion dynamics.
Example 12:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, associating
The computational methods of distortion are Dunion(T, G)=D1(T, G)=F1(T, T ', H (G, G ')), wherein: D1(T, G) is target view
Distortion, T is that texture information works as block;G is side information current block;T ' is that texture information current coded unit is compiled with present mode
The reconstructed block obtained after code, decoding;G ' is that side information current coded unit is with the reconstruction obtained after present mode coding, decoding
Block;H (x, y) for calculate x and between the function of absolute value of each two corresponding point difference;F1(x, y, z) for combining depth coding
Distortion z rebuild after calculating texture x and coding square error between texture y and function, concrete grammar be the big brief summary according to z and
Camera parameters, it is thus achieved that the project migration in target view that distortion z causes, recycle this side-play amount find between x and y knot and
The position of the corresponding point of degree of depth distortion also calculates error;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ
Runion(T, G), wherein λ is a fixing global parameter.
Example 13:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, associating
The computational methods of distortion are
Wherein: Dunion(T, G) is Combine distortion, D1(T, G) is the distortion that target view is corresponding, TjTexture for viewpoint j to be encoded is believed
Breath current block, GjFor the depth information sequence current block of viewpoint j to be encoded, Tj' it is the texture information current block of viewpoint j to be encoded
With the reconstructed block obtained after present mode coding, decoding, Gj' it is that the depth information sequence current block of viewpoint j to be encoded is with currently
The reconstructed block obtained after pattern-coding, decoding, W1(x, y) for utilizing texture x and depth information y to obtain target view respective image
Information virtual regarding generation method, Eu(x, y) for the function of the square error between pixel u of calculating x and y, c1, j, uFor waiting to compile
The distortion that code viewpoint j the causes weight when calculating target view corresponding blocks pixel u, its computational methods are, when pixel u is virtual
Depending on generation method W1(x, y) in can need not interpolation from viewpoint j to be encoded and directly obtain, and directly can not obtain from another viewpoint
, then c1, j, u=1, if interpolation can not be need not from viewpoint j to be encoded and directly obtain, can directly obtain from another viewpoint, then
c1, j, u=0, if being unsatisfactory for any of the above-described point, thenWherein hjFor the baseline between viewpoint j to be encoded and target view
Distance;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing overall situation ginseng
Number.
Example 14:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, associating
The computational methods of distortion are Wherein Dunion(T, G) is for combining mistake
Very, D1(T, G) is the distortion that target view is corresponding, and T is texture information current block and current block left side block and the right block, and G is
Depth information sequence current block, T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, and G ' is deep
Degree information sequence current block with the reconstructed block that obtains after present mode coding, decoding, I (x, y) for calculating the function of sensitive factor,
Act as utilizing x, y and the surrounding pixel thereof of this function are calculated the change complexity of this image-region, and image change is multiple
Miscellaneous degree is the highest, and this value is the biggest, I (T, Tn')=(Tn′-TN, l])2+(Tn′-TN, r)2, wherein Tn' for pixel n in reconstructed block,
Tn.lFor T in original imagenOne, the left side of ' respective pixel pixel, if Tn' on left picture boundary, then Tn.l=Tn, Tn.rFor
T in original imagenOne, the right of ' respective pixel pixel, if Tn' on image right margin, then Tn.l=Tn, (x, y) for meter for H
Calculate the function H (G of difference between x and yn, Gn')=| Gn′-Gn|;The computational methods of associating rate distortion costs are J=Dunion(T, G)
+λRunion(T, G), wherein λ is a fixing global parameter.
Example 15:
This example describes a kind of 3 D video Rate-distortion optimization method, and current coded unit is image block in this instance;Use
Side information be depth information;Use associating rate-distortion optimization to complete optimization model to select;In optimization model selects, use
The rate-distortion optimization algorithm of texture information independence obtains optimum pattern, in the combination of all patterns of texture optimization model and the degree of depth
The middle pattern using associating rate-distortion optimization to choose depth coding optimum, and optimum mode combinations is applied to degree of depth current block
Coding on;In associating rate-distortion optimization, the computational methods of unicode rate are Runion(T, G)=R (G), wherein R (G) is deep
Degree information sequence current block is with present mode encoder bit rate;In associating rate-distortion optimization, target view number is set to 1, associating
The computational methods of distortion are Wherein Dunion(T, G) is for combining mistake
Very, D1(T, G) is the distortion that target view is corresponding, and T is texture information current block and current block left side block and the right block, and G is
Depth information sequence current block, T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, and G ' is deep
Degree information sequence current block with the reconstructed block that obtains after present mode coding, decoding, I (x, y) for calculating the function of sensitive factor,
Act as utilizing x, y and the surrounding pixel thereof of this function are calculated the change complexity of this image-region, and image change is multiple
Miscellaneous degree is the highest, and this value is the biggest, Wherein Tn' in reconstructed block
Pixel n, Tn.lFor T in original imagenOne, the left side of ' respective pixel pixel, if Tn' on left picture boundary, then Tn.l=
Tn, Tn.rFor T in original imagenOne, the right of ' respective pixel pixel, if Tn' on image right margin, then Tn.l=Tn,
Tn.l' for TnOne, the left side of ' respective pixel pixel, if Tn' on left picture boundary, then Tn.l'=Tn', Tn.r' for Tn' right
Answer one, the right pixel of pixel, if Tn' on current block right margin, then Tn.r'=Tn', (x, y) for calculating between x and y for H
Function H (the G of differencen, Gn')=(Gn′-Gn)2;The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T,
G), wherein λ is a fixing global parameter.
What above example one to example 15 described is to use a kind of 3 D video Rate-distortion optimization method of the present invention
Embodiment, the enforcement introducing several utilization a kind of 3 D video rate-distortion optimization device of the present invention in detail below is real
Example.
Example 16:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 1: by coding and reconstruction module, connection
Close code check computing module, Combine distortion computing module, associating rate distortion costs computing module, poll optimization module composition;Described
Coding with rebuild module need texture information current coded unit and side information current coded unit as input, and with combine
Code check computing module, Combine distortion computing module are connected;Described unicode rate computing module calculates mould with combining rate distortion costs
Block is connected;Described Combine distortion computing module need camera parameters as input, and with combine rate distortion costs computing module
It is connected;Described associating rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and coding and weight
Modeling block is connected, and exports optimum code mode combinations.Wherein: the unit that (1) described device processes is image block;(2) described
The 3 D video side information that device uses is depth information;(3) described parameters optimization is the volume of texture information and depth information
Pattern combines;(4) described coding with rebuild module for according to currently processed coding mode combination carry out texture information with
The Video coding of depth information current block, and obtain reconstruction image;(5) effect of the computing module of described unicode rate is to calculate
Unicode rate, in this module, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is texture letter
Breath current block is with present mode encoder bit rate, and R (G) is that depth information sequence current block is with present mode encoder bit rate;(6) described
Combine distortion computing module act as calculate Combine distortion, in this module, target view number is set to 1, Combine distortion
Computational methods are Dunion(T, G)=D1(T, G)=E (W1(T, G), W1(T ', G ')), wherein Dunion(T, G) is Combine distortion, D1
(T, G) is the distortion that target view is corresponding, and T is texture information current block, and G is depth information sequence current block, and T ' is texture letter
Breath current block is with the reconstructed block obtained after present mode coding, decoding, and G ' is that depth information sequence current block is compiled with present mode
The reconstructed block obtained after code, decoding, W1(x, y) for utilizing texture x and depth information y to obtain target view respective image information
Virtual regarding generation method, (x, y) for calculating the function of total square error between x and y for E;(7) described associating rate distortion costs calculates
Module act as calculate associating rate distortion costs, in this module, combine rate distortion costs computational methods be J=Dunion
(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter;(8) described poll optimize module act as control device
Each mode combinations to be selected is carried out rate-distortion optimization and selects the optimization model combination that associating rate distortion costs is minimum, and will fortune
For texture and the optimum mode combinations output of degree of depth current block.
Example 17:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Described pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding and reconstruction module
Need texture information current coded unit and side information current coded unit as input, and with unicode rate computing module,
Combine distortion computing module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described
Closing distortion computation module needs camera parameters as input and connected with combining rate distortion costs computing module;Described associating
Rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and
Output optimum code mode combinations.Wherein: the unit that (1) described device processes is image block;(2) three-dimensional that described device uses
Video side information is depth information;(3) effect of described pattern preliminary election module is the various coding modes to texture and the degree of depth
Carry out primary election, in this module, use the rate-distortion optimization algorithm of texture information independence from the K kind pattern that texture information encodes
Obtain optimum N kind pattern (wherein K >=N), using the combination of texture N kind optimization model and all patterns of the degree of depth as treating modeling
Formula;(4) described coding is used for carrying out texture information and depth information according to currently processed coding mode combination with rebuilding module
The Video coding of current block, and obtain reconstruction image;(5) effect of described unicode rate computing module is to calculate unicode rate,
In this module, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is texture information current block
With present mode encoder bit rate, R (G) is that depth information sequence current block is with present mode encoder bit rate;(6) described Combine distortion
Computing module act as calculate Combine distortion, in this module, target view number is set to 1, the computational methods of Combine distortion
For Runion(T, G)=D1(T, G)=E (W1(T, G), W1(T ', G ')), wherein Dunion(T, G) is Combine distortion, D1(T, G) is mesh
Mark distortion corresponding to viewpoint, T is texture information current block, and G is depth information sequence current block, T ' be texture information current block with
The reconstructed block obtained after present mode coding, decoding, G ' is that depth information sequence current block is to obtain after present mode coding, decoding
The reconstructed block arrived, W1(x, y) virtual the regarding for utilizing texture x and depth information y to obtain target view respective image information generates
Method, (x, y) for calculating the function of total square error between x and y for E;(7) effect of described associating rate distortion costs computing module
For calculating associating rate distortion costs, in this module, the computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion
(T, G), wherein λ is a fixing global parameter;(8) each is treated modeling by the control device that act as of described poll optimization module
Formula combination carries out rate-distortion optimization and selects the minimum optimization model of associating rate distortion costs and combine, and will apply to texture and
The optimum mode combinations output of degree of depth current block.
Example 18:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Described pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding and reconstruction module
Need texture information current coded unit and side information current coded unit as input, and with unicode rate computing module,
Combine distortion computing module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described
Closing distortion computation module needs camera parameters as input and connected with combining rate distortion costs computing module;Described associating
Rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and
Output optimum code mode combinations.Wherein: the unit that (1) described device processes is image macro;(2) the three of described device use
Dimension video side information is parallax information;(3) effect of described pattern preliminary election module is the various coding moulds to texture and parallax
Formula carries out primary election, in this module, use texture information combine uncoded parallax information information rate-distortion optimization algorithm from
The K kind pattern of texture information coding obtains the N kind pattern (wherein K >=N) of optimum, uses parallax information independence rate distortion excellent
Change algorithm from the L kind pattern that parallax information encodes, obtain the M kind pattern (wherein L > M) of optimum, wherein texture information rate distortion
Optimized algorithm is that distortion to diverse location carries out different weightings on the basis of texture information independence rate-distortion optimization algorithm
And obtain, position includes interior of articles, background, object boundary etc., and the judgement of position utilizes analysis parallax information to obtain,
Using the combination of texture N kind optimization model and parallax M kind optimization model as treating lectotype;(4) described coding is used with rebuilding module
In carrying out the Video coding of texture information and depth information current block according to currently processed coding mode combination, and obtain reconstruction
Image;(5) effect of the computing module of described unicode rate is to calculate unicode rate, in this module, the calculating of unicode rate
Method is Runion(T, G)=aTR(T)+aGR (G), wherein R (T) is that texture information current macro is with present mode encoder bit rate
Estimated value, R (G) is that parallax information sequence current macro is with present mode encoder bit rate, aT, aGFor real number, respectively by texture information
Determine with parallax information calculated distortion meansigma methods when selecting several optimization model;(6) described Combine distortion calculates mould
Block act as calculate Combine distortion, in this module, target view number is set to 2, and the computational methods of Combine distortion are Wherein Dunion(T, G) is Combine distortion, Di(T, G) is
The distortion that target view i is corresponding, T is texture information current macro, and G is parallax information sequence current macro, and T ' is texture information
Current macro is with the reconstruction macro block obtained after present mode coding, decoding, and G ' is that parallax information sequence current macro is to work as front mould
The reconstruction macro block obtained after formula coding, decoding, Wi(x, y) in order to utilize texture x and parallax information y to obtain, target view i is corresponding to be schemed
Virtual regarding generation method, V as informationi(x, y) for utilizing texture x and parallax information y to obtain target view i respective image information
Simplification projecting method, (x, y) for calculating the function of total absolute value error between x and y for E;(7) described associating rate distortion costs meter
That calculates module act as calculating associating rate distortion costs, and in this module, the computational methods of associating rate distortion costs are J=Dunion
(T, G)+λ Runion(T, G), wherein λ is the global parameter according to coded quantization parameter determination;(8) described poll optimizes module
It act as controlling device each mode combinations to be selected is carried out rate-distortion optimization and selects the optimum that associating rate distortion costs is minimum
Mode combinations, and the optimum mode combinations output of texture and degree of depth current block will be applied to.
Example 19:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Described pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding and reconstruction module
Need texture information current coded unit and side information current coded unit as input, and with unicode rate computing module,
Combine distortion computing module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described
Closing distortion computation module needs camera parameters as input and connected with combining rate distortion costs computing module;Described associating
Rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and
Output optimum code mode combinations.Wherein: the unit that (1) described device processes is image block;(2) three-dimensional that described device uses
Video side information is depth information;(3) effect of described pattern preliminary election module is the various coding modes to texture and the degree of depth
Carry out primary election, in this module, use texture information to combine the rate-distortion optimization algorithm of uncoded depth information information from stricture of vagina
The K kind pattern of reason information coding obtains the N kind pattern (wherein K >=N) of optimum, uses depth information to combine uncoded stricture of vagina
The rate-distortion optimization algorithm of reason information obtains the M kind pattern (wherein L > M) of optimum from the L kind pattern of encoding depth information, its
Distortion D=D in middle texture information rate-distortion optimization algorithm1(T, G)=E (W1(T, G), W1(T ', G)), D1(T, G) is target
The distortion of viewpoint 1 correspondence, T is texture information current block, and G is depth information sequence current block, T ' be texture information current block with
The reconstructed block obtained after present mode coding, decoding, W1(x, y) for utilizing texture x and depth information y to obtain target view 1 phase
Answering the virtual regarding generation method of image information, (x, y) for calculating the function of total square error between x and y for E;Depth information rate is lost
Distortion D=D in true optimized algorithm1(T, G)=E (W1(T, G '), W1(T, G)), D1(T, G) is the mistake of target view 1 correspondence
Very, T is texture information current block, and G is depth information sequence current block, T ' be texture information current block with present mode coding,
The reconstructed block obtained after decoding, W1(x, y) for utilizing texture x and depth information y to obtain the void of target view 1 respective image information
Intending regarding generation method, (x, y) for calculating the function of total square error between x and y for E;By texture N kind optimization model and degree of depth M kind
The combination of optimization model is as treating lectotype;(4) described coding is used for according to currently processed coding mode group with rebuilding module
Close the Video coding carrying out texture information with depth information current block, and obtain reconstruction image;(5) described unicode rate calculates mould
Block act as calculate unicode rate, in this module, the computational methods of unicode rate are Runion(T, G)=aTR(T)+aGR
(G), wherein R (T) is the texture information current block estimated value with present mode encoder bit rate, and R (G) is that depth information sequence is current
Block is with the estimated value of present mode encoder bit rate, aT, aGFor real number, determined by the quantization parameter of texture information and depth information respectively
Fixed;(6) described Combine distortion computing module act as calculate Combine distortion, in this module, target view number is set to 3,
The computational methods of Combine distortion are Wherein Dunion
(T, G) is Combine distortion, Di(T, G) is the distortion that target view i is corresponding, and T is texture information current block, and G is depth information sequence
Row current block, T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, and G ' is depth information sequence
Current block is with the reconstructed block obtained after present mode coding, decoding, Wi(x, y) and Vi(x y) respectively utilizes texture x and the degree of depth
Information y obtains different virtual regarding generation method of two kinds of target view i respective image information, and (x, y) for calculating between x and y for E
The function of average every pixel square error, biFor real number, for target view i weight, by parallax between target view i and coded views
Size determines;(7) described associating rate distortion costs computing module act as calculate associating rate distortion costs, in this module,
The computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein encoded sequence according to λ
Code check adjusts with distortion dynamics;(8) each mode combinations to be selected is entered by the control device that act as of described poll optimization module
Row rate-distortion optimization also selects the optimization model combination that associating rate distortion costs is minimum, and will apply to texture and the degree of depth is current
The optimum mode combinations output of block.
Example 20:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding needs with rebuilding module
Texture information current coded unit and side information current coded unit as input, and with unicode rate computing module, combine
Distortion computation module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described combine mistake
True computing module needs camera parameters as input and connected with combining rate distortion costs computing module;Described rate of combining is lost
True cost computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and exports
Optimum code mode combinations.Wherein: the unit that (1) described device processes is image block;(2) 3 D video that described device uses
Side information is depth information;(3) effect of described pattern preliminary election module is that the various coding modes to texture and the degree of depth are carried out
Primary election, in this module, uses the rate-distortion optimization algorithm of texture information independence to obtain from the K kind pattern that texture information encodes
Optimum N kind pattern (wherein K >=N), using the combination of texture N kind optimization model and all patterns of the degree of depth as treating lectotype;
(4) described coding is current with depth information for carrying out texture information according to currently processed coding mode combination with reconstruction module
The Video coding of block, and obtain reconstruction image;(5) effect of described unicode rate computing module is to calculate unicode rate, at this
In module, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is that texture information current block is to work as
Premode encoder bit rate, R (G) is that depth information sequence current block is with present mode encoder bit rate;(6) described Combine distortion calculates
Module act as calculate Combine distortion, in this module, target view number is set to 1, and the computational methods of Combine distortion are
Dunion(T, G)=D1(T, G)=F1(T, T ', H (G, G ')), wherein: D1(T, G) is the distortion of target view, and T is texture information
Work as block;G is side information current block;T ' is that texture information current coded unit is with the weight obtained after present mode coding, decoding
Build block;G ' is that side information current coded unit is with the reconstructed block obtained after present mode coding, decoding;(x, y) for calculating x for H
The function of the absolute value of each two corresponding point difference between and;F1(x, y, z) for combine depth coding distortion z calculate texture x with
Rebuilding the function of square error sum between texture y after coding, concrete grammar is the big brief summary according to z and camera parameters, it is thus achieved that
The project migration in target view that distortion z causes, recycles this side-play amount and finds between x with y the relative of knot and degree of depth distortion
The position that should put also calculates error;(7) described associating rate distortion costs computing module act as calculate associating rate distortion costs,
In this module, the computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ is one fixing
Global parameter;(8) the control device that act as of described poll optimization module carries out rate-distortion optimization to each mode combinations to be selected
And select the optimization model combination that associating rate distortion costs is minimum, and the optimum mould of texture and degree of depth current block will be applied to
Formula combination output.
Example 21:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding needs with rebuilding module
Texture information current coded unit and side information current coded unit as input, and with unicode rate computing module, combine
Distortion computation module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described combine mistake
True computing module needs camera parameters as input and connected with combining rate distortion costs computing module;Described rate of combining is lost
True cost computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and exports
Optimum code pattern.Wherein: the image block that unit is two viewpoints that (1) described device processes;(2) the three of described device use
Dimension video side information is depth information;(3) effect of described pattern preliminary election module is the various coding moulds to texture and the degree of depth
Formula carries out primary election, in this module, uses the rate-distortion optimization algorithm of texture information independence to obtain two viewpoint current blocks optimum
NN kind mode combinations, using the combination of texture NN kind optimization model and the combination of all mode combinations of the degree of depth as treating lectotype;(4)
Described coding regards with depth information two for carrying out texture information according to currently processed coding mode combination with rebuilding module
The Video coding of some current block, and obtain reconstruction image;(5) effect of described unicode rate computing module is to calculate joint code
Rate, in this module, the computational methods of unicode rate are Runion(T, G)=R (T1)+R(T2)+R(G1)+R(G2), wherein R (T1)
For the texture information current block of viewpoint 1 to be encoded with present mode encoder bit rate, R (G1) it is the depth information of viewpoint 1 to be encoded
Sequence current block is with present mode encoder bit rate;R(T2) it is that the texture information current block of viewpoint 2 to be encoded is with present mode coding
Code check, R (G2) it is that the depth information sequence current block of viewpoint 2 to be encoded is with present mode encoder bit rate;(6) described Combine distortion
Computing module act as calculate Combine distortion, in this module, target view number is set to 1, the computational methods of Combine distortion
For Wherein: Dunion(T, G)
For Combine distortion, D1(T, G) is the distortion that target view is corresponding, TjFor the texture information current block of viewpoint j to be encoded, GjFor treating
The depth information sequence current block of coded views j, Tj' be viewpoint j to be encoded texture information current block with present mode coding,
The reconstructed block obtained after decoding, Gj' be viewpoint j to be encoded depth information sequence current block with present mode coding, decoding after
The reconstructed block obtained, W1(x, y) for utilizing texture x and depth information y to obtain the virtual regarding raw of target view respective image information
One-tenth method, Eu(x, y) for the function of the square error between pixel u of calculating x and y, c1, j, uThe mistake caused for viewpoint j to be encoded
The very weight when calculating target view corresponding blocks pixel u, its computational methods are, when pixel u virtual regarding generation method W1(x,
Y) interpolation can be need not from viewpoint j to be encoded and directly obtain in, and can not directly obtain from another viewpoint, then c1, j, u=1, if
Interpolation can not be need not from viewpoint j to be encoded and directly obtain, can directly obtain from another viewpoint, then c1, j, u=0, if being unsatisfactory for
Any of the above-described, thenWherein hjFor the parallax range between viewpoint j to be encoded and target view;(7) described associating
Rate distortion costs computing module act as calculate associating rate distortion costs, in this module, combine rate distortion costs calculating
Method is J=Dunion(T, G)+λ Runion(T, G), wherein λ is a fixing global parameter;(8) described poll optimizes module
It act as controlling device and each is treated that lectotype carries out rate-distortion optimization and selects the optimization model that associating rate distortion costs is minimum,
And the optimum mode combinations output of two viewpoint current blocks of the degree of depth will be applied to.
Example 22:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 2: by pattern preliminary election module, coding
With rebuild module, unicode rate computing module, Combine distortion computing module, combine rate distortion costs computing module, poll optimization
Module forms.Described pattern preliminary election module optimizes module with coding with reconstruction module, poll and is connected;Described coding and reconstruction module
Need texture information current coded unit and side information current coded unit as input, and with unicode rate computing module,
Combine distortion computing module is connected;Described unicode rate computing module is connected with combining rate distortion costs computing module;Described
Closing distortion computation module needs camera parameters as input and connected with combining rate distortion costs computing module;Described associating
Rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and is connected with rebuilding module with coding, and
Output optimum code mode combinations.Wherein: the unit that (1) described device processes is image block;(2) three-dimensional that described device uses
Video side information is depth information;(3) effect of described pattern preliminary election module is the various coding modes to texture and the degree of depth
Carry out primary election, in this module, use the rate-distortion optimization algorithm of texture information independence to obtain texture information present encoding block
Excellent pattern, using all for degree of depth patterns as treating lectotype;(4) described coding is used for current according to texture information with rebuilding module
The pattern of encoding block optimum carries out the Video coding of texture information current block, carries out degree of depth letter according to currently processed coding mode
The Video coding of breath current block, and obtain reconstruction image;(5) effect of described unicode rate computing module is to calculate joint code
Rate, in this module, the computational methods of unicode rate are Runion(T, G)=R (T)+R (G), wherein R (T) is that texture information is current
Block is with optimization model encoder bit rate, and R (G) is that depth information sequence current block is with present mode encoder bit rate;(6) described mistake is combined
True computing module act as calculate Combine distortion, in this module, target view number is set to 1, the calculating side of Combine distortion
Method is Wherein Dunion(T, G) is Combine distortion, D1(T, G)
For the distortion that target view is corresponding, T is texture information current block and current block left side block and the right block, and G is depth information sequence
Row current block, T ' is texture information current block with the reconstructed block obtained after present mode coding, decoding, and G ' is depth information sequence
Current block is with the reconstructed block obtained after present mode coding, decoding, and (x, y) for calculating the function of sensitive factor, the work of this function for I
With for utilizing x, y and surrounding pixel thereof are calculated the change complexity of this image-region, and image change complexity is the highest,
This value is the biggest, I (T, Tn')=(Tn′-TN, l])2+(Tn′-TN, r)2, wherein Tn' for pixel u in reconstructed block, Tn.lFor original graph
T in XiangnOne, the left side of ' respective pixel pixel, if Tn' on left picture boundary, then Tn.l=Tn, Tn.rFor in original image
TnOne, the right of ' respective pixel pixel, if Tn' on image right margin, then Tn.l=Tn, (x, y) for calculating between x and y for H
The function of difference, H (Gn, Gn')=| Gn′-Gn|;(7) rate is combined in the calculating that act as of described associating rate distortion costs computing module
Distortion cost, in this module, the computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ
It it is a fixing global parameter;(8) each is treated that lectotype carries out rate mistake by the control device that act as of described poll optimization module
Really optimize and select the optimization model that associating rate distortion costs is minimum, and by defeated for the optimum pattern that applies to degree of depth current block
Go out.
Example 23:
A kind of 3 D video rate-distortion optimization device of the present invention, as shown in Figure 3: include coding with rebuild module,
Unicode rate computing module, Combine distortion computing module, associating rate distortion costs computing module, poll optimize module.Described volume
Code needs texture information current coded unit and side information current coded unit as input and joint code with rebuilding module
Rate computing module, Combine distortion computing module are connected;Described unicode rate computing module with combine rate distortion costs computing module
It is connected;Described Combine distortion computing module need camera parameters as input, and with combine rate distortion costs computing module phase
Even;Described associating rate distortion costs computing module optimizes module with poll and is connected;Described poll optimizes module and coding and reconstruction
Module, and export optimal transformation coefficient quantization value.Wherein: the unit that (1) described device processes is image block;(2) described device
The 3 D video side information used is depth information;(3) described coding is used for according to currently processed conversion with rebuilding module
The combination of coefficient quantization value carries out the Video coding of texture information and depth information current block, and obtains reconstruction image;(4) described
The effect of the computing module closing code check is to calculate unicode rate, and in this module, the computational methods of unicode rate are Runion(T, G)
=R (T)+R (G), wherein R (T) be texture information current block be that depth information sequence is current with present mode encoder bit rate, R (G)
Block is with present mode encoder bit rate;(5) described Combine distortion computing module act as calculate Combine distortion, in this module,
Target view number is set to 1, and the computational methods of Combine distortion are Dunion(T, G)=D1(T, G)=E (W1(T, G), W1(T ',
G ')), wherein Dunion(T, G) is Combine distortion, D1(T, G) is the distortion that target view is corresponding, and T is texture information current block, G
For depth information sequence current block, T ' is that texture information current block is to obtain after current transform coefficient quantized value coding, decoding
Reconstructed block, G ' is that depth information sequence current block is with the reconstructed block obtained after current transform coefficient quantized value coding, decoding, W1
(x, y) for utilizing texture x and depth information y to obtain the virtual regarding generation method of target view respective image information, (x y) is E
Calculate the function of total square error between x and y;(6) rate is combined in the calculating that act as of described associating rate distortion costs computing module
Distortion cost, in this module, the computational methods of associating rate distortion costs are J=Dunion(T, G)+λ Runion(T, G), wherein λ
It it is a fixing global parameter;(7) described poll optimize module act as control device to each quantization of transform coefficients to be selected
Value combination carries out rate-distortion optimization and selects the optimal transformation coefficient quantization value combination that associating rate distortion costs is minimum, and will use
In texture and the optimum quantization of transform coefficients value output of degree of depth current block.
It should be pointed out that, that the application of the present invention is extensive, method is very flexible, can not enumerate at this.In every case it is
The algorithm utilizing the essence spirit of the present invention and design and the device of exploitation and system are all within the protection domain of this patent.
Claims (11)
1. a 3 D video Rate-distortion optimization method, it is characterised in that including: (1) uses texture information and side information associating
Calculate unicode rate;(2) texture information and side information combined calculation Combine distortion are used;(3) utilize unicode rate and combine
Distortion computation associating rate distortion costs;Described associating rate distortion costs is used for texture information and the combined coding of side information or use
Coding in side information;Described associating rate distortion costs refers to use connection for the combined coding of texture information and side information
Close rate distortion costs and combine the coding mode determining texture information and side information;Determine in coding mode combining, by texture
The combination of the N number of pattern in information coding mode and M pattern in side information coding mode is selected associating rate distortion generation
The mode combinations that valency is optimum;Wherein: N is integer, 2≤N≤K, K are the number of texture information coding mode;M is integer, 1 <
=M≤L, L are the number of side information coding mode.
A kind of 3 D video Rate-distortion optimization method the most according to claim 1, it is characterised in that described side information is
Depth information.
3. a 3 D video Rate-distortion optimization method, it is characterised in that including: (1) uses texture information and side information associating
Calculate unicode rate;(2) texture information and side information combined calculation Combine distortion are used;(3) utilize unicode rate and combine
Distortion computation associating rate distortion costs;Described associating rate distortion costs is used for texture information and the combined coding of side information or use
Coding in side information;Described associating rate distortion costs refers to use connection for the combined coding of texture information and side information
Close rate distortion costs and combine the optimization quantized value determining texture information and side information conversion coefficient;Conversion coefficient is determined combining
Optimization quantized value in, by NL quantized value in texture information quantization of transform coefficients value and side information quantization of transform coefficients value
In ML quantized value combination in select the quantized value combination that associating rate distortion costs is optimum;Wherein: NL is integer, 2≤NL
≤ KL, KL are the number of texture information quantization of transform coefficients value to be selected;ML is integer, and 1≤ML≤LL, LL are side information
The number of quantization of transform coefficients value to be selected.
A kind of 3 D video Rate-distortion optimization method the most according to claim 3, it is characterised in that described side information is
Depth information.
5. a 3 D video Rate-distortion optimization method, it is characterised in that including: (1) uses texture information and side information associating
Calculate unicode rate;(2) texture information and side information combined calculation Combine distortion are used;(3) utilize unicode rate and combine
Distortion computation associating rate distortion costs;Described associating rate distortion costs is used for texture information and the combined coding of side information or use
Coding in side information;Described unicode rate RunionThe computational methods of (T, G) are Runion(T, G)=aTR(T)+aGR(G);Its
In: Runion(T, G) is unicode rate;T is texture information current coded unit;G is side information current coded unit;R (T) is
Following one of both: (1) texture information current coded unit encoder bit rate;(2) texture information current coded unit encoder bit rate
Estimated value;R (G) is following one of both: (1) side information current coded unit encoder bit rate;(2) side information present encoding
In cell encoding rate estimation value;aTFor real number, for texture information current coded unit encoder bit rate weighted value;aGFor real number, for
Side information current coded unit encoder bit rate weighted value.
A kind of 3 D video Rate-distortion optimization method the most according to claim 5, it is characterised in that described side information is
Depth information.
7. a 3 D video Rate-distortion optimization method, it is characterised in that including: (1) uses texture information and side information associating
Calculate unicode rate;(2) texture information and side information combined calculation Combine distortion are used;(3) utilize unicode rate and combine
Distortion computation associating rate distortion costs;Described associating rate distortion costs is used for texture information and the combined coding of side information or use
Coding in side information;Described Combine distortion DunionThe computational methods of (T, G) areWherein:
Dunion(T, G) is Combine distortion;T is texture information current coded unit;G is side information current coded unit;Di(T, G) is
The distortion that target view i is corresponding;biFor real number, for target view i weight;P=1,2 ... for target view number;I=1,
2 ... P is target view label.
A kind of 3 D video Rate-distortion optimization method the most according to claim 7, it is characterised in that described side information is
Depth information.
A kind of 3 D video Rate-distortion optimization method the most according to claim 7, it is characterised in that described target view i
Corresponding distortion DiThe computational methods of (T, G) are one of following four method: the distortion D that (1) described target view i is correspondingi(T,
G) computational methods are Di(T, G)=E (Wi(T,G),Vi(T',G'));Wherein: Di(T, G) is the distortion of target view i;T is
Texture information current coded unit;G is side information current coded unit;T ' is that texture information current coded unit is compiled with current
The reconstruction coding unit obtained after code parameter coding, decoding;G ' is that side information current coded unit is compiled with present encoding parameter
The reconstruction coding unit obtained after code, decoding;Wi(x, y) and Vi(x, y) for utilizing texture x and side information y to obtain at viewpoint i
The function of respective image information;(x, y) for calculating the function of difference between x and y for E;(2) distortion that described target view i is corresponding
DiThe computational methods of (T, G) are Di(T, G)=F (T, T', H (G, G'));Wherein: Di(T, G) is the distortion of target view i;T is
Texture information current coded unit;G is side information current coded unit;T ' is that texture information current coded unit is compiled with current
The reconstruction coding unit obtained after code parameter coding, decoding;G ' is that side information current coded unit is compiled with present encoding parameter
The reconstruction coding unit obtained after code, decoding;(x, y) for calculating the function of difference between x and y for H;(x, y z) supplement F for utilizing
Information coding distortion z rebuilds the function of difference between texture y after calculating texture x and coding;(3) described current coded unit includes
The respective image the most described T of multiple viewpoints to be encoded is by multiple TjComposition, described G is by multiple GjComposition, described target view i
Corresponding distortion DiThe computational methods of (T, G) areWherein: Di(T, G) is target view i pair
The distortion answered;TjFor the corresponding unit to be encoded of texture information in viewpoint j to be encoded;GjCorresponding for side information in viewpoint j to be encoded
Unit to be encoded;Di,u(Tj,Gj) for utilizing viewpoint j to be encoded to generate the distortion that pixel u of corresponding target view i is corresponding;ci,j,u
For real number, the weight of viewpoint j to be encoded during for calculating the Combine distortion of pixel u of target view i;I=1,2 ... P is target
Viewpoint label;P=1,2 ... for target view number;J=1,2 ... Q is viewpoint label to be encoded;Q=1,2 ... for treating
Coded views number;U=1,2 ... V is respective pixel piont mark;V=1,2 ... for corresponding pixel points number;(4) described mesh
The distortion D that mark viewpoint i is correspondingiThe computational methods of (T, G) areWherein: Di(T, G) is target
The distortion of viewpoint i;Tt is texture information current coded unit and the most each coding unit;G is that side information is currently compiled
Code unit;T ' is that texture information current coded unit is with the reconstruction coding unit obtained after present encoding parameter coding, decoding;G’
For side information current coded unit with the reconstruction coding unit obtained after present encoding parameter coding, decoding;N is present encoding
Pixel n in unit;S is sum of all pixels in current coded unit;(x, y) for calculating the function of difference between x and y for H;I(x,y)
For calculating the function of sensitive factor.
A kind of 3 D video Rate-distortion optimization method the most according to claim 9, it is characterised in that described side information
For depth information.
11. 1 kinds of 3 D video rate-distortion optimization devices, it is characterised in that this device is by encoding and rebuild module, unicode rate meter
Calculate module, Combine distortion computing module, associating rate distortion costs computing module, poll optimization module composition;Described coding and weight
Modeling block need texture information current coded unit and side information current coded unit or one of both as input, and with connection
Close code check computing module, Combine distortion computing module is connected, and described unicode rate computing module calculates with combining rate distortion costs
Module be connected, described Combine distortion computing module need camera parameters as input, and with combine rate distortion costs calculating mould
Block is connected, and associating rate distortion costs computing module optimizes module with poll and is connected, and poll optimizes module and coding and rebuilds module
It is connected, and exports optimized parameter;Described coding is used for carrying out texture information and benefit according to the coding parameter of input with rebuilding module
Fill the coding of information or one of both, and obtain rebuilding image, described unicode rate computing module be used for calculating texture information and
Side information combined coding code check, described Combine distortion computing module combines mistake for calculate texture information and side information
Very, described associating rate distortion costs computing module is used for utilizing unicode rate to calculate associating rate distortion costs, wheel with Combine distortion
Ask optimization module to be used for judging rate distortion costs and making Optimal Decision-making, when optimal conditions not yet meets, adjusting and optimizing parameter
Setting, when optimal conditions meets, by optimum coding parameter output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110374120.4A CN103139554B (en) | 2011-11-22 | 2011-11-22 | A kind of 3 D video Rate-distortion optimization method and optimization device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110374120.4A CN103139554B (en) | 2011-11-22 | 2011-11-22 | A kind of 3 D video Rate-distortion optimization method and optimization device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103139554A CN103139554A (en) | 2013-06-05 |
CN103139554B true CN103139554B (en) | 2016-12-21 |
Family
ID=48498790
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201110374120.4A Active CN103139554B (en) | 2011-11-22 | 2011-11-22 | A kind of 3 D video Rate-distortion optimization method and optimization device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103139554B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104717515A (en) * | 2015-03-24 | 2015-06-17 | 上海大学 | Texture video and depth map code rate distributing method based on 3D-HEVC |
CN109246407B (en) * | 2017-04-28 | 2020-09-25 | 微软技术许可有限责任公司 | Image coding |
CN107862713B (en) * | 2017-09-22 | 2021-04-06 | 贵州电网有限责任公司 | Camera deflection real-time detection early warning method and module for polling meeting place |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101521819A (en) * | 2008-02-27 | 2009-09-02 | 深圳市融合视讯科技有限公司 | Method for optimizing rate distortion in video image compression |
CN101835056A (en) * | 2010-04-29 | 2010-09-15 | 西安电子科技大学 | Allocation method for optimal code rates of texture video and depth map based on models |
-
2011
- 2011-11-22 CN CN201110374120.4A patent/CN103139554B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101521819A (en) * | 2008-02-27 | 2009-09-02 | 深圳市融合视讯科技有限公司 | Method for optimizing rate distortion in video image compression |
CN101835056A (en) * | 2010-04-29 | 2010-09-15 | 西安电子科技大学 | Allocation method for optimal code rates of texture video and depth map based on models |
Non-Patent Citations (1)
Title |
---|
《Joint depth/texture bit-allocation for multi-view video compression》;Yannick Morvan等;《PICTURE CODING SYMPOSIUM》;20071107(第26期);第1-4页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103139554A (en) | 2013-06-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN100576934C (en) | Virtual visual point synthesizing method based on the degree of depth and block information | |
CN101668205B (en) | Self-adapting down-sampling stereo video compressed coding method based on residual error macro block | |
Shao et al. | Asymmetric coding of multi-view video plus depth based 3-D video for view rendering | |
CN103402097B (en) | A kind of free viewpoint video depth map encoding method and distortion prediction method thereof | |
CN102801997B (en) | Stereoscopic image compression method based on interest depth | |
CN103428499B (en) | The division methods of coding unit and the multi-view point video encoding method of use the method | |
CN104469386B (en) | A kind of perception method for encoding stereo video of the proper appreciable error model based on DOF | |
CN102685532A (en) | Coding method for free view point four-dimensional space video coding system | |
CN101309412A (en) | Multichannel video stream encoder and decoder based on deep image rendering | |
WO2016119551A1 (en) | Code distortion prediction method for depth map of two-dimensional free viewpoint video | |
CN104838648B (en) | For creating the apparatus and method with reconstructed video stream | |
CN103873867B (en) | Free viewpoint video depth map distortion prediction method and free viewpoint video depth map coding method | |
CN103248909A (en) | Method and system of converting monocular video into stereoscopic video | |
CN103139554B (en) | A kind of 3 D video Rate-distortion optimization method and optimization device | |
CN104159095B (en) | A kind of multiple views texture video and the bit rate control method of depth map encoding | |
CN107071423A (en) | Application process of the vision multi-channel model in stereoscopic video quality objective evaluation | |
CN102790895B (en) | Multi-viewpoint video encoding and viewpoint synthesis predication method based on least square | |
CN106791876B (en) | A kind of depth map fast intra-frame predicting method based on 3D-HEVC | |
CN101888561B (en) | Multi-view video transmission error control method for rate distortion optimization dynamic regulation | |
CN101521829A (en) | Method and device for processing range image sequence | |
CN106888374B (en) | A kind of 3 d video encoding method, apparatus and video processing equipment | |
CN103220532B (en) | The associated prediction coded method of three-dimensional video-frequency and system | |
Zhang et al. | Efficient rendering distortion estimation for depth map compression | |
Chung et al. | Bit allocation algorithm with novel view synthesis distortion model for multiview video plus depth coding | |
CN102892004B (en) | Multi-view point video coding code rate control method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |