CN106658006B - A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization - Google Patents
A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization Download PDFInfo
- Publication number
- CN106658006B CN106658006B CN201710020210.0A CN201710020210A CN106658006B CN 106658006 B CN106658006 B CN 106658006B CN 201710020210 A CN201710020210 A CN 201710020210A CN 106658006 B CN106658006 B CN 106658006B
- Authority
- CN
- China
- Prior art keywords
- image
- quantization parameter
- jpeg
- bit rate
- mad
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
- H04N19/149—Data rate or code amount at the encoder output by estimating the code amount by means of a model, e.g. mathematical model or statistical model
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/70—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Algebra (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
Abstract
A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization devises a kind of bit rate control method of near lossless compression for JPEG-LS standard.Encoder bit rate is pre-established under high code rate about the mathematical model between the pre- error of measurement of average absolute and coded quantization parameter NEAR.In encoding specific image process, in blocks, quantization parameter is adjusted using the model.The logic of adjusting is the closest scheduled code rate requirement of the sum of estimated bit number needed under the quantization parameter of setting of the bit number for consuming encoded part and uncoded part.The method can avoid quantization parameter from floating up and down on a large scale as far as possible, to reach the distortion performance of near-optimization.
Description
Technical field
The invention belongs to image compression encoding fields, and in particular to the code rate of the nearly non-destructive prediction coding of image controls skill
Art.
Background technique
It stays indoors, it is excellent to read the most world.Currently, information technology brings earth-shaking variation to people's lives.
And image is wherein important medium and means.Have no to rant out, image obtains the weight for having become people's daily life with transmission
Want component part.In addition, image acquisition and transmission technology also have important answer in fields such as national defence, agricultural production and disaster reduction and prevention
With.In recent years, great powers in the world all competitively research and develop, establish earth observation systems applied to fields such as military affairs, agriculturals, are being believed with expectation
The breath epoch gain the initiative.However higher and higher image resolution ratio causes image transmission to become a problem.And compression of images is
Solve the main method of transmission problem.
Existing Image Compression is divided into two class of lossless compression and lossy compression.Lossless compression can completely restore former
Beginning image, perfectness between reconstruction image and original image are very beneficial for the preservation of image information, but lossless compression
Multiplying power usually lower (between 1~2), therefore helps transmission limited.Lossy compression allows reconstruction signal to be slightly distorted, and is changed with this
Carry out higher compression multiplying power, therefore it is usually more much higher than lossless compression to compress multiplying power.It can essence by adjusting parameter appropriate
Really control compression multiplying power.However, the lossy compression of big multiplying power, will lead to picture quality reduction.The near lossless compression of image is pressure
A half-way house in demagnification rate and picture quality.Near lossless compression often can be under 2-4 times of compression multiplying power, and image loses
It is very very small.In general, it is near lossless compression, such as China and Pakistani cooperative research and development that the field for using compression of images is often mostly
Remote sensing satellite --- resource three just uses near lossless compression method.
Lossless compression mainly uses predictive coding method, and more representative is JPEG-LS and CALIC.And lossy compression
Mostly use the coding method based on transformation, such as JPEG2000 and SPIHT.Lossy coding method based on integer transform is often
It can be used for lossless compression and near lossless compression.However based on the compaction coding method of transformation in nearly lossless and lossless compression performance
Often below based on the method for predictive coding.Especially when image texture enriches, compaction coding method based on transformation (such as
JPEG2000 and SPIHT) near lossless compression performance usually it is 2dB lower than coding method JPEG-LS and CALIC based on prediction with
On.Although predictive coding is relatively more suitable for the near lossless compression of image, there is a geneogenous defect --- it is difficult to reality
Now accurate code rate control.Predictive coding method generally realizes near lossless compression by being quantified to prediction error.Generally
Based on integer prediction technique and unified integer quantization parameter is used, final encoder bit rate is often difficult to control.This is resulted in
Predictive coding cannot be applied in the occasion for much having particular/special requirement to compression bit rate.
Hou Shuwei etc. a kind of its patent " bit rate control method of JPEG-LS compression of images " (number of patent application:
201010617932.2 publication number CN102088602A) in disclose a kind of JPEG-LS bit rate control method.Wu Xianyun etc. exists
Its patent " the dynamic code rate control method based on JPEG-LS standard " applied (number of patent application: 201210434247.5, it is public
The number of opening CN102938838A) in disclose a kind of dynamic JPEG-LS bit rate control method.Open the paper that firm grade is delivered at it
" a kind of new JPEG-LS dynamic code rate control algolithm based on priori data table " (electronics and information journal, 36 (4), 823-
827,2014) proposed in it is a kind of quickening rate convergence to target bit rate method.These types of bit rate control method has one
Common trait attempts that the encoder bit rate of all regional areas of whole image is made to reach scheduled code by dynamic adjusting quantization parameter
Rate, and the distortion performance of coding is not considered.It opens and waits quietly in its patent " JPEG_LS bit rate control method under high code rate " (patent
Application number: 201210156691.5, publication number CN102695055A) disclose a kind of bit rate control method under high code rate.It should
Method although it is contemplated that coding distortion performance, however there are two defects.First is that its code rate model is excessively complicated, first calculate
The variance of relic, then the kurtosis value of residual error is calculated, so that it is determined that the generalized Gaussian distribution of residual error;Then every piece suitable is calculated again
Code rate, then an optimal quantization parameter is calculated by code rate, whole process is designed into the estimation of multiple amounts.The second is all residual
Poor block uses the same quantization parameter, and JPEG-LS is using integer quantization parameter, this necessarily causes encoder bit rate not total
Scheduled requirement can be reached.And the accuracy for rate control accuracy being placed on completely code rate model is not a real side
Case.
Many foreign scholars also proposed the bit rate control method of JPEG-LS coding.It is mainly concentrated in reduction image
The visibility of distortion.Its main principle is the visual masking effect according to human eye, image texture region abundant introduce compared with
Big noise;And smooth region introduces lesser noise.These methods can reduce image in the case where encoder bit rate is certain
It is distorted the visibility to human eye.However, the distortion of image is objective reality.In many specific application fields, image
Quality is not what human eye can determine that.One typical example is exactly Mapping remote sensing technology field, is needed the same area not
Image with visual angle is matched.This process is generally automatically performed by computer.Under normal circumstances, higher image peak value letter
The image than (or lower mean square error) of making an uproar will obtain higher matching precision.However, research discovery is excellent for human eye feature
Change, image Y-PSNR is often led to as the bit rate control method of target to human eye visibility using reduction distortion and is sharply declined
(or mean square error sharply increases).
Summary of the invention
JPEG-LS or other predictive coding methods are generally by adjusting quantization parameter NEAR (conveniently for writing, textual
Point it is abbreviated as T) adjust the code rate of coding.However, different adjusting methods may cause the image fault reconstructed after encoding and decoding
It makes a big difference.In view of the various disadvantages of above-mentioned bit rate control method, the present invention provides a kind of distortion performance near-optimization
JPEG-LS image near lossless compression bit rate control method.
The bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization, pre-establishes general
High code rate under encoder bit rate R about the mathematical model between average absolute pre- error of measurement MAD and coded quantization parameter TAnd adjust T in real time in blocks in encoding specific image process accordingly comprising the steps of:
Step 1, according to the feature of image, initial coded quantization parameter T is set*;
Step 2, image is encoded using JPEG-LS algorithm, initial coded quantization parameter T1It is set as quantization ginseng
Number T*, every to have encoded a block and just adjust first encoding quantization parameter, i.e. Ti+1=Ti+ Δ T, the method that wherein Δ T is determined is to make
The consumed bit number of encoded block estimates the sum of bit number that need to be consumed closest to target bit rate R with uncoded piecetIt is required that as follows
Formula:
Here k is integer, BN1→iThe total bit number that i block is consumed before presentation code, PNj→iIt indicates from j-th piece
To i-th piece of total number of samples,Indicate the estimated value of the pre- error of measurement of the average absolute of uncoded part, N is indicated
Total block data.
Simply, encoder bit rate R is established about the mathematical model between MAD and T with a segmented model
Specially when MAD be greater than threshold value Th,Using Log-Linear modelOtherwise general linear model is usedWherein α (T) and β (T) is the model coefficient controlled by T, threshold value Th
For the value between 0.5~2.
To avoid inaccurate estimation as far as possibleBring code rate deviation and performance loss are measured in step 2
Change used in parameter regulationEstimated by following formula:
HereIt is to acquire prediction residual, then calculated preceding i after the encoded block in the preceding face i decodes
The pre- error of measurement of the average absolute of block sampling point, MAD are the pre- error of measurements of average absolute of entire original image sampling point, and ω (i) is that a codomain is
The weight function in [0,1] section, when encoded the piece number i very little, value works as encoded the piece number close to total the piece number close to zero
When, value is close to 1.
Coding efficiency, not preparatory piecemeal before image coding, whole image predictive coding are improved to reduce prediction residual
It is continuously, only to adjust quantization parameter in blocks in an encoding process.Initial coded quantization parameter T*Setting method
Specifically: each sampling point of original image is predicted first, acquires prediction residual, the flat of image sampling point is estimated with this
Equal absolute prediction difference MAD;Further according to the pre- error of measurement MAD of resulting average absolute, optimal quantization parameter T is searched*As primary quantity
Change parameter, the difference of best i.e. target bit rate and model code rateIt is minimum.
For the Error propagation problems for avoiding the dependence between image block from may cause, another encoding scheme is image
First piecemeal, each image block are encoded using JPEG-LS algorithm independent prediction.Initial coded quantization parameter T*Estimation method then
Are as follows: first to each block, prediction residual is sought, and thus estimates the pre- error of measurement MAD of average absolute of blocki, 1≤i≤N;It searches again
Optimal quantization parameter T*As initial quantization parameters, even if the difference of target bit rate and averaging model code rateIt is minimum.The pre- error of measurement MAD of the average absolute of blockj(1≤j≤N), with fixed point
Decoding end is passed to as secondary information after number encoder;And adjust the average absolute prediction of uncoded part required for quantization parameter
The estimation of differenceIt is directly calculated by secondary information, i.e.,
Bit rate control method provided by the invention is by simply establishing under general high code rate encoder bit rate R about average
Mathematical model between absolute prediction difference MAD and coded quantization parameter TTo optimize the encoding amount of topography
The adjusting for changing parameter, finally ensure that the distortion performance of near-optimization.In addition, quantifying parameter regulation also by introducing, avoid
By inaccurate code rate model and estimated value inaccurate in an encoding processAnd cause final code rate that cannot receive
The problem of holding back target bit rate.Not only code rate model had been introduced to guarantee distortion performance, but also introduced quantization parameter adjustment mechanism to promote
It is the place that bit rate control method provided by the invention is better than the prior art into the convergence of encoder bit rate.
Detailed description of the invention
The flow diagram of Fig. 1 block not nearly Lossless Image Compression of absolute coding
The encoder bit rate of image block of Fig. 2 MAD greater than 1 and the logarithmic linear relationship display diagram of MAD
The flow diagram of the nearly Lossless Image Compression of Fig. 3 block absolute coding
Specific embodiment
Embodiment 1
Present embodiment provides the code rate controlling party of the JPEG-LS image near lossless compression of distortion performance near-optimization
Method, as shown in Figure 1.In image encoding process, quantization parameter adjustment is carried out in blocks, is joined by reasonably adjustment quantization
Numerical value keeps all pieces to have consistent quantization parameter as far as possible, and last encoder bit rate is made to reach scheduled object code
Rate.The specific implementation method of bit rate control method provided by the invention is as follows.
Firstly, calculating the prediction residual of image, and thus calculate the MAD of sampling point.According to intermediate value side used by JPEG-LS
Edge detects the prediction technique of MED, acquires the prediction residual of each sampling point, original image becomes residual image D, then as follows
(1) MAD is calculated
W, H respectively indicate the width and height of image in formula.
Secondly, being asked using code rate model according to the pre- error of measurement MAD of the resulting average absolute of step 1 and approaching object code as far as possible
Rate RtRequired quantization parameter T*。
The study found that different regions uses identical quantization parameter when carrying out predictive coding to image, on the whole may be used
To obtain almost optimal distortion performance.However, due to quantization parameter be it is discrete, not necessarily using same quantization parameter
Expected code rate can be reached.Therefore, it can only have to take the second best, obtain the rate of near-optimization using identical quantization parameter as far as possible
Distortion performance.In addition, when one timing of quantization parameter, the distortion of compression of images bring is very close, and encodes code in high code rate
There are more stable correlations by the MAD of rate and image.Include character image, natural land, satellite image and artificial for several
Composite diagram is divided into the segment of 64x64 size, and the relationship between test code rate and MAD is carried out by coded trial.Fig. 2 gives
The encoder bit rate of block of the MAD greater than 1 and the logarithmic relationship scatter diagram of MAD, wherein (a), (b), (c) and (d) respectively indicates quantization
The relational graph that parameter is 0,1,2 and 3.This shows that when MAD is greater than 1, code rate and MAD relationship are in Log-Linear relationship.However
Further tests showed that be similar to linear relationship when MAD is less than 1.5 or so.It is complicated in order to simplify calculating and model
Degree, the mathematical model established between code rate and MAD and quantization parameter T with a segmented modelHere it uses
One three piecewise function, such as following formula:
Wherein α (T) and β (T) is the model coefficient controlled by T.By regression model, acquire when quantization parameter be 0,1,2,
3,4,5 and 6 when coefficient value such as the following table 1.Although the case where only listing T≤6 meets the need of near lossless compression enough
It asks.
Table 1, the model coefficient value list as 0≤T≤6
α,β\T | 0 | 1 | 2 | 3 | 4 | 5 | 6 |
α1(T) | 2.3475 | 0.9607 | 0.5484 | 0.3438 | 0.2288 | 0.1192 | 0.0549 |
β1(T) | 1.0806 | 1.0064 | 0.8779 | 0.7888 | 0.7199 | 0.6765 | 0.6388 |
α2(T) | 0.8475 | 0.6576 | 0.5340 | 0.4874 | 0.3419 | 0.2354 | 0.1782 |
β2(T) | 1.3881 | 0.6820 | 0.6334 | 0.5317 | 0.6386 | 0.6980 | 0.7003 |
α3(T) | 0.0915 | 0.0966 | 0.0817 | 0.0689 | 0.0719 | 0.0219 | 0.0064 |
β3(T) | 2.9002 | 1.7967 | 1.5381 | 1.3687 | 1.1785 | 1.1160 | 1.0043 |
Need exist for illustrating that the coefficient value in table 1 is the regression coefficient calculated when piecemeal size is 64x64.Research
Show that other piecemeal sizes can be slightly different, but bring difference very little (less than the error of model itself).Therefore, table 1 is
Numerical value is similarly applied to the case where other piecemeal sizes.
Formula (2) establishes the mathematical model between code rate and MAD and quantization parameter TIt can be used to estimate
The code rate of image coding and optimal coded quantization parameter.By two points of iterative process, found for code search optimal
Quantization parameter T*, that is, find a several T*, make target bit rate RtAnd the difference of model code rateMinimum,
T*∈{0,1,2,3,…}。
Finally, the quantization parameter T according to estimated by step 2*, predictive coding is carried out to image, it is every to have encoded a block, it holds
Quantization parameter of row is adjusted.Quantization parameter of the quantization parameter as next block coding after adjusting, i.e. T1=T*, Ti+1=Ti
+ Δ T, the method that wherein Δ T is determined be make the consumed bit number in encoded part and the bit number that need to consume of uncoded part estimation it
With closest to target bit, such as following formula:
Here k is integer, BN1→iThe total bit number that i block is consumed before presentation code, PNj→iIt indicates from j-th piece
To i-th piece of total number of samples,Indicate the estimated value of the pre- error of measurement of the average absolute of uncoded part, N is indicated
Total block data.
Formula (3) needs to estimate the pre- error of measurement of the average absolute of uncoded partIt simply, can be direct
Use the pre- error of measurement of average absolute of encoded block portion point(it is pre- that average absolute is calculated by the image after coding and decoding
It surveys residual error and obtains, which can synchronization gain in an encoding process).But when i is smaller, estimation bring error is larger, can band
Carry out biggish performance loss.In addition estimation method is the pre- error of measurement MAD of average absolute using general image, and excludes front and compiled
The absolute prediction difference of code part can get preferably estimation, i.e.,But due toRelative to former MAD1→iIn the presence of accidentally
Difference, when i is larger, estimation bring error is larger, leads to the code rate finally encoded and expected code rate RtThere is certain deviation.
The advantages of comprehensive two kinds of estimations, the coding of front face piecemeal using second of estimation, estimated using the first by aft section block coding
Meter, such as following formula:
Here ω (i) is the weight function that a codomain is [0,1] section, and when encoded the piece number i very little, value is connect
It is bordering on zero, and when encoded the piece number is close to total the piece number, value is close to 1.Preferably ω (i) design scheme is
It completes to encode according to above three step, target bit rate RtThe pre- error of measurement MAD of average absolute with image is as secondary letter
Breath is transmitted to decoding end together, so as to the identical process in decoding end repeated encoding end.Block size is a key parameter, is on the one hand divided
Make the adjustment capability of code rate stronger at more blocks;On the other hand, bigger that there is more stable statistical property.Table 1,2 is given
The experimental result of present implementation is gone out, block is the band comprising 4096 sampling points.Since image encodes as a whole, this
Implementation method obtains optimal Y-PSNR.
Embodiment 2
Embodiment 1 provides bit rate control method and carries out quantization parameter adjusting in blocks, and whole image is as one
Binary encoding.Good distortion performance can be obtained in this way.However if there is mistake (such as transmission process in some sampling point
The mistake of middle introducing), it will cause the subsequent all sampling points of mistake sampling point that can not decode.To avoid Error propagation problems, this reality
The mode of applying provides a kind of splits' positions mode such as Fig. 3.Piecemeal is carried out to image first, then each piece is individually compiled respectively
Code.Every encoded carries out a quantization parameter after a block and adjusts, and encoder bit rate finally on the whole is made to approach scheduled target
Code rate, specific implementation method are as follows.
Firstly, image block and calculating each piece of prediction residual, every piece of MAD is thus calculated.After piecemeal, according to
The prediction technique of intermediate value edge detection MED used by JPEG-LS, acquires every piece of prediction residual, and former block becomes residual block
Bk(assuming that image is divided into N block, here 1≤k≤N), the as follows MAD of (6) calculation block
W in formulab、HbRespectively indicate the width and height of block.
Secondly, according to the resulting all pre- error of measurement MAD of average absolute of step 1k(1≤k≤N), estimation meet target bit rate Rt
It is required that optimal quantization parameter T*.The mathematical model established by formula (2)It was iteratively solved by one two points
Journey searches optimal quantization parameter T*, make target bit rate RtAnd the difference of the averaging model code rate of blockIt is minimum.
Finally, being encoded respectively to each piece, quantization parameter is carried out after coding and adjusts determining next piece of coded quantization
Parameter, i.e. Ti+1=Ti+ΔT.First piece of quantization parameter T1 is set as optimal quantization parameter T estimated by previous step*.It adjusts every time
The method such as formula (3) that section amount Δ T is determined.In addition, due to each piece of absolute coding, the pre- error of measurement MAD of every piece of average absolutek(1≤
K≤N) it needs to be transmitted to decoding end together as secondary information (each piece of MAD is with the fixed point number encoder of 10 bits).Therefore formula
(3) the pre- error of measurement of the average absolute of rest block inIt can directly calculate, i.e.,In order to guarantee the independence of every block coding, adjust every time
Amount Δ T is also transmitted to decoding end as secondary information together.Δ T belongs to a number in set { -1,0,1 } when specific implementation, therefore
Bit number needed for coded delta T can be ignored.
Block size is key parameter in the present embodiment, and the adjustment capability of code rate can be made more by being on the one hand divided into more blocks
By force, can also will likely code stream mistake be limited in smaller range.On the other hand, smaller piecemeal processing can reduce prediction and compile
The performance of code.A suitable block size can be selected as the case may be.Table 1,2 gives the piecemeal mould of present embodiment offer
The experimental result of formula (block size 64x64).Since image is separately encoded with piecemeal, the edge sampling point of block can not give a forecast, and cause
The Y-PSNR of present embodiment is lower than embodiment 1.However, with it is firm it is equal proposed in the paper that it is delivered based on elder generation
The JPEG-LS dynamic code rate control algolithm comparison of tables of data is tested, present embodiment still has a clear superiority.
The compression result of 2. target bit rate 3.0 of table compares
The compression result of 3. target bit rate 4.0 of table compares
Claims (6)
1. a kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization, passes through piecemeal tune
Section prediction residual quantization parameter controls encoder bit rate, which is characterized in that establishes under general high code rate encoder bit rate R about flat
Mathematical model between equal absolute prediction difference MAD and coded quantization parameter TEncoding specific image process
It is middle to adjust T using the model comprising the steps of:
Step 1, according to the feature of image, initial coded quantization parameter T is set*;
Step 2, image is encoded using JPEG-LS algorithm, initial coded quantization parameter T1It is set as T*, every to have encoded
First encoding quantization parameter, i.e. T are adjusted after one blocki+1=Ti+ Δ T, the method that wherein Δ T is determined is to make encoded part institute
Consumption bit number and uncoded part estimate the sum of bit number that need to be consumed closest to target bit rate RtRequirement, such as following formula:
Here k is integer, BN1→iThe total bit number that i block is consumed before presentation code, PNj→iIt indicates from j-th piece to i-th
Total number of samples of a block,Indicate the estimated value of the pre- error of measurement of the average absolute of uncoded part, N indicates total block
Number.
2. the code rate control of the JPEG-LS image near lossless compression of described distortion performance near-optimization according to claim 1
Method processed, which is characterized in that the encoder bit rate R of foundation is about the mathematical model between MAD and TIt is a segmentation
Model, specially when MAD be greater than threshold value Th,Using Log-Linear modelOtherwise general linear model is usedWherein α (T) and β (T) is the model coefficient controlled by T, threshold value Th
For the value between 0.5~2.
3. the code rate control of the JPEG-LS image near lossless compression of described distortion performance near-optimization according to claim 1
Method processed, which is characterized in that image is subjected to predictive coding as a whole, only adjusts quantization in blocks in an encoding process
Parameter, the setting method of initial coded quantization parameter T* described in step 1 specifically: each first to original image
Sampling point is predicted, prediction residual is acquired, the pre- error of measurement MAD of average absolute for acquiring image sampling point with this;Further according to resulting
MAD searches optimal quantization parameter T* as initial quantization parameters, the i.e. difference of target bit rate and model code rateIt is minimum.
4. the code rate control of the JPEG-LS image near lossless compression of described distortion performance near-optimization according to claim 1
Method processed, which is characterized in that used in step 2Estimated by following formula:
HereIt is first to acquire prediction residual, then calculated preceding i block after front i encoded blocks decode
The pre- error of measurement of the average absolute of sampling point;MAD is the pre- error of measurement of average absolute of entire original image sampling point;ω (i) be a codomain be [0,
1] weight function in section.
5. the code rate control of the JPEG-LS image near lossless compression of described distortion performance near-optimization according to claim 1
Method processed, which is characterized in that each image block independent prediction encodes, initial coded quantization parameter T described in step 1*'s
Estimation method specifically: first to each block, seek prediction residual, and the thus pre- error of measurement MAD of average absolute of calculation blocki, 1
≤i≤N;Optimal quantization parameter T is searched again*As initial quantization parameters, even if the difference of target bit rate and averaging model code rateIt is minimum.
6. the code rate control of the JPEG-LS image near lossless compression of the distortion performance near-optimization according to described in claim 5
Method processed, which is characterized in that the pre- error of measurement MAD of the average absolute of blockj, 1≤j≤N, to pinpoint after number encoder as secondary information transmitting
To decoding end;Adjust the estimation of the pre- error of measurement of average absolute of uncoded part required for quantization parameterDirectly
It connects and is calculated by secondary information, i.e.,
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710020210.0A CN106658006B (en) | 2017-01-12 | 2017-01-12 | A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710020210.0A CN106658006B (en) | 2017-01-12 | 2017-01-12 | A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106658006A CN106658006A (en) | 2017-05-10 |
CN106658006B true CN106658006B (en) | 2019-08-13 |
Family
ID=58844043
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710020210.0A Active CN106658006B (en) | 2017-01-12 | 2017-01-12 | A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106658006B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109510984B (en) * | 2018-10-26 | 2021-02-05 | 和宇健康科技股份有限公司 | Compression coding quantization method |
CN109561306B (en) * | 2018-10-26 | 2021-08-06 | 上海九吾尊易信息科技有限公司 | Bandwidth compression quantization method |
CN116438794B (en) * | 2022-05-31 | 2023-12-12 | 上海玄戒技术有限公司 | Image compression method, device, electronic equipment, chip and storage medium |
CN117676142A (en) * | 2022-09-08 | 2024-03-08 | 华为技术有限公司 | Image coding and decoding method, device, encoder, decoder and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102088602A (en) * | 2010-12-22 | 2011-06-08 | 西安空间无线电技术研究所 | Code rate control method for JPEG-LS (joint photographic experts group-lossless standard) image compression |
CN102695055A (en) * | 2012-05-18 | 2012-09-26 | 西安电子科技大学 | JPEG_LS (Joint Pho-tographic Experts Group-Lossless Standard) bit rate control method under high bit rate |
CN102938838A (en) * | 2012-10-19 | 2013-02-20 | 西安电子科技大学 | Joint Photographic Experts Group Lossless and near_lossless compression of conituous_tone still image (JPEG-LS) standard based dynamic code rate control method |
CN103686172A (en) * | 2013-12-20 | 2014-03-26 | 电子科技大学 | Code rate control method based on variable bit rate in low latency video coding |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7672523B2 (en) * | 2003-03-03 | 2010-03-02 | The Hong Kong University Of Science And Technology | Efficient rate allocation for multi-resolution coding of data |
-
2017
- 2017-01-12 CN CN201710020210.0A patent/CN106658006B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102088602A (en) * | 2010-12-22 | 2011-06-08 | 西安空间无线电技术研究所 | Code rate control method for JPEG-LS (joint photographic experts group-lossless standard) image compression |
CN102695055A (en) * | 2012-05-18 | 2012-09-26 | 西安电子科技大学 | JPEG_LS (Joint Pho-tographic Experts Group-Lossless Standard) bit rate control method under high bit rate |
CN102938838A (en) * | 2012-10-19 | 2013-02-20 | 西安电子科技大学 | Joint Photographic Experts Group Lossless and near_lossless compression of conituous_tone still image (JPEG-LS) standard based dynamic code rate control method |
CN103686172A (en) * | 2013-12-20 | 2014-03-26 | 电子科技大学 | Code rate control method based on variable bit rate in low latency video coding |
Also Published As
Publication number | Publication date |
---|---|
CN106658006A (en) | 2017-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106658006B (en) | A kind of bit rate control method of the JPEG-LS image near lossless compression of distortion performance near-optimization | |
US11297319B2 (en) | Method for encoding images and corresponding terminals | |
CN101478693A (en) | Method for evaluating star-loaded optical remote sensing image compression quality | |
CN109495744B (en) | Large-magnification remote sensing image compression method based on joint generation countermeasure network | |
CN107318014B (en) | The video quality evaluation method of view-based access control model marking area and space-time characterisation | |
CN103501438B (en) | A kind of content-adaptive method for compressing image based on principal component analysis | |
CN107040771B (en) | A kind of Encoding Optimization for panoramic video | |
CN113850747B (en) | Underwater image sharpening processing method based on light attenuation and depth estimation | |
Liu et al. | CG-Cast: Scalable wireless image SoftCast using compressive gradient | |
Tong et al. | Learning no-reference quality metric by examples | |
CN115908311B (en) | Lens forming detection equipment and method based on machine vision | |
Hou et al. | No-reference video quality evaluation by a deep transfer CNN architecture | |
CN111402327B (en) | Outdoor photo sun position estimation method based on full convolution neural network | |
CN110260914B (en) | Engineering safety monitoring system region division method based on measuring point space-time characteristics | |
CN102547363A (en) | No-reference image quality evaluating method on basis of contourlet transform domain image energy features | |
Vora et al. | Analysis of compressed image quality assessments, m | |
CN114972332B (en) | Bamboo laminated wood crack detection method based on image super-resolution reconstruction network | |
CN101938655B (en) | Compressed sensing image decoding method based on non-uniform quantizing noise model | |
CN100373952C (en) | Rapid kinematics estimation of video frequency object based on MPEG-4 | |
US20060039619A1 (en) | Image compression using a color visual model | |
CN100469143C (en) | Method for compression quantization and anti-quantization of video-frequency data | |
CN111325720B (en) | Multi-visual feature integrated reference-free super-resolution image quality evaluation method | |
CN115115573A (en) | Transform model-based rivet defect detection method and system | |
Liu et al. | Image inpainting algorithm based on KSVD and improved CDD | |
Wan et al. | Reduced reference image quality assessment based on entropy of classified primitives |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |