CN109191386A - A kind of quick Gamma bearing calibration and device based on BPNN - Google Patents
A kind of quick Gamma bearing calibration and device based on BPNN Download PDFInfo
- Publication number
- CN109191386A CN109191386A CN201810792378.8A CN201810792378A CN109191386A CN 109191386 A CN109191386 A CN 109191386A CN 201810792378 A CN201810792378 A CN 201810792378A CN 109191386 A CN109191386 A CN 109191386A
- Authority
- CN
- China
- Prior art keywords
- value
- register
- rgb
- matrix
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012549 training Methods 0.000 claims abstract description 73
- 238000012360 testing method Methods 0.000 claims abstract description 45
- 239000011159 matrix material Substances 0.000 claims description 66
- 238000000034 method Methods 0.000 claims description 24
- 238000012937 correction Methods 0.000 abstract description 24
- 210000005036 nerve Anatomy 0.000 abstract description 3
- 238000002834 transmittance Methods 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/80—Geometric correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Picture Signal Circuits (AREA)
- Processing Of Color Television Signals (AREA)
- Controls And Circuits For Display Device (AREA)
- Spectrometry And Color Measurement (AREA)
Abstract
The quick Gamma bearing calibration and device that the invention discloses a kind of based on BPNN, include the steps that training BPNN model obtain Initial Value Prediction model, using Initial Value Prediction model prediction register initial value, determine register value calibration model and RGB register corrected value determined by the model prediction；This Gamma bearing calibration and device based on data, model is trained by reverse transmittance nerve network, it is predicted based on the model trained, the data for acquiring a small number of test samples, are input in trained BPNN model and go, BPNN model determines the characteristic of test sample according to these data, the characteristic is both different from test sample, training sample can be contacted again, so that the determination of register initial value and exact value was both simple and efficient, can be made and be calculated to a nicety；Its trained model of institute has independent learning ability and predictive ability, effectively overcomes traditional algorithm and obtains the problem of initial value is difficult, approximate algorithm correction takes a long time, can rapidly correct to panel into Gamma, improve accuracy and efficiency.
Description
Technical field
The invention belongs to machine learning and technical field of image processing, more particularly, to a kind of based on the quick of BPNN
Gamma bearing calibration and device.
Background technique
Gamma correction is to be commonly utilized in one of camera work, video image and computer graphics technology；
Gamma refers to non-linear, this non-linear light usually intrinsic by device between the electric signal and optical brightness of characterization gray scale
Electrical characteristics determine.In the manufacturing, since the deviation of technique and used material are different, OLED device is driven
Gamma voltage curve is different, this brings comparable difficulty to Gamma correction.
Gamma correction refers to change Gamma value to match the middle gray of monitor；Referring to Fig.1, existing Gamma correction
Process include start correction, obtain RGB register initial value, using approximate algorithm obtain register exact value, obtain error and
Correct duration.There are mainly two types of the current methods for determining register value initial value, first is that mean value method, first mixes up several
The Gamma of panel determines suitable register value, these register values are taken it is average as initial value, if panel to be measured institute
The initial value and this mean difference needed is larger, needs to adjust another crowd of panel to obtain new initial value, in this way meeting again
Greatly increase workload；Second is that using the register value of the upper a piece of panel for mixing up Gamma as next panel's to be measured
Initial value；In this method, if the panel property difference previous panel debugged to be debugged with the latter is biggish
Words can generate large effect to subsequent adjusting.The current method for determining register exact value is approximate algorithm, approaches step-length
Selection be this method difficult point: step-length is too long, and precision does not reach requirement；Step-length is too short, inefficiency；Current algorithm is to sound out
Property go determine, it is more (existing algorithm is 6-7 time average) each to tie up number of corrections a little, causes to correct and takes a long time, and has exceeded
User is to the time-consuming requirement of correction (user requires number of corrections to be no more than 4 times).
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides one kind to be based on BPNN (Back
Propagation Neural Net, reverse transmittance nerve network) quick Gamma bearing calibration and device, its object is to
Solve the problems, such as that the prior art obtains register initial value difficulty, determines that register exact value takes a long time.
To achieve the above object, according to one aspect of the present invention, a kind of quick Gamma correction based on BPNN is provided
Method includes the following steps:
(1) using the color coordinates of mould group and brightness as input, using mould group RGB register value as output training BPNN
Model obtains Initial Value Prediction model；
(2) color coordinates and brightness value given user is obtained as the input of the Initial Value Prediction model, prediction
Obtain several groups register value；The mould group of test sample is provided as using this several groups register value, by the color of test sample
RGB register corresponding to a group in coordinate and brightness value measured value closest to setting value is posted as test sample RGB
The initial value of storage value；
(3) the first matrix is constituted according to the color coordinates of training sample and brightness, RGB register value constitutes the second matrix,
It is crossed by the difference of element and a reference value in the first matrix and constitutes third matrix, by the difference of element and a reference value in the second matrix
It was worth and constitutes the 4th matrix；And BPPN model is trained using first, third matrix as input, the second, the 4th matrix as output,
Obtain register value calibration model；
(4) using the initial value of the RGB register value as the input of the register value calibration model, by the register
It is worth corrected value of the output of calibration model as RGB register.
Preferably, the above-mentioned quick Gamma bearing calibration based on BPNN, step (1) specifically: N number of mould group is selected to make
For training sample, this N number of training sample is randomly divided into m block, every piece of training sample includes N/m training sample；
Data using every piece of training sample include mould group color coordinates and brightness as input, mould group register value
As output, BPNN model is trained and obtains m Initial Value Prediction model.
(2) register initial value is determined using Initial Value Prediction model prediction, specific as follows:
(2.1) color coordinates and brightness value given user leads to step (1) as the input of Initial Value Prediction model
The input of model generated predicts m group register value；
(2.2) this m group register value is respectively written into test sample, reads the color coordinates and brightness value of m group mould group.
(2.3) this m group color coordinates and brightness value are made comparisons with user's specified value, selects most to connect from this m group
One group of nearly user's given value, using the corresponding register value of this group as the initial value of test sample RGB register value.
(3) it determines register value calibration model and determines corrected value using the model；
(3.1) training sample data are acquired, and construct two matrixes: first be made of the color coordinates of mould group and brightness
Matrix A, the second matrix B being made of the register value of mould group；
(3.2) for the first matrix A, benchmark is regarded with matrix center row, other rows subtract this benchmark, obtain one group of face
The Δ value of chromaticity coordinates and brightness, this class value constitute third matrix Δ A；The second matrix B handle using same method
To the 4th matrix Δ B；
(3.3) register value is obtained using BPNN training pattern as output as input, (B, Δ B) with (A, Δ A)
Calibration model.
Preferably, the above-mentioned quick Gamma bearing calibration based on BPNN, step (4) includes following sub-step:
(4.1) mould group is written in the register initial value for obtaining step (2), is sat using the color that CA-310 acquires the mould group
It is marked with and brightness xylv, the mould group color coordinates and brightness is subtracted into user's specified value, obtain one group of difference DELTA xylv；
(4.2) using mould group color coordinates and brightness collection value xylv and difference DELTA xylv as register value calibration model
Input, register value rgb and register value difference DELTA rgb is predicted, by register value rgb and register value difference DELTA
Rgb is combined to obtain the corrected value (rgb+ Δ rgb) of register value；The register value of this correction is on initial value basis
On be corrected until reach within error range, obtained exact value.
To achieve the above object, order according to the invention on one side, provides a kind of quick school Gamma based on BPNN
Equipment, including Initial Value Prediction model training unit, RGB register value initial value acquiring unit, register value calibration model
Training unit and RGB register corrected value acquiring unit；
Wherein, Initial Value Prediction model training unit is used for using the color coordinates of mould group and brightness as input, with mould
Group RGB register value obtains Initial Value Prediction model as training BPNN model is exported；
The color coordinates and brightness value that RGB register value initial value acquiring unit is used to give user are as described first
The input of initial value prediction model, prediction obtain several groups register value；It is provided as testing using this several groups register value
The mould group of sample, will be corresponding to a group in the color coordinates of test sample and brightness value measured value closest to setting value
Initial value of the RGB register as test sample RGB register value；
Register value calibration model training unit be used to be constituted according to the color coordinates of training sample and brightness the first matrix,
RGB register value constitutes the second matrix, is crossed by the difference of element and a reference value in the first matrix and constitute third matrix, by second
The difference of element and a reference value in matrix, which is crossed, constitutes the 4th matrix；And using first, third matrix as input, with second, the
Four matrixes obtain register value calibration model as training BPPN model is exported；
RGB register corrected value acquiring unit is using the initial value of test sample RGB register value as the register value
The input of calibration model, the corrected value by the output of the register value calibration model as RGB register.
Preferably, above-mentioned quick Gamma means for correcting, Initial Value Prediction model training unit are instructed according to following methods
Practice and obtain Initial Value Prediction model:
It selects N number of mould group as training sample, this N number of training sample is randomly divided into m block, every piece of training sample includes
N/m training sample；
Data using every piece of training sample include mould group color coordinates and brightness as input, mould group register value
As output, BPNN model is trained and obtains m Initial Value Prediction model；Wherein, N, m, N/m are natural number.
Preferably, above-mentioned quick Gamma means for correcting, RGB register value initial value acquiring unit is according to lower section
Method determines the initial value of test sample RGB register value:
(2.1) color coordinates and brightness value given user is pre- by the Initial Value Prediction model as input
Measure m group register value；
(2.2) the m group register value is respectively written into test sample, reads color coordinates and the brightness of m group mould group
Value；
(2.3) the m group color coordinates and brightness value are made comparisons with user's specified value, is selected most from this m group
Close to one group of user's given value, using the corresponding register value of this group as the initial value of test sample RGB register value.
Preferably, above-mentioned quick Gamma means for correcting, register value calibration model training unit use following methods
Training obtains register value calibration model:
(3.1) training sample data are acquired, and construct two matrixes: first be made of the color coordinates of mould group and brightness
Matrix A, the second matrix B being made of the register value of mould group；
(3.2) for the first matrix A, benchmark is regarded with matrix center row, other rows subtract this benchmark, obtain one group of face
The difference DELTA of chromaticity coordinates and brightness and a reference value, this group of difference DELTA constitute third matrix Δ A；Using same method to the second square
Battle array B is handled to obtain the 4th matrix Δ B；
(3.3) BPPN model is trained as output as input, (B, Δ B) with (A, Δ A), obtains register value straightening die
Type.
Preferably, above-mentioned quick Gamma means for correcting, RGB register corrected value acquiring unit use following methods
Obtain the corrected value of register value:
(4.1) mould group is written into register initial value, the mould group color coordinates and brightness xylv is acquired, by the mould
Group color coordinates and brightness subtract user's specified value, obtain one group of difference DELTA xylv；
(4.2) using mould group color coordinates and brightness collection value xylv and difference DELTA xylv as register value calibration model
Input, register value rgb and register value difference DELTA rgb is predicted, by register value rgb and register value difference DELTA
Rgb is combined to obtain the corrected value (rgb+ Δ rgb) of register value
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
(1) a kind of quick Gamma bearing calibration and device based on BPNN provided by the invention is based on Back propagation neural
Network (BPNN) selects several pieces of panel as training sample, acquires the data of these training samples, and by these data into
Row study, trains model, these models can make good prediction, excessive artificial interference not needed, by reversely passing
Broadcast neural network (BPNN) quickly, efficiently, accurately solve initial value problem；It does not need constantly to change step-length, but according to mind
Go out model through network training, predict gamma curve using this model, obtain corresponding register value, such that often
A number of corrections tied up a little effectively reduces, while quickly output calibration error and the time, the good register of output correction of a final proof
Value is convenient for burning；It is corrected in this way, effectively Gamma can be carried out to panel, can successfully evade the limitation of traditional algorithm,
Human cost is greatly reduced simultaneously.
(2) a kind of quick Gamma bearing calibration and device based on BPNN provided by the invention, using neural network model
Initial value problem can be effectively solved, model has independent learning ability, this model can be given according to different test panel
Its different and suitable register initial value is given, and obtained initial value is more accurate, reduces subsequent work to the greatest extent
The workload of work actually also ensures that subsequent correction time-consuming can effectively reduce.
(3) different panel, due to the influence of technique and material etc., characteristic be will be different, this is to the school Gamma
Just causing very big difficulty；Using the quick Gamma bearing calibration provided by the invention based on BPNN and device, due to training
Model out has independent learning ability, can give the suitable register correction of one group according to different test panel
Value (this value is on initial value basis), can well solve bad shadow caused by the difference between different panel
It rings, also, actual measurement shows that required precision can be properly arrived at correction 1~2 time, it is time-consuming that Gamma correction greatly reduces.
Detailed description of the invention
Fig. 1 is the flow diagram that the prior art carries out Gamma correction；
Fig. 2 is the flow diagram of quick Gamma bearing calibration provided by the invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
One embodiment of quick Gamma bearing calibration provided by the invention based on BPNN, reference Fig. 2, including it is as follows
Step:
(1) mould group data, training Initial Value Prediction model are acquired；
In embodiment, using measurement optical correlation parameter equipment for example Konica company CA-310 acquire training sample
This color coordinates and brightness value；The value (rgb) for changing mould group register, obtains multiple groups color coordinates and brightness value
(xylv)；Data using every piece of training sample include that color coordinates and brightness value are input, register value as output, right
BPNN model, which is trained, obtains Initial Value Prediction model；
16 mould groups are arbitrarily selected in embodiment, and for example OLED mobile phone screen is as training sample, by this 16 training samples
4 blocks are randomly divided into, each piece includes 4 training samples, and symbiosis is at 4 Initial Value Prediction models.
(2) register initial value is determined using Initial Value Prediction model prediction, specific as follows:
(2.1) color coordinates and brightness value given user passes through above-mentioned 4 Initial Value Prediction models as input
Predict 4 groups of register values；
(2.2) this 4 groups of register values are respectively written into test sample, utilize CA-310 read test sample, that is, mould group face
Chromaticity coordinates and brightness value obtain four groups of color coordinates and brightness value；
(2.3) this four groups of color coordinates and brightness value are made comparisons with user's specified value, selection is given closest to user
One group of definite value, using the corresponding register value of this group as the initial value of test sample RGB register value.
It is extracted due to having done a small amount of data to test sample, and these a small amount of data contain test sample and are different from
The characteristic of other test samples, therefore the different initial value of different test samples can be predicted in this step.
(3) RGB register corrected value is determined using model prediction.
(3.1) training sample data are acquired, and construct two matrixes: first be made of the color coordinates of mould group and brightness
Matrix A, the second matrix B being made of the register value of mould group；
(3.2) for the first matrix A, benchmark is regarded with matrix center row, other rows subtract this benchmark, obtain one group of face
The Δ value of chromaticity coordinates and brightness, this class value constitute third matrix Δ A；The second matrix B handle using same method
To the 4th matrix Δ B；
(3.3) register value is obtained using BPNN training pattern as output as input, (B, Δ B) with (A, Δ A)
Calibration model；
(3.4) by step (2) obtain register initial value be written mould group, using CA-310 acquire mould group color coordinates with
And brightness (xylv), this xylv is subtracted into user's specified value, obtains one group of Δ xylv value；
(3.5) using above-mentioned xylv and Δ xylv as the input of register value calibration model, register value is predicted
(rgb) and register value difference (Δ rgb) it, is combined (rgb) and (Δ rgb) to obtain the corrected value (rgb of register value
+Δrgb)。
Two class models are trained in embodiment, one kind is to train what is obtained to be used for the initial of Initial Value Prediction in step (1)
Value prediction model, another kind of is the register for determining characteristic and carrying out register value correction that training obtains in step (3)
It is worth calibration model.Different test samples, the correcting register value predicted is different；If once predicted obtained
Corrected value cannot reach the required precision of user's proposition, then repeat step (3) and predicted again.In embodiment, two are carried out
The secondary available higher register value exact value of one group of accuracy of prediction；Here by corresponding xylv (chromaticity coordinates and
Brightness) error size determines accuracy.
The quick Gamma means for correcting based on BPNN that embodiment provides, including Initial Value Prediction model training unit,
RGB register value initial value acquiring unit, register value calibration model training unit and RGB register corrected value obtain single
Member；
Wherein, Initial Value Prediction model training unit is used for using the color coordinates of mould group and brightness as input, with mould
Group RGB register value obtains Initial Value Prediction model as training BPNN model is exported；
The color coordinates and brightness value that RGB register value initial value acquiring unit is used to give user are as described first
The input of initial value prediction model, prediction obtain several groups register value；It is provided as testing using this several groups register value
The mould group of sample, will be corresponding to a group in the color coordinates of test sample and brightness value measured value closest to setting value
Initial value of the RGB register as test sample RGB register value；
Register value calibration model training unit be used to be constituted according to the color coordinates of training sample and brightness the first matrix,
RGB register value constitutes the second matrix, is crossed by the difference of element and a reference value in the first matrix and constitute third matrix, by second
The difference of element and a reference value in matrix, which is crossed, constitutes the 4th matrix；And using first, third matrix as input, with second, the
Four matrixes obtain register value calibration model as training BPPN model is exported；
RGB register corrected value acquiring unit is using the initial value of test sample RGB register value as the register value
The input of calibration model, the corrected value by the output of the register value calibration model as RGB register.
The above-mentioned Gamma bearing calibration and system that embodiment provides are based entirely on the Gamma correction of data, by anti-
Model is trained to Propagation Neural Network, and test sample is predicted, including the determination of register initial value and exact value
It determines；Whole process is predicted based on the model trained, is acquired the data of a small number of test samples, is input to trained
It is gone in BPNN model, BPNN model determines the characteristic of test sample according to these data, which can both be different from test specimens
This, and training sample can be contacted, so that the determination of register initial value and exact value was both simple and efficient, can make accurate
Ground prediction.Its trained model of institute has independent learning ability and predictive ability, has effectively evaded traditional algorithm and has obtained just
The problem of initial value is difficult and approximate algorithm correction takes a long time, and reduce cost of labor, can rapidly to panel into
Gamma correction, improves accuracy and efficiency.
For the Gamma of mould group correction, even if the mould group of same model, inputs identical register value, utilizes survey
Color coordinates and brightness also can be variant when the equipment for measuring optical parameter reads data, how to go to make up this difference to be difficult
Point.BPNN is also known as reverse transmittance nerve network, and marrow is to calculate error using gradient descent method, and error is reversely passed
It broadcasts, modifies weight and amount of bias makes gradient and mean square deviation reach minimum as far as possible, so that model reaches pre- well
Survey effect；The present invention is predicted using the model that BPNN is trained, and the difference between different mould groups can be offset, so that not
Gamma with mould group can be corrected effectively；For the correction of OLED mould group, Gamma value is made to float 2.2 or so, with
So that the display effect of OLED mould group more meets human eye vision curve.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (10)
1. a kind of quick Gamma bearing calibration based on BPNN, which comprises the steps of:
(1) using the color coordinates of mould group and brightness as input, using mould group RGB register value as output training BPNN model
Obtain Initial Value Prediction model；
(2) color coordinates and brightness value given user is as the input of the Initial Value Prediction model, if prediction obtains
Dry group register value；It is provided as the mould group of test sample using this several groups register value, the color of test sample is sat
Be marked with and brightness value measured value in RGB register corresponding to a group closest to setting value deposited as test sample RGB
The initial value of device value；
(3) the first matrix is constituted according to the color coordinates of training sample and brightness, RGB register value constitutes the second matrix, by the
The difference of element and a reference value in one matrix, which is crossed, constitutes third matrix, by the difference mistake of the element in the second matrix and a reference value
Constitute the 4th matrix；And BPPN model is trained as input, using the second, the 4th matrix as output using first, third matrix, it obtains
Obtain register value calibration model；
(4) using the initial value of the RGB register value as the input of the register value calibration model, by the register value
Corrected value of the output of calibration model as RGB register.
2. quickly Gamma bearing calibration as described in claim 1, which is characterized in that the step (1) specifically: selection is N number of
Mould group is randomly divided into m block as training sample, by this N number of training sample, and every piece of training sample includes N/m training sample；
Data using every piece of training sample include mould group color coordinates and brightness as input, mould group register value as
Output is trained BPNN model and obtains m Initial Value Prediction model；Wherein, N, m, N/m are natural number.
3. quickly Gamma bearing calibration as claimed in claim 1 or 2, which is characterized in that the step (2) includes following son
Step:
(2.1) color coordinates and brightness value given user goes out m by the Initial Value Prediction model prediction as input
Group register value；
(2.2) the m group register value is respectively written into test sample, reads the color coordinates and brightness value of m group mould group；
(2.3) the m group color coordinates and brightness value are made comparisons with user's specified value, is selected from this m group closest
One group of user's given value, using the corresponding register value of this group as the initial value of test sample RGB register value.
4. quickly Gamma bearing calibration as claimed in claim 1 or 2, which is characterized in that the step (3) includes following son
Step:
(3.1) training sample data are acquired, and construct two matrixes: the first matrix being made of the color coordinates of mould group and brightness
A, the second matrix B being made of the register value of mould group；
(3.2) for the first matrix A, benchmark is regarded with matrix center row, other rows subtract this benchmark, obtain one group of color and sit
The difference DELTA of mark and brightness and a reference value, this group of difference DELTA constitute third matrix Δ A；Using same method to the second matrix B
It is handled to obtain the 4th matrix Δ B；
(3.3) BPPN model is trained as output as input, (B, Δ B) with (A, Δ A), obtains register value calibration model.
5. quickly Gamma bearing calibration as claimed in claim 1 or 2, which is characterized in that the step (4) includes following son
Step:
(4.1) mould group is written in the register initial value for obtaining step (2), acquires the mould group color coordinates and brightness
The mould group color coordinates and brightness are subtracted user's specified value, obtain one group of difference DELTA xylv by xylv；
(4.2) using mould group color coordinates and brightness collection value xylv and difference DELTA xylv as the defeated of register value calibration model
Enter, predict register value rgb and register value difference DELTA rgb, by register value rgb and register value difference DELTA rgb
It is combined to obtain the corrected value (rgb+ Δ rgb) of register value.
6. a kind of quick Gamma means for correcting based on BPNN, which is characterized in that including Initial Value Prediction model training unit,
RGB register value initial value acquiring unit, register value calibration model training unit and RGB register corrected value obtain single
Member；
The Initial Value Prediction model training unit is used for using the color coordinates of mould group and brightness as input, with mould group RGB
Register value obtains Initial Value Prediction model as training BPNN model is exported；
The color coordinates and brightness value that the RGB register value initial value acquiring unit is used to give user are as described first
The input of initial value prediction model, prediction obtain several groups register value；It is provided as testing using this several groups register value
The mould group of sample, will be corresponding to a group in the color coordinates of test sample and brightness value measured value closest to setting value
Initial value of the RGB register as test sample RGB register value；
The register value calibration model training unit be used to be constituted according to the color coordinates of training sample and brightness the first matrix,
RGB register value constitutes the second matrix, is crossed by the difference of element and a reference value in the first matrix and constitute third matrix, by second
The difference of element and a reference value in matrix, which is crossed, constitutes the 4th matrix；And using first, third matrix as input, with second, the
Four matrixes obtain register value calibration model as training BPPN model is exported；
The RGB register corrected value acquiring unit is using the initial value of test sample RGB register value as the register value
The input of calibration model, the corrected value by the output of the register value calibration model as RGB register.
7. the quick Gamma means for correcting based on BPNN as claimed in claim 6, which is characterized in that the Initial Value Prediction
Model training unit obtains Initial Value Prediction model according to following methods training:
It selects N number of mould group as training sample, this N number of training sample is randomly divided into m block, every piece of training sample includes N/m
A training sample；
Data using every piece of training sample include mould group color coordinates and brightness as input, mould group register value as
Output is trained BPNN model and obtains m Initial Value Prediction model；Wherein, N, m, N/m are natural number.
8. the quick Gamma means for correcting based on BPNN as claimed in claims 6 or 7, which is characterized in that the RGB deposit
Device value initial value acquiring unit determines the initial value of test sample RGB register value according to following methods:
(2.1) color coordinates and brightness value given user goes out m by the Initial Value Prediction model prediction as input
Group register value；
(2.2) the m group register value is respectively written into test sample, reads the color coordinates and brightness value of m group mould group；
(2.3) the m group color coordinates and brightness value are made comparisons with user's specified value, is selected from this m group closest
One group of user's given value, using the corresponding register value of this group as the initial value of test sample RGB register value.
9. the quick Gamma means for correcting based on BPNN as claimed in claims 6 or 7, which is characterized in that the register value
Calibration model training unit obtains register value calibration model using following methods training:
(3.1) training sample data are acquired, and construct two matrixes: the first matrix being made of the color coordinates of mould group and brightness
A, the second matrix B being made of the register value of mould group；
(3.2) for the first matrix A, benchmark is regarded with matrix center row, other rows subtract this benchmark, obtain one group of color and sit
The difference DELTA of mark and brightness and a reference value, this group of difference DELTA constitute third matrix Δ A；Using same method to the second matrix B
It is handled to obtain the 4th matrix Δ B；
(3.3) BPPN model is trained as output as input, (B, Δ B) with (A, Δ A), obtains register value calibration model.
10. the quick Gamma means for correcting based on BPNN as claimed in claims 6 or 7, which is characterized in that the RGB deposit
Device corrected value acquiring unit obtains the corrected value of register value using following methods:
(4.1) mould group is written into register initial value, the mould group color coordinates and brightness xylv is acquired, by the mould group face
Chromaticity coordinates and brightness subtract user's specified value, obtain one group of difference DELTA xylv；
(4.2) using mould group color coordinates and brightness collection value xylv and difference DELTA xylv as the defeated of register value calibration model
Enter, predict register value rgb and register value difference DELTA rgb, by register value rgb and register value difference DELTA rgb
It is combined to obtain the corrected value (rgb+ Δ rgb) of register value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810792378.8A CN109191386B (en) | 2018-07-18 | 2018-07-18 | BPNN-based rapid Gamma correction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810792378.8A CN109191386B (en) | 2018-07-18 | 2018-07-18 | BPNN-based rapid Gamma correction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109191386A true CN109191386A (en) | 2019-01-11 |
| CN109191386B CN109191386B (en) | 2020-11-06 |
Family
ID=64936253
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810792378.8A Active CN109191386B (en) | 2018-07-18 | 2018-07-18 | BPNN-based rapid Gamma correction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109191386B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109935206A (en) * | 2019-04-15 | 2019-06-25 | 京东方科技集团股份有限公司 | Display device luminance compensation method, device and equipment neural network based |
| CN110728362A (en) * | 2019-12-19 | 2020-01-24 | 武汉精立电子技术有限公司 | Module Gamma adjusting method based on LSTM neural network |
| CN113223453A (en) * | 2020-01-21 | 2021-08-06 | 武汉精立电子技术有限公司 | Module Gamma initial value prediction method |
| CN113223452A (en) * | 2020-01-21 | 2021-08-06 | 武汉精立电子技术有限公司 | Characteristic curve-based module Gamma initial value prediction method |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0934537A1 (en) * | 1996-10-27 | 1999-08-11 | Ge Medical Systems Israel, Ltd. | Gamma camera with two sequential correction maps |
| CN101907448A (en) * | 2010-07-23 | 2010-12-08 | 华南理工大学 | Depth measurement method based on binocular three-dimensional vision |
| CN102419861A (en) * | 2010-09-27 | 2012-04-18 | 上海中医药大学 | Color image correcting method based on topology subdivision of uniform color space |
| CN103208037A (en) * | 2013-04-26 | 2013-07-17 | 国电南瑞南京控制系统有限公司 | Online correction based power prediction method applicable to new energy power station |
| CN104464623A (en) * | 2014-12-05 | 2015-03-25 | 西安诺瓦电子科技有限公司 | Method and device for adjusting low gray scales of LED |
| CN106057145A (en) * | 2016-06-12 | 2016-10-26 | 北京印刷学院 | Display device colorimetric characterization model gain-offset-gamma determination method based on coordinate descent optimization |
| CN106355570A (en) * | 2016-10-21 | 2017-01-25 | 昆明理工大学 | Binocular stereoscopic vision matching method combining depth characteristics |
| CN106384573A (en) * | 2016-11-04 | 2017-02-08 | 武汉精测电子技术股份有限公司 | OLED module Gamma adjustment and calibration method based on linear interpolation calculation |
| CN106404712A (en) * | 2016-10-19 | 2017-02-15 | 重庆城市管理职业学院 | Adaptive model correcting method and system based on GT-KF-PLC near infrared spectrum |
| CN106412547A (en) * | 2016-08-29 | 2017-02-15 | 厦门美图之家科技有限公司 | Image white balance method and device based on convolutional neural network, and computing device |
| CN107016819A (en) * | 2017-06-05 | 2017-08-04 | 中国民航大学 | A kind of airfield pavement accumulated ice early warning system and its method for early warning |
| CN107346653A (en) * | 2017-06-28 | 2017-11-14 | 武汉精测电子技术股份有限公司 | A kind of GAMMA curves adjusting process and device based on deep learning |
| CN107358255A (en) * | 2017-06-28 | 2017-11-17 | 武汉精测电子技术股份有限公司 | GAMMA based on convolutional neural networks ties up a classification learning method and device |
| CN107886170A (en) * | 2017-09-30 | 2018-04-06 | 珠海格力电器股份有限公司 | Control method, device and system of cooking appliance, storage medium and processor |
| CN107966638A (en) * | 2017-12-29 | 2018-04-27 | 国网北京市电力公司 | Method and apparatus, storage medium and the processor of correction error |
| CA2987846A1 (en) * | 2016-12-07 | 2018-06-07 | Idemia Identity & Security France | Image processing system |
-
2018
- 2018-07-18 CN CN201810792378.8A patent/CN109191386B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0934537A1 (en) * | 1996-10-27 | 1999-08-11 | Ge Medical Systems Israel, Ltd. | Gamma camera with two sequential correction maps |
| CN101907448A (en) * | 2010-07-23 | 2010-12-08 | 华南理工大学 | Depth measurement method based on binocular three-dimensional vision |
| CN102419861A (en) * | 2010-09-27 | 2012-04-18 | 上海中医药大学 | Color image correcting method based on topology subdivision of uniform color space |
| CN103208037A (en) * | 2013-04-26 | 2013-07-17 | 国电南瑞南京控制系统有限公司 | Online correction based power prediction method applicable to new energy power station |
| CN104464623A (en) * | 2014-12-05 | 2015-03-25 | 西安诺瓦电子科技有限公司 | Method and device for adjusting low gray scales of LED |
| CN106057145A (en) * | 2016-06-12 | 2016-10-26 | 北京印刷学院 | Display device colorimetric characterization model gain-offset-gamma determination method based on coordinate descent optimization |
| CN106412547A (en) * | 2016-08-29 | 2017-02-15 | 厦门美图之家科技有限公司 | Image white balance method and device based on convolutional neural network, and computing device |
| CN106404712A (en) * | 2016-10-19 | 2017-02-15 | 重庆城市管理职业学院 | Adaptive model correcting method and system based on GT-KF-PLC near infrared spectrum |
| CN106355570A (en) * | 2016-10-21 | 2017-01-25 | 昆明理工大学 | Binocular stereoscopic vision matching method combining depth characteristics |
| CN106384573A (en) * | 2016-11-04 | 2017-02-08 | 武汉精测电子技术股份有限公司 | OLED module Gamma adjustment and calibration method based on linear interpolation calculation |
| CA2987846A1 (en) * | 2016-12-07 | 2018-06-07 | Idemia Identity & Security France | Image processing system |
| CN107016819A (en) * | 2017-06-05 | 2017-08-04 | 中国民航大学 | A kind of airfield pavement accumulated ice early warning system and its method for early warning |
| CN107346653A (en) * | 2017-06-28 | 2017-11-14 | 武汉精测电子技术股份有限公司 | A kind of GAMMA curves adjusting process and device based on deep learning |
| CN107358255A (en) * | 2017-06-28 | 2017-11-17 | 武汉精测电子技术股份有限公司 | GAMMA based on convolutional neural networks ties up a classification learning method and device |
| CN107886170A (en) * | 2017-09-30 | 2018-04-06 | 珠海格力电器股份有限公司 | Control method, device and system of cooking appliance, storage medium and processor |
| CN107966638A (en) * | 2017-12-29 | 2018-04-27 | 国网北京市电力公司 | Method and apparatus, storage medium and the processor of correction error |
Non-Patent Citations (3)
| Title |
|---|
| BISWAJIT BISWAS等: "Remote sensing image fusion using PCNN model parameter estimation by Gamma distribution in shearlet domain", 《PROCEDIA COMPUTER SCIENCE》 * |
| 李国芳等: "基于改进 Gamma 和改进 BP 算法的人脸识别研究", 《人工智能》 * |
| 王殿伟等: "一种光照不均匀图像的自适应校正算法", 《系统工程与电子技术》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109935206A (en) * | 2019-04-15 | 2019-06-25 | 京东方科技集团股份有限公司 | Display device luminance compensation method, device and equipment neural network based |
| CN109935206B (en) * | 2019-04-15 | 2021-08-20 | 京东方科技集团股份有限公司 | Neural network-based display device brightness compensation method, device and equipment |
| CN110728362A (en) * | 2019-12-19 | 2020-01-24 | 武汉精立电子技术有限公司 | Module Gamma adjusting method based on LSTM neural network |
| CN110728362B (en) * | 2019-12-19 | 2020-05-22 | 武汉精立电子技术有限公司 | Module Gamma adjusting method based on LSTM neural network |
| CN113223453A (en) * | 2020-01-21 | 2021-08-06 | 武汉精立电子技术有限公司 | Module Gamma initial value prediction method |
| CN113223452A (en) * | 2020-01-21 | 2021-08-06 | 武汉精立电子技术有限公司 | Characteristic curve-based module Gamma initial value prediction method |
| CN113223452B (en) * | 2020-01-21 | 2022-06-03 | 武汉精立电子技术有限公司 | Characteristic curve-based module Gamma initial value prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109191386B (en) | 2020-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109191386A (en) | A kind of quick Gamma bearing calibration and device based on BPNN | |
| CN105427788B (en) | Method and system for automatically adjusting brightness and chromaticity of display device | |
| CN108039143A (en) | A kind of method and device of gamma circuit adjustment | |
| CN109409441A (en) | Based on the coastal waters chlorophyll-a concentration remote sensing inversion method for improving random forest | |
| CN106851138B (en) | A kind of image processing method based on HDR | |
| CN110232445B (en) | Cultural relic authenticity identification method based on knowledge distillation | |
| CN104252056B (en) | The detection method and device of a kind of substrate | |
| WO2014195951A4 (en) | System and method for measurement of refractive error of an eye based on subjective distance metering | |
| JP2020077326A (en) | Photographing method and photographing device | |
| CN103559484B (en) | The method for quickly identifying of measuring instrument graduation mark | |
| CN105976767A (en) | Area source brightness uniformity adjusting method, device and system | |
| CN107229560A (en) | A kind of interface display effect testing method, image specimen page acquisition methods and device | |
| CN103763550A (en) | Method for fast measuring crosstalk of stereoscopic display | |
| CN104458597A (en) | Camera-based method, device and system for detecting product color based on | |
| Szafir et al. | Adapting color difference for design | |
| CN108986721B (en) | Detection graph generation method for display panel detection | |
| CN110044485B (en) | Image type fabric color measuring method | |
| CN108986765B (en) | Display screen white point calibration method covering visual angle color cast | |
| CN109064445B (en) | Animal quantity statistical method and system and storage medium | |
| CN106023238A (en) | Color data calibration method for camera module | |
| CN106774884B (en) | Method and device for measuring lens parameters | |
| EP4286939A1 (en) | Imaging condition setting system, imaging condition setting method, and program | |
| CN101276462B (en) | Method for processing medical image | |
| CN111932642B (en) | Method, device and equipment for measuring and calculating volume of structural crack and storage medium | |
| CN108168600A (en) | A kind of injecting products Intelligentized control method based on secondary detection |
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 |