CN104917936B - One kind is based on the relevant time domain high pass asymmetric correction method of gray scale - Google Patents
One kind is based on the relevant time domain high pass asymmetric correction method of gray scale Download PDFInfo
- Publication number
- CN104917936B CN104917936B CN201510290343.0A CN201510290343A CN104917936B CN 104917936 B CN104917936 B CN 104917936B CN 201510290343 A CN201510290343 A CN 201510290343A CN 104917936 B CN104917936 B CN 104917936B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- munderover
- correction
- frame
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Processing (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
Abstract
It is disclosed by the invention a kind of based on the relevant time domain high pass asymmetric correction method of gray scale, be related to it is a kind of for infrared imaging field based on the relevant time domain high pass asymmetric correction method of gray scale, belong to infrared imagery technique field.The present invention is worth more relevant Nonuniformity Correction model using with incident radiation, and precorrection is carried out to input picture as calibration reference source using with edge-protected spatial domain low-pass filtering result.The correction bias of each frame is calculated with reference to temporal high pass filter, the mapping relations for changing bias and gray scale according to the variable quantity of every frame same position incident radiation complete next frame correction bias, eliminate " ghost " in correction course, improve infrared imaging quality.The probability that " ghost " and " overcorrect " that the infrared imaging system Non-uniformity Correction Algorithm that the present invention can reduce real time implementation occurs occurs, improves infrared imaging quality, and calculation amount and memory space are less, facilitate hardware realization.
Description
Technical field
It is used for the present invention relates to one kind based on the relevant time domain high pass asymmetric correction method of gray scale more particularly to one kind
Infrared imaging field based on the relevant time domain high pass asymmetric correction method of gray scale, belong to infrared imagery technique field.
Background technology
Fixed pattern noise caused by infrared imaging device heterogeneity be influence its image quality key factor, it is necessary to
Nonuniformity Correction (Nonuniformity Correction) method is introduced in subsequent image procossing and eliminates noise.It is non-equal
The method of even property correction mainly has scaling correction (the Calibration Based Non- based on radiation reference source at present
Uniformity Correction, CBNUC) and adaptively correcting (Scene Based Non-uniformity based on scene
Correction,SBNUC)。
CBNUC needs to add in the homogeneous radiation reference source that radiation baffle is needed as correcting algorithm before the detectors, still
Since algorithm needs are constantly inserted into reference source in visual field, the continuous imaging of imaging system is affected to a certain extent, one
A little application fields are greatly limited.Therefore SBNUC technologies obtain very big attention in recent years, the algorithm without necessarily referring to source, according to
NUC processes can be completed by scene information.Current main SBNUC algorithms have constant statistic law, neural network, time domain height
Pass filter method, registering class correction method.Movement of these algorithms dependent on the diversity and interframe of scene, by statistical method or
Iteration can correct heterogeneity to person's registration technique to a certain extent frame by frame.It on the other hand, can be with the adaptive of hardware real time implementation
Correcting algorithm is answered to receive very big concern.But these adaptive algorithms easily generate " ghost ", the condition of convergence is not in correction course
Can rationally it cause " overcorrect ".Such as the neural network algorithm based on mean filter, details life is easily lost in filter window
Into improperly reference source, cause the appearance of " ghost ";The subsequent neural network algorithm based on bilateral filtering can be to a certain degree
The reference source for keeping details generation stable, but in the case where front and rear frame scene changes are big, the correction parameter of neutral net is difficult to
It correctly updates, causes " ghost ".
The content of the invention
For solve can real time implementation infrared imaging system Non-uniformity Correction Algorithm occur " ghost " and " overcorrect " ask
Topic, disclosed by the invention a kind of based on the relevant time domain high pass asymmetric correction method of gray scale, technical problems to be solved are
Reduce the probability that " ghost " and " overcorrect " of the infrared imaging system Non-uniformity Correction Algorithm appearance of real time implementation occurs, improve
Infrared imaging quality.
The purpose of the present invention is what is be achieved through the following technical solutions.
It is disclosed by the invention a kind of based on the relevant time domain high pass asymmetric correction method of gray scale, use and incident radiation
Be worth more relevant Nonuniformity Correction model, using with edge-protected spatial domain low-pass filtering result as calibration reference source pair
Input picture carries out precorrection, and calibration reference source is used to prevent " overcorrect " as desired value.It is calculated with reference to temporal high pass filter
The correction bias of each frame, the correction bias implementation method is, according to the variation of every frame same position incident radiation
Amount changes bias and the mapping relations of gray scale complete next frame correction bias, eliminates " ghost " in correction course, improves
Infrared imaging quality.
It is disclosed by the invention a kind of based on the relevant time domain high pass asymmetric correction method of gray scale, include the following steps:
Step 1, precorrection is carried out to input picture.Using the time domain specification of gray scale, precorrection bias matrix is carried out about
Beam.I.e. when incident radiation value changes, the heterogeneity of detector cells also changes, and corresponding bias also changes, here it is
Described gray scale is related.Therefore, present frame (k+1 frames) and previous frame gray scale difference away from it is big when, present frame bias bk+1(i,j,t)
=0;Present frame (k+1 frames) is with previous frame gray scale difference away from hour, present frame biasWherein, t tables
Show ambient temperature.
Because present frame (k+1 frames) and previous frame gray scale difference away from it is big when, present frame bk+1(i, j, t)=0, it is inclined to present frame
Put value bk+1(i, j, t) is recalculated, and eliminates previous frame to present frame bias bk+1The influence of (i, j, t), correction for reduction
The probability that " ghost " occurs in the process.
Step 2, airspace filter is carried out to the image after precorrection, estimates incident radiation value.The airspace filter is preferred
Adaptively selected property spatial domain averaging low-pass filter carries out airspace filter, and adaptively selected property spatial domain averaging low-pass filter can
High gradient information in protecting window during iteration correction, retains the marginal information of image, mitigates " overcorrect ", meanwhile,
Adaptively selected property spatial domain averaging low-pass wave energy is enough more accurately to estimate input signal, " ghost during further correction for reduction
The probability that shadow " occurs.The spatial domain estimate of incident radiation during kth frameFor:
Wherein, s × s is window size, ykIt is explorer response value, δ is the adaptively selected factor, is defined as follows:
Wherein, TspThreshold value, T are rejected for adaptive high gradientspIt is related with heteropical form of image, TspState shape
The numerical value of formula such as formula (3), wherein α and β are related with heteropical form, d1It is the spatial domain threshold value coefficient of dilution, 0.5 to 3
Between change, with serious, the d of heterogeneity phenomenon1It can increase, if heterogeneity shape is " transverse direction " striped, α=0, β=
1;If heterogeneity shape is " longitudinal direction " striped, α=1, β=0;If heterogeneity shape is " grid " property or " water
Line ", α=0.5, β=0.5.
Step 3, bias is estimated based on temporal high pass filter.The biasing estimate of kth frameFor:
For the influence of Removing Random No, bias is averaged in time domain.
Wherein,For temperature t when, detect the estimate of first (i, j) biasing.Assuming that nkThe number of (i, j) in time domain
Distribution value meets the random normal that average is 0 and is distributed, then when K is sufficiently large:
Step 4, original input picture is corrected with some calibration models, some calibration models such as formula (8),
X (i, j, t)=yk+1(i,j,t)-bk+1(i,j,t) (8)
Wherein x (i, j, t) is as output image, bk+1(i, j, t) is the bias that kth frame calculates, bk+1(i, j, t) is used
In next frame Nonuniformity Correction, 1 iteration of return to step.
Step 5, step 1 is repeated to 4 based on the relevant time domain high pass asymmetric correction method of gray scale, to each frame figure
As being corrected processing, as the time is progressive, the infrared imaging quality exported after Nonuniformity Correction gradually steps up.
Advantageous effect:
Of the invention and existing neural network nonuniformity correcting algorithm, constant statistics Non-uniformity Correction Algorithm, time domain
High pass filtering algorithm is compared, and has following advantage:
1st, the present invention is filtered using adaptively selected property spatial domain averaging low-pass filter, treated incident spoke
Estimate is penetrated for protecting strong gradient, retains image detail, while can prevent " overcorrect ", overcome neural network algorithm and when
The image degradation phenomena that domain high pass filtering algorithm occurs when handling static image;
2nd, the present invention estimates bias using temporal high pass filter device, and adjusts bias according to incident radiation value, of equal value
In adding estimation, it can effectively eliminate what corrected neural network algorithm and conventional Time-domain high pass filtering algorithm occurred
" ghost " phenomenon.
3rd, the calculation amount and memory space that the present invention needs are less, facilitate hardware realization.
Description of the drawings
Fig. 1 is the flow chart based on the relevant time domain high pass Non-uniformity Correction Algorithm of gray scale of the present invention;
Fig. 2 (a) is VOx focus planar detectors acquisition the 11st frame of original image using domestic 384 × 288;
Fig. 2 (b) is the 11st frame precorrection biasing absolute value gray-scale map;
Fig. 2 (c) is the spatial domain estimation image of the 11st frame incident radiation;
Fig. 2 (d) is the correction biasing absolute value gray-scale map that the 11st frame is calculated;
Fig. 2 (e) is the output figure after the correction of the 11st frame.
Fig. 3 (a1) gathers the 1st frame figure of original image using domestic 384 × 288 VOx focus planar detectors;
Fig. 3 (a2) gathers the 11st frame figure of original image using domestic 384 × 288 VOx focus planar detectors;
Fig. 3 (a3) gathers the 35th frame figure of original image using domestic 384 × 288 VOx focus planar detectors;
Fig. 3 (b1) is using constant statistics algorithm to the 1st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (b2) is using constant statistics algorithm to the 11st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (b3) is using constant statistics algorithm to the 35th frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (c1) is using neural network algorithm to the 1st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (c2) is using neural network algorithm to the 11st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (c3) is using neural network algorithm to the 35th frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (d1) is using temporal high-pass filtering correction to the 1st frame figure of original video sequence Nonuniformity Correction;
Fig. 3 (d2) is using temporal high-pass filtering correction to the 11st frame figure of original video sequence Nonuniformity Correction;
Fig. 3 (d3) is using temporal high-pass filtering correction to the 35th frame figure of original video sequence Nonuniformity Correction;
Fig. 3 (e1) is using inventive algorithm to the 1st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (e2) is using inventive algorithm to the 11st frame figure of result of original video sequence Nonuniformity Correction;
Fig. 3 (e3) is using inventive algorithm to the 35th frame figure of result of original video sequence Nonuniformity Correction;
It is handled using this algorithm, neutral net, constant statistics and temporal high-pass filtering correction with heteropical video
Series processing effect contrast figure, including the 1st frame, the treatment effect comparison of the 11st frame and the 35th frame.
Specific embodiment
In order to better illustrate objects and advantages of the present invention, the content of the invention is done further with example below in conjunction with the accompanying drawings
Explanation.
Embodiment 1:
The video sequence that this example is gathered using domestic 384 × 288 VOx focus planar detectors carries out contrast test and tests
The card present invention is for " striped " heteropical treatment effect of true detector.Video stream sequence shoots static for detector
Radiator constantly cuts the treatment effect and convergence rate of verification algorithm in visual field with hand.
Using disclosed in the present embodiment it is a kind of based on the relevant time domain high pass asymmetric correction method of gray scale to above-mentioned
Video stream sequence is corrected processing, and the neutral net of the infrared imaging quality that corrects that treated and prior art is non-
Even property correcting algorithm, constant statistics Non-uniformity Correction Algorithm, the infrared imaging quality of temporal high-pass filtering correction correction process
It is compared, illustrates a kind of beneficial effect based on the relevant time domain high pass asymmetric correction method of gray scale disclosed in the present embodiment
Fruit.
The method flow diagram of the present embodiment is as shown in Figure 1, a kind of high based on the relevant time domain of gray scale disclosed in the present embodiment
Logical asymmetric correction method, specific implementation step are as follows:
Step 1, precorrection is carried out to input picture.For the 11st frame original graph (such as attached drawing 2 (a)), precorrection
Bias uses what the 10th frame was calculated, and using the 11st frame and the time domain specification of the 10th frame gray scale, precorrection is biased
Value matrix carries out gray scale related constraint.I.e. the gray scale difference of the 11st frame of same position and the 10th frame away from it is little when, biasOtherwise, the 11st frame bias becomes b11(i, j, t)=0.It is related to the 11st frame bias gray scale
Biasing absolute value gray-scale map such as Fig. 2 (b) after constraint.It can be seen that region existing for the right side finger of the 11st two field picture with
10th frame grey scale change is larger, and the zero setting position of precorrection bias is more, and " ghost " occurs generally during reaching correction for reduction
The purpose of rate.
Step 2, airspace filter is carried out to the image after precorrection, estimates incident radiation value.The airspace filter is preferred
Adaptively selected property spatial domain averaging low-pass filter carries out airspace filter, and adaptively selected property spatial domain averaging low-pass filter can
High gradient information in protecting window during iteration correction, retains the marginal information of image, mitigates " overcorrect ", meanwhile,
Adaptively selected property spatial domain averaging low-pass wave energy is enough more accurately to estimate input signal, " ghost during further correction for reduction
The probability that shadow " occurs.According to formula (1), the spatial domain estimate of the incident radiation of the 11st frame is calculatedSuch as Fig. 2
(c), wherein, window size is 7 × 7, and according to formula (3), adaptive high gradient rejects threshold value TspFor 3.0881, in 7 × 7 windows
Pixel of the shade of gray more than 3.0881 is not involved in adaptively selected property spatial domain averaging low-pass ripple.
Step 3, bias is estimated based on temporal high pass filter.According to formula (7), the pre- of the 12nd frame is calculated in the 11st frame
Correction biasing estimateIts absolute value gray-scale map such as Fig. 2 (d).
Step 4, some calibration models according to formula (8) correct the 11st frame original input picture, obtain the 11st
Output image after frame correction, as shown in Fig. 2 (e).As the bias that the 11st frame calculates, return to step 1 is used
In the heterogeneity precorrection of the 12nd two field picture.
Step 5, step 1 is repeated to 4 based on the relevant time domain high pass asymmetric correction method of gray scale, to each frame figure
As being corrected processing, as the time is progressive, the infrared imaging quality exported after Nonuniformity Correction gradually steps up.
With reference to attached drawing 3 (a)-(e), the present embodiment method, neutral net, constant statistics and temporal high-pass filtering correction are used
Processing is with heteropical video sequence treatment effect comparison diagram, including the 1st frame, the treatment effect pair of the 11st frame and the 35th frame
Than, it can be clearly seen that, when occurring hand in scene and blocking radiator below, due to neutral net, constant statistics and time domain
Correction parameter does not establish correlativity with detector output gray level value in high pass filtering algorithm correction course, after 11 frames all
There is ghost in more or less, is showed in 35 frames clearly.The present embodiment method then avoids this kind of scene well
The ghost that lower Nonuniformity Correction occurs.Pass through the comparison of Fig. 3 (c2) and Fig. 3 (e2), it can be seen that the present embodiment method is subtracting
Light " overcorrect " aspect of crossing will be got well compared with neural network algorithm.
The scope of the present invention is not only limited to embodiment, the present embodiment for explaining the present invention, it is all with of the invention in phase
With the change under the conditions of principle and design or modification within protection domain disclosed by the invention.
Claims (1)
1. one kind is based on the relevant time domain high pass asymmetric correction method of gray scale, it is characterised in that:Using with incident radiation value
More relevant Nonuniformity Correction model, using with edge-protected spatial domain low-pass filtering result as calibration reference source to defeated
Enter image and carry out precorrection, calibration reference source is used to prevent " overcorrect " as desired value;It is calculated with reference to temporal high pass filter every
The correction bias of one frame, the correction bias implementation method is, according to the variable quantity of every frame same position incident radiation
The mapping relations for changing bias and gray scale complete next frame correction bias, eliminate " ghost " in correction course, improve red
Outer image quality;
The specific implementation step of the method is as follows,
Step 1, precorrection is carried out to input picture, using the time domain specification of gray scale, precorrection bias matrix is constrained, when
Previous frame (k+1 frames) and previous frame gray scale difference away from it is big when, present frame bias bk+1(i, j, t)=0;Present frame (k+1 frames) and upper one
Frame gray scale difference is away from hour, present frame biasWherein, t represents ambient temperature;
Step 2, airspace filter is carried out to the image after precorrection, estimates incident radiation value;The airspace filter is adaptive
Selective spatial domain averaging low-pass filter carries out airspace filter, and adaptively selected property spatial domain averaging low-pass filter can protect window
High gradient information in mouthful during iteration correction, retains the marginal information of image, mitigates " overcorrect ", meanwhile, adaptively
Selective spatial domain averaging low-pass wave energy is enough more accurately to estimate input signal, and " ghost " occurs during further correction for reduction
Probability;The spatial domain estimate of incident radiation during kth frameFor:
<mrow>
<msub>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>m</mi>
<mo>=</mo>
<mo>-</mo>
<mi>s</mi>
</mrow>
<mi>s</mi>
</munderover>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>n</mi>
<mo>=</mo>
<mo>-</mo>
<mi>s</mi>
</mrow>
<mi>s</mi>
</munderover>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>m</mi>
<mo>=</mo>
<mo>-</mo>
<mi>s</mi>
</mrow>
<mi>s</mi>
</munderover>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>n</mi>
<mo>=</mo>
<mo>-</mo>
<mi>s</mi>
</mrow>
<mi>s</mi>
</munderover>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, s × s is window size, ykIt is explorer response value, δ is the adaptively selected factor, is defined as follows:
<mrow>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mrow>
<mo>|</mo>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mo><</mo>
<msub>
<mi>T</mi>
<mrow>
<mi>s</mi>
<mi>p</mi>
</mrow>
</msub>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mo>|</mo>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mi>m</mi>
<mo>,</mo>
<mi>j</mi>
<mo>+</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>|</mo>
<mo>></mo>
<msub>
<mi>T</mi>
<mrow>
<mi>s</mi>
<mi>p</mi>
</mrow>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, I (i, j) represents that k takes the explorer response y of arbitrary frame numberk(i,j);TspThreshold value, T are rejected for adaptive high gradientsp
It is related with heteropical form of image, TspThe numerical value of expression form such as formula (3), wherein α and β have with heteropical form
It closes, d1It is the spatial domain threshold value coefficient of dilution, changes between 0.5 to 3, with serious, the d of heterogeneity phenomenon1It is corresponding to increase, such as
Fruit heterogeneity shape be " transverse direction " striped, α=0, β=1;If heterogeneity shape is " longitudinal direction " striped, α=1, β=0;
If heterogeneity shape is " grid " property or " water wave ", α=0.5, β=0.5;
<mrow>
<msub>
<mi>T</mi>
<mrow>
<mi>s</mi>
<mi>p</mi>
</mrow>
</msub>
<mo>=</mo>
<msub>
<mi>d</mi>
<mn>1</mn>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>&alpha;</mi>
<mfrac>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>2</mn>
</mrow>
<mi>M</mi>
</munderover>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>-</mo>
<mn>1</mn>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mrow>
<mo>(</mo>
<mrow>
<mi>M</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
<mi>N</mi>
</mrow>
</mfrac>
<mo>+</mo>
<mi>&beta;</mi>
<mfrac>
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>2</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>-</mo>
<mn>1</mn>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mi>M</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>N</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
M represents the row sum of image, and N represents the row sum of image;
Step 3, bias, the biasing estimate of kth frame are estimated based on temporal high pass filterFor:
<mrow>
<msub>
<mover>
<mi>b</mi>
<mo>^</mo>
</mover>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>n</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
For the influence of Removing Random No, bias is averaged in time domain;
<mrow>
<mover>
<mi>b</mi>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>K</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<msub>
<mover>
<mi>b</mi>
<mo>^</mo>
</mover>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, K represents the totalframes of multi-frame mean;For temperature t when, detect the estimate of first (i, j) biasing, it is assumed that nk
The numeric distribution of (i, j) in time domain meets the random normal that average is 0 and is distributed, then when K is sufficiently large:
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<msub>
<mi>n</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>0</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mover>
<mi>b</mi>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>K</mi>
</mfrac>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>y</mi>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>k</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>t</mi>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
Step 4, original input picture is corrected with some calibration models, some calibration models such as formula (8),
X (i, j, t)=yk+1(i,j,t)-bk+1(i,j,t) (8)
Wherein x (i, j, t) is as output image, bk+1(i, j, t) is the bias that kth frame calculates, bk+1(i, j, t) is used for down
One frame Nonuniformity Correction, 1 iteration of return to step,
Step 5, repeat step 1 to 4 based on the relevant time domain high pass asymmetric correction method of gray scale, to each two field picture into
Row correction process, as the time is progressive, the infrared imaging quality exported after Nonuniformity Correction gradually steps up.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510290343.0A CN104917936B (en) | 2015-05-29 | 2015-05-29 | One kind is based on the relevant time domain high pass asymmetric correction method of gray scale |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510290343.0A CN104917936B (en) | 2015-05-29 | 2015-05-29 | One kind is based on the relevant time domain high pass asymmetric correction method of gray scale |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104917936A CN104917936A (en) | 2015-09-16 |
CN104917936B true CN104917936B (en) | 2018-05-18 |
Family
ID=54086609
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510290343.0A Expired - Fee Related CN104917936B (en) | 2015-05-29 | 2015-05-29 | One kind is based on the relevant time domain high pass asymmetric correction method of gray scale |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104917936B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106017695B (en) * | 2016-07-20 | 2019-02-19 | 上海航天控制技术研究所 | Adaptive infrared asymmetric correction method based on state estimation |
CN109272520B (en) * | 2018-09-18 | 2020-11-03 | 浙江大学 | Self-adaptive infrared focal plane non-uniform correction method combining motion guidance and edge detection |
CN109360168B (en) * | 2018-10-16 | 2021-02-12 | 烟台艾睿光电科技有限公司 | Method and device for removing stripes of infrared image, infrared detector and storage medium |
CN109813442B (en) * | 2019-03-27 | 2020-05-12 | 北京理工大学 | Multi-frame processing-based internal stray radiation non-uniformity correction method |
CN109934790A (en) * | 2019-03-27 | 2019-06-25 | 北京理工大学 | Infrared imaging system asymmetric correction method with adaptive threshold |
CN110445953B (en) * | 2019-08-02 | 2020-11-06 | 浙江大华技术股份有限公司 | Method and device for reducing dynamic stripe noise, electronic equipment and storage device |
CN112769471B (en) * | 2019-11-01 | 2022-08-26 | 华为技术有限公司 | Optical fiber testing method based on optical time domain reflectometer and optical time domain reflectometer |
CN110852976B (en) * | 2019-11-22 | 2023-04-18 | 昆明物理研究所 | Infrared image brightness unevenness correction method and computer program product |
CN113240664B (en) * | 2021-06-03 | 2023-06-09 | 郑州航空工业管理学院 | Infrared detection false alarm detection method based on scene space-time significance and application thereof |
CN117372285B (en) * | 2023-12-05 | 2024-02-20 | 成都市晶林科技有限公司 | Time domain high-pass filtering method and system for static and dynamic region distinction |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289788A (en) * | 2011-06-17 | 2011-12-21 | 中国电子科技集团公司第二十八研究所 | Strip non-uniformity real-time correction method in multi-channel infrared detector |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104813652B (en) * | 2012-09-18 | 2018-03-09 | 菲力尔系统公司 | The noise reduction of pixel in thermal image |
-
2015
- 2015-05-29 CN CN201510290343.0A patent/CN104917936B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102289788A (en) * | 2011-06-17 | 2011-12-21 | 中国电子科技集团公司第二十八研究所 | Strip non-uniformity real-time correction method in multi-channel infrared detector |
Non-Patent Citations (1)
Title |
---|
Space Low-Pass and Temporal High-Pass Nonuniformity Correction Algorithm;Weixian Qian;《Optical Review》;20101231;第24-29页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104917936A (en) | 2015-09-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104917936B (en) | One kind is based on the relevant time domain high pass asymmetric correction method of gray scale | |
CN102778296B (en) | Total variation-based self-adaptation non-uniformity correction method for infrared focal plane | |
CN103279931B (en) | Mist elimination image denoising method based on absorbance | |
CN106169181B (en) | A kind of image processing method and system | |
CN102968765B (en) | Method for correcting infrared focal plane heterogeneity based on sigma filter | |
CN105205794B (en) | A kind of synchronous enhancing denoising method of low-light (level) image | |
US11087439B2 (en) | Hybrid framework-based image bit-depth expansion method and device | |
CN107133923B (en) | Fuzzy image non-blind deblurring method based on adaptive gradient sparse model | |
CN102968776A (en) | Linear filter and non-linear filter combined heterogeneity correction method | |
CN107886486A (en) | Based on dark channel prior and variation Retinex underwater picture Enhancement Methods | |
CN109934790A (en) | Infrared imaging system asymmetric correction method with adaptive threshold | |
CN110445953B (en) | Method and device for reducing dynamic stripe noise, electronic equipment and storage device | |
CN106342194B (en) | A kind of Infrared Image Non-uniformity Correction method of ground scene | |
CN105046658A (en) | Low-illumination image processing method and device | |
CN104881845A (en) | Method And Apparatus For Processing Image | |
CN104867122A (en) | Infrared self-adaptive non-uniformity correction and detail enhanced cascade processing method | |
CN101527038A (en) | Improved method for enhancing picture contrast based on histogram | |
Indu et al. | A noise fading technique for images highly corrupted with impulse noise | |
CN105931203A (en) | Infrared image stripe filtering method based on statistical relative stripe removal method | |
CN102750679B (en) | Blind deblurring method for image quality evaluation | |
CN113781367B (en) | Noise reduction method after low-illumination image histogram equalization | |
CN109360167B (en) | Infrared image correction method and device and storage medium | |
CN105654431A (en) | Deblurring method for image with presence of shielding | |
CN105005967A (en) | Method and apparatus for correcting non-uniformity of infrared imaging based on combined space-time filtering | |
Veerakumar et al. | Salt and pepper noise removal in video using adaptive decision based median filter |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180518 |
|
CF01 | Termination of patent right due to non-payment of annual fee |