CN109658342A - The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system - Google Patents
The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system Download PDFInfo
- Publication number
- CN109658342A CN109658342A CN201811280676.5A CN201811280676A CN109658342A CN 109658342 A CN109658342 A CN 109658342A CN 201811280676 A CN201811280676 A CN 201811280676A CN 109658342 A CN109658342 A CN 109658342A
- Authority
- CN
- China
- Prior art keywords
- component
- norm
- remote sensing
- sensing image
- variation
- 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.)
- Pending
Links
- 238000007323 disproportionation reaction Methods 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 97
- 238000005286 illumination Methods 0.000 claims abstract description 72
- 238000012937 correction Methods 0.000 claims abstract description 41
- 238000012545 processing Methods 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 3
- 238000002474 experimental method Methods 0.000 description 24
- 238000003708 edge detection Methods 0.000 description 11
- 230000006872 improvement Effects 0.000 description 9
- 238000001914 filtration Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 5
- 101100117236 Drosophila melanogaster speck gene Proteins 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 230000000149 penetrating effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000003595 mist Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
- 239000003643 water by type Substances 0.000 description 1
- 230000003313 weakening effect 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
- G06T5/77—Retouching; Inpainting; Scratch removal
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system, method include the following steps: remote sensing image being decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;Using L2Norm constrains illumination component, using L1Norm constrains reflecting component, obtains the Variation Model of double norm mixed constraints;The Variation Model of double norm mixed constraints is solved, the illumination component and reflecting component after being corrected, and according to after correction illumination component and reflecting component corrected after remote sensing image.Technical solution provided by the invention utilizes norm L1The texture and detailed information for constraining reflected image, utilize L2The slickness of norm constraint ray images can eliminate brightness irregularities phenomenon well compared with prior art and keep the texture and detailed information of image, improve the effect corrected to remote sensing image brightness disproportionation variation.
Description
Technical field
The invention belongs to Remote Sensing Image Processing Technology fields, and in particular to a kind of remote sensing image of double norm mixed constraints is bright
Spend uneven variation bearing calibration and system.
Background technique
When obtaining remote sensing images, due to the influence of the external factor such as sensor oneself factor and illumination, cloud and mist, obtain
Optical image can have difference in varying degrees in brightness, and the quality of image be caused to decline, so affect image into one
Step handles (such as target interpretation, the even color of image, image mosaic) and applies.Therefore the correction of research remote sensing image brightness disproportionation is asked
Topic eliminates the luminance difference inside image, has important practical significance and application value.
The method of currently used image brilliance unevenness correction can be divided into three classes: statistical information method, mathematical model method and frequency
Rate domain filter method.
Statistical information method is corrected using the statistical information of image, and this method commonly utilizes histogram equalization
Method carries out the bearing calibration of brightness disproportionation correction and opportunity mean variance.But histogram equalization method carries out brightness disproportionation correction
It calculates fairly simple, and does not account for pixel space distribution in calculating process, versatility is strong;The school of opportunity mean variance
Correction method can solve the inconsistent problem of the uniform variance of brightness disproportionation, but be easy colour cast, change atural object property.
According to local luminance mean value, using the Luminance Distribution inside Model fitting image, then mathematical model method is
To the method that different piece is compensated in various degree, the most commonly used is adaptive template methods for this method.Due to causing image bright
Spend that external factor unevenly distributed is more complicated, atural object itself distribution is also irregular, therefore uses mathematical model side
When method carries out gamma correction, it is difficult to choose most suitable model automatically, and some irregular luminance areas in image
It will lead to the brightness change that can not accurately be fitted image.
Frequency filtering method is method of greatest concern, such as the common Mask method based on additive model, passes through Gauss
Low-pass filtering method obtains ray images, then subtracts it from raw video, achievees the purpose that gamma correction.This method letter
It is single practical, the concern of more scholar is obtained, and derived many improved methods.
Incident components are separated using homomorphic filtering method with reflecting component such as, enhance high frequency while weakening low-frequency component
Ingredient eliminates inhomogeneous illumination purpose;Or the Retinex bearing calibration based on human eye visual perception theory, image is passed through
After logarithmic transformation, recycles low-pass filtering to obtain illumination component, color distortion can be avoided while correcting luminance unevenness.?
On the basis of this, and single scale Retinex and multi-Scale Retinex Algorithm are developed.Although frequency filtering method is for yin
The special areas such as shadow, speck are insensitive, but raw video and the difference of background video obtained by the way of low-pass filtering or
Ratio operation is likely to result in more information loss, easily causes the degeneration of the quality of image, while the design and ginseng of filter
Several selections need certain experience and skill.
With the development of Theory of Variational Principles, there is the Imaging enhanced method based on variation Retinex, this method is also introduced into
Into remote sensing image processing, it converts ray images estimation problem to the Optimal solution problem of Variation Model, is remote sensing image brightness
Inhomogeneity correction problem provides a kind of new approaches.But only constrained using a kind of norm, it cannot obtain preferably correcting knot
Fruit.
Summary of the invention
The purpose of the present invention is to provide a kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints,
For carrying out uneven variation correction to remote sensing image brightness, solve carrying out uneven variation to remote sensing image brightness in the prior art
The bad problem of timing calibration result;Correspondingly, the present invention provides a kind of remote sensing image brightness of double norm mixed constraints
Uneven variation corrects system.
To achieve the above object, present invention provide the technical scheme that
A kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints, includes the following steps:
(1) remote sensing image is decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;
(2) L is used2Norm constrains illumination component, using L1Norm constrains reflecting component, obtains double models
The Variation Model of number mixed constraints;
(3) Variation Model of double norm mixed constraints is solved, the illumination component and reflection after being corrected
Component, and according to after correction illumination component and reflecting component corrected after remote sensing image.
Technical solution provided by the present invention, utilizes L1The texture and detailed information of norm constraint reflected image, utilize L2Model
The slickness of number constraint ray images can eliminate brightness irregularities phenomenon well compared with prior art and keep image
Texture and detailed information improve the effect corrected to remote sensing image brightness disproportionation variation, while having higher operation efficiency.
As the further improvement that the Variation Model to double norm mixed constraints is solved, to double norm mixed constraints
The step of Variation Model is solved are as follows:
Auxiliary quantity is introduced using division Bregman method, makes L1Norm and L2The bound term of norm is not directly relevant to;
Penalty term is added, converts unconstrained problem for constrained optimization problem;
By carrying out optimization processing to illumination component, auxiliary component and reflecting component, the illumination component after being corrected
And reflecting component.
As the further improvement for optimizing processing to illumination component, auxiliary component and transmitting component, using alternating iteration
Method carries out optimization processing to the illumination component, auxiliary component and reflecting component;When the residual error of illumination component is less than limit difference
When, corresponding illumination component and auxiliary component are optimal illumination component and auxiliary component.
As further limiting for the Variation Model to double norm mixed constraints, the variation mould of double norm mixed constraints
Type is
Wherein, l is illumination component, and r is reflecting component,For L1Norm, λ1、λ2For non-negative parameter,To protect
True item constrains the similarity between l+r and s;It is to constrain the texture and minutia of reflected image, passes through ginseng
Number λ1Carry out weight adjustment;It is to constrain the spatial smoothness of ray images, passes through parameter lambda2Carry out weight tune
It is whole.
As the further improvement to illumination component and reflecting component solution, remote sensing image is decomposed into ray images and anti-
After projection picture, by taking Logarithmic calculation to obtain corresponding illumination component and reflecting component;Illumination component after being corrected and anti-
After penetrating component, the ray images and reflected image after correction are calculated by corresponding fetching number, and according to the light after correction
Remote sensing image after being corrected according to image and reflected image.
A kind of remote sensing image brightness disproportionation variation correction system of double norm mixed constraints, including processor and memory,
The computer program for executing on a processor is stored on the memory;The processor executes the computer program
Shi Shixian following steps:
(1) remote sensing image is decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;
(2) L is used2Norm constrains illumination component, using L1Norm constrains reflecting component, obtains double models
The Variation Model of number mixed constraints;
(3) Variation Model of double norm mixed constraints is solved, the illumination component and reflection after being corrected
Component, and according to after correction illumination component and reflecting component corrected after remote sensing image.
As the further improvement that the Variation Model to double norm mixed constraints is solved, to double norm mixed constraints
The step of Variation Model is solved are as follows:
Auxiliary quantity is introduced using division Bregman method, makes L1Norm and L2The bound term of norm is not directly relevant to;
Penalty term is added, converts unconstrained problem for constrained optimization problem;
By carrying out optimization processing to illumination component, auxiliary component and reflecting component, the illumination component after being corrected
And reflecting component.
As the further improvement for optimizing processing to illumination component, auxiliary component and transmitting component, using alternating iteration
Method carries out optimization processing to the illumination component, auxiliary component and reflecting component;When the residual error of illumination component is less than limit difference
When, corresponding illumination component and auxiliary component are optimal illumination component and auxiliary component.
As further limiting for the Variation Model to double norm mixed constraints, the variation mould of double norm mixed constraints
Type is
Wherein, l is illumination component, and r is reflecting component,For L1Norm, λ1、λ2For non-negative parameter,For fidelity
, constrain the similarity between l+r and s;It is to constrain the texture and minutia of reflected image, passes through parameter
λ1Carry out weight adjustment;It is to constrain the spatial smoothness of ray images, passes through parameter lambda2Carry out weight adjustment.
As the further improvement to illumination component and reflecting component solution, remote sensing image is decomposed into ray images and anti-
After projection picture, by taking Logarithmic calculation to obtain corresponding illumination component and reflecting component;Illumination component after being corrected and anti-
After penetrating component, the ray images and reflected image after correction are calculated by corresponding fetching number, and according to the light after correction
Remote sensing image after being corrected according to image and reflected image.
Detailed description of the invention
Fig. 1 is the remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints in embodiment of the present invention method
Flow chart;
Fig. 2 a is the first original image in embodiment of the present invention method experiment 1;
Fig. 2 b is the result figure that first original image uses the correction of VFR method in embodiment of the present invention method experiment 1;
Fig. 2 c provides bearing calibration using the present embodiment by the first original image in embodiment of the present invention method experiment 1
Result figure;
Fig. 3 a is the second original image in embodiment of the present invention method experiment 1;
Fig. 3 b is the result figure that second original image uses the correction of VFR method in embodiment of the present invention method experiment 1;
Fig. 3 c provides bearing calibration using the present embodiment by the second original image in embodiment of the present invention method experiment 1
Result figure;
Fig. 4 a is the third original image in embodiment of the present invention method experiment 1;
Fig. 4 b is the result figure that third original image uses the correction of VFR method in embodiment of the present invention method experiment 1;
Fig. 4 c provides bearing calibration using the present embodiment by third original image in embodiment of the present invention method experiment 1
Result figure;
Fig. 5 a is that the correcting image of the first original image in embodiment of the present invention method experiment 1 is distributed line chart;
Fig. 5 b is that the correcting image of the second original image in embodiment of the present invention method experiment 1 is distributed line chart;
Fig. 5 c is that the correcting image of third original image in embodiment of the present invention method experiment 1 is distributed line chart;
Fig. 6 a is the edge detection that first original image uses after the correction of VFR method in embodiment of the present invention method experiment 2
As a result;
Fig. 6 b is the edge detection that second original image uses after the correction of VFR method in embodiment of the present invention method experiment 2
As a result;
Fig. 6 c is the edge detection that third original image uses after the correction of VFR method in embodiment of the present invention method experiment 2
As a result;
Fig. 7 a provides bearing calibration school using the present embodiment by the first original image in embodiment of the present invention method experiment 2
Edge detection results after just;
Fig. 7 b provides bearing calibration school using the present embodiment by the second original image in embodiment of the present invention method experiment 2
Edge detection results after just;
Fig. 7 c provides bearing calibration school using the present embodiment by third original image in embodiment of the present invention method experiment 2
Edge detection results after just.
Specific embodiment
The purpose of the present invention is to provide a kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints,
For carrying out uneven variation correction to remote sensing image brightness, solve carrying out uneven variation to remote sensing image brightness in the prior art
The bad problem of timing calibration result;Correspondingly, the present invention provides a kind of remote sensing image brightness of double norm mixed constraints
Uneven variation corrects system.
To achieve the above object, present invention provide the technical scheme that
A kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints, includes the following steps:
(1) remote sensing image is decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;
(2) L is used2Norm constrains illumination component, using L1Norm constrains reflecting component, obtains double models
The Variation Model of number mixed constraints;
(3) Variation Model of double norm mixed constraints is solved, the illumination component and reflection after being corrected
Component, and according to after correction illumination component and reflecting component corrected after remote sensing image.
Embodiments of the present invention are described further with reference to the accompanying drawing.
Embodiment of the method:
The present embodiment provides a kind of remote sensing image brightness disproportionation variation bearing calibrations of double norm mixed constraints, for distant
Feel image brilliance and carry out uneven variation correction, solves carrying out uneven variation timing school to remote sensing image brightness in the prior art
The bad problem of plus effect.
The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints provided by the present embodiment, process is such as
Shown in Fig. 1, remote sensing image is HSI colour model by rgb color model conversion by the present embodiment, only to the channel brightness I therein
Carrying out processing, specific step is as follows:
(1) according to Retinex theory, remote sensing image is decomposed into ray images and reflected image, specifically:
If raw video is S (x, y), ray images are L (x, y), and reflected image is R (x, y), then have
S (x, y)=L (x, y) × R (x, y);
It is calculated to simplify, above-mentioned formula is taken into logarithm, converts plus and minus calculation for product calculation:
If s=log (S), l=log (L), r=log (R) then have
S (x, y)=l (x, y)+r (x, y);
Ray images reflect illumination condition when obtaining remote sensing image, and reflected image then corresponds to the attribute of atural object itself,
Ray images are first estimated, then it is removed from raw video, obtain the original original appearance of atural object, brightness disproportionation can eliminated
While keep image in atural object color information, achieve the purpose that gamma correction;
(2) L based on illumination component is established according to Retinex algorithm2The Variation Model of norm constraint, specifically:
The Variation Model obtained first according to Retinex algorithm are as follows:
Then L is used2Norm constrains it, is represented by
WhereinIndicate L2Norm,For image gradient, α, β are non-negative parameter;
Reflecting component is added in Variation Model, and uses L1Norm constrains Variation Model, obtains double norms
The Variation Model of mixed constraints:
WhereinFor L1Norm, λ1、λ2For non-negative parameter;For fidelity term, constrain similar between l+r and s;It is to constrain the texture and minutia of reflected image, passes through parameter lambda1Carry out weight adjustment;It is to illumination
The spatial smoothness of image is constrained, and parameter lambda is passed through2Carry out weight adjustment;
(3) Variation Model is solved;
Auxiliary variable is introduced using division Bregman method, makes L1Norm and L2The bound term of norm is not directly relevant to, and is obtained
It arrives
Wherein
Four secondary penalty items are added, convert unconstrained problem for constrained optimization problem
Wherein
Wherein i is the line number of location of pixels in image, and j is the row number of location of pixels in image.
In order to simplify above-mentioned minimization problem, the present embodiment uses alternative iteration method, the solution of Variation Model is decomposed into 3
A subproblem:
Volume reflection optimizes subproblem:
To its derivation, can obtain
It arranges, can obtain
WhereinIndicate the gradient in the direction x,Indicate the gradient in the direction y, I indicates that unit matrix, Δ indicate Laplce
Operator,
Illumination component optimizes subproblem:
To its derivation, can obtain
It arranges, can obtain
Auxiliary variable optimizes subproblem:
The approximate solution of the two formula is calculated by iterative method:
Wherein
Wherein k is the number of iterations.
Work as residual error | | lk+1-lk||/||lk| | stop iteration when less than limit difference ε, obtains optimal l and r, then refer to the two
Ray images and reflected image can be obtained in transformation of variables, and then the remote sensing images after being corrected.
Experiment porch is the computer of Intel Core i5CPU, 8G memory, and programming language is Matlab programming language.?
This in verification process needs to be arranged four parameters, λ1Influence less, there is stability to result;λ2Smaller, problem gets over positive definite;γ1
Smaller, convergence rate is faster, γ2Bigger, ray images are more smooth.It is found through experiments that following parameter can take different data
Obtain better result: λ1=0.1, λ2=0.0001, γ1=0.0002, γ2=200, guaranteeing to receive while correcting outcome quality
Fast speed is held back, therefore is all made of this group of parameter during the experiment.The parameter of VFR method is set as α=0.00001, and β=
0.1, all experiments are all made of same group of parameter.
Test 1: brightness uniformity comparison
Test data is the non-uniform remote sensing image of Luminance Distribution of 3 1024 × 1024 sizes, such as Fig. 2 a, Fig. 3 a and figure
Shown in 4a.Fig. 2 a (image I) and Fig. 3 a (image II) is respectively the excessively dark gray scale rural area image and city chromatic image in part, figure
4a (image III) is the islands and reefs image of local overexposure.VFR method and context of methods is respectively adopted to handle 3 width images, schemes
2b, Fig. 3 b, the test result that Fig. 4 b is VFR method, Fig. 2 c, Fig. 3 c and Fig. 4 c provide the test of bearing calibration by the present embodiment
As a result.
In terms of improvement of visual effect, VFR method is more satisfactory to the correction result of image I and image II, and global illumination is more equal
It is even.But there are the images of waters speck, as shown in Figure 4 b, there is local speck in the upper right corner.In comparison, this implementation
The obtained 3 width correction result Luminance Distribution of the provided bearing calibration of example is all relatively uniform, and improvement of visual effect is preferable.
In order to further compare the uniformity of two methods correction result image brilliance, correcting image is divided into the 16 of 4 × 4
A image blocks calculate the luminance mean value of each image blocks and draw Luminance Distribution according to serial number from top to bottom, from left to right
Line chart, as shown in Figure 5.As can be seen that and the brightness distribution curve of the correction result of the provided bearing calibration of the present embodiment more
Add smoothly, and the luminance mean value dispersion degree of each image blocks of VFR method is relatively large, illustrates correction side provided by the present embodiment
Method has apparent advantage compared with VFR method on brightness uniformity.
Test 2: edge detection comparison
The superiority and inferiority of image texture and details ability is kept in order to compare the provided bearing calibration of the present embodiment and VFR method
Property, this experiment is using Canny operator to 6 width correcting images progress edge detection, and wherein Fig. 6 a, Fig. 6 b and Fig. 6 c are respectively VFR
The edge detection results of method correcting image, Fig. 7 a, Fig. 7 b and Fig. 7 c are respectively the edge detection knot of VFR method correcting image
Fruit.
Comparison diagram 6a, Fig. 6 b, Fig. 6 c and Fig. 7 a, Fig. 7 b, Fig. 7 c are it is found that the provided bearing calibration correcting image of the present embodiment
Edge detection results be better than VFR method, road extraction edge is than more complete, as shown in Fig. 7 (a) and Fig. 7 (b).And VFR method
The road edge detected in correcting image is imperfect, is broken as shown in Fig. 6 (a) or even among road, such as Fig. 6 (b) institute
Show.Comparison diagram 6 (c) and Fig. 7 (c) are as can be seen that the islands and reefs edge of the Image detection of VFR method processing is also not so good as embodiment and is mentioned
The bearing calibration of confession is accurate.This is because VFR method uses only L2Norm constrains the slickness of ray images, and this
Bearing calibration provided by embodiment has comprehensively considered the heterogeneity of ray images and reflected image, using L1Norm and L2Model
Number mixed constraints, therefore image texture and details ability is kept to be better than VFR method.
Test 3: quantitative assessing index comparison
In order to the provided bearing calibration of quantitative assessment the present embodiment correction as a result, count respectively the entropy of each correcting image,
The value of average gradient and standard deviation, three kinds of video quality evaluation parameters is bigger, illustrates that the texture of image and detailed information are abundanter.
Secondly, counting the standard deviation of each image blocks luminance mean value, image blocks standard deviation is bigger, illustrates each piece of luminance mean value dispersion degree
Bigger, the distribution of image internal brightness is more uneven.When the finally processing of statistics VFR method and the provided bearing calibration of the present embodiment
Between, statistical result is as shown in table 1.
As shown in Table 1, entropy, average gradient and the standard deviation of the provided bearing calibration correcting image of the present embodiment are superior to
VFR method, average gradient and standard deviation improve 15% or more, illustrate the line of the provided bearing calibration correcting image of the present embodiment
Reason and details are richer.In addition, the piecemeal average value standard deviation of the provided bearing calibration correcting image of the present embodiment be decreased to 3 with
Interior, about the 1/4 of VFR methods and results illustrates the provided bearing calibration method correcting image internal brightness distribution of the present embodiment more
Uniformly, consistent with improvement of visual effect, demonstrate the validity of double norm constraints.In terms of the speed of service, the provided school of the present embodiment
Correction method and VFR method are the iterative solution based on Variation Model, and VFR method is solved using steepest descent method, and this implementation
The there is provided bearing calibration of example is optimized using Bregman iterative method is divided, with faster speed, efficiency improve 7 times with
On.
Table 1
Have known to above-mentioned experiment: the present embodiment is unevenly distributed problem for remote sensing image internal brightness, proposes a kind of double
The remote sensing image brightness disproportionation variation bearing calibration of norm mixed constraints, is had more compared to using the VFR method of single norm constraint
Advantage can preferably eliminate brightness irregularities phenomenon and keep the texture and detailed information of image;It is used most with VFR method
Fast descent method is compared, and division Bregman method used by the present embodiment is more efficient, and solving speed improves 7 times or more.
System embodiment:
The present embodiment provides a kind of remote sensing image brightness disproportionation variations of double norm mixed constraints to correct system, including storage
Device and processor are stored with the computer program for executing on a processor on memory, when processor executes the computer
When program, the remote sensing image brightness disproportionation variation correction side of double norm mixed constraints provided in above method embodiment is realized
Method.
Claims (10)
1. a kind of remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints, which is characterized in that including walking as follows
It is rapid:
(1) remote sensing image is decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;
(2) L is used2Norm constrains illumination component, using L1Norm constrains reflecting component, and it is mixed to obtain double norms
The Variation Model of contract beam;
(3) Variation Model of double norm mixed constraints is solved, the illumination component and reflecting component after being corrected,
And according to after correction illumination component and reflecting component corrected after remote sensing image.
2. the remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints according to claim 1, feature
The step of being, the Variation Model of double norm mixed constraints solved are as follows:
Auxiliary quantity is introduced using division Bregman method, makes L1Norm and L2The bound term of norm is not directly relevant to;
Penalty term is added, converts unconstrained problem for constrained optimization problem;
By carrying out optimization processing to illumination component, auxiliary component and reflecting component, the illumination component after being corrected and anti-
Penetrate component.
3. the remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints according to claim 2, feature
It is, optimization processing is carried out to the illumination component, auxiliary component and reflecting component using alternative iteration method;Work as illumination component
Residual error when being less than limit difference, corresponding illumination component and auxiliary component are optimal illumination component and auxiliary component.
4. the remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints according to claim 1, feature
It is, the Variation Model of double norm mixed constraints is
Wherein, l is illumination component, and r is reflecting component,For L1Norm, λ1、λ2For non-negative parameter,For fidelity term,
Constrain the similarity between l+r and s;It is to constrain the texture and minutia of reflected image, passes through parameter lambda1Into
The adjustment of row weight;It is to constrain the spatial smoothness of ray images, passes through parameter lambda2Carry out weight adjustment.
5. the remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints according to claim 1, feature
Be, after remote sensing image is decomposed into ray images and reflected image, by take Logarithmic calculation obtain corresponding illumination component and
Reflecting component;After illumination component and reflecting component after being corrected, the light after correction is calculated by corresponding fetching number
According to image and reflected image, and according to after correction ray images and reflected image corrected after remote sensing image.
6. a kind of remote sensing image brightness disproportionation variation of double norm mixed constraints corrects system, including processor and memory, institute
State the computer program being stored on memory for executing on a processor;It is characterized in that, described in the processor execution
Following steps are realized when computer program:
(1) remote sensing image is decomposed into ray images and reflected image, obtains corresponding illumination component and reflecting component;
(2) L is used2Norm constrains illumination component, using L1Norm constrains reflecting component, and it is mixed to obtain double norms
The Variation Model of contract beam;
(3) Variation Model of double norm mixed constraints is solved, the illumination component and reflecting component after being corrected,
And according to after correction illumination component and reflecting component corrected after remote sensing image.
7. the remote sensing image brightness disproportionation variation of double norm mixed constraints according to claim 6 corrects system, feature
The step of being, the Variation Model of double norm mixed constraints solved are as follows:
Auxiliary quantity is introduced using division Bregman method, makes L1Norm and L2The bound term of norm is not directly relevant to;
Penalty term is added, converts unconstrained problem for constrained optimization problem;
By carrying out optimization processing to illumination component, auxiliary component and reflecting component, the illumination component after being corrected and anti-
Penetrate component.
8. the remote sensing image brightness disproportionation variation of double norm mixed constraints according to claim 7 corrects system, feature
It is, optimization processing is carried out to the illumination component, auxiliary component and reflecting component using alternative iteration method;Work as illumination component
Residual error when being less than limit difference, corresponding illumination component and auxiliary component are optimal illumination component and auxiliary component.
9. the remote sensing image brightness disproportionation variation of double norm mixed constraints according to claim 6 corrects system, feature
It is, the Variation Model of double norm mixed constraints is
Wherein, l is illumination component, and r is reflecting component,For L1Norm, λ1、λ2For non-negative parameter,For fidelity term,
Constrain the similarity between l+r and s;It is to constrain the texture and minutia of reflected image, passes through parameter lambda1Into
The adjustment of row weight;It is to constrain the spatial smoothness of ray images, passes through parameter lambda2Carry out weight adjustment.
10. the remote sensing image brightness disproportionation variation of double norm mixed constraints according to claim 6 corrects system, feature
Be, after remote sensing image is decomposed into ray images and reflected image, by take Logarithmic calculation obtain corresponding illumination component and
Reflecting component;After illumination component and reflecting component after being corrected, the light after correction is calculated by corresponding fetching number
According to image and reflected image, and according to after correction ray images and reflected image corrected after remote sensing image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811280676.5A CN109658342A (en) | 2018-10-30 | 2018-10-30 | The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811280676.5A CN109658342A (en) | 2018-10-30 | 2018-10-30 | The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109658342A true CN109658342A (en) | 2019-04-19 |
Family
ID=66110486
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811280676.5A Pending CN109658342A (en) | 2018-10-30 | 2018-10-30 | The remote sensing image brightness disproportionation variation bearing calibration of double norm mixed constraints and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109658342A (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7034963B2 (en) * | 2001-07-11 | 2006-04-25 | Applied Materials, Inc. | Method for adjusting edges of grayscale pixel-map images |
US20070217713A1 (en) * | 2004-12-16 | 2007-09-20 | Peyman Milanfar | Robust reconstruction of high resolution grayscale images from a sequence of low resolution frames |
CN101408975A (en) * | 2008-11-28 | 2009-04-15 | 哈尔滨工业大学 | Method for smoothing self-regulation totality variation image base on edge confidence degree |
CN101441765A (en) * | 2008-11-19 | 2009-05-27 | 西安电子科技大学 | Self-adapting regular super resolution image reconstruction method for maintaining edge clear |
CN103530848A (en) * | 2013-09-27 | 2014-01-22 | 中国人民解放军空军工程大学 | Double exposure implementation method for inhomogeneous illumination image |
CN105551000A (en) * | 2015-12-11 | 2016-05-04 | 中国科学院遥感与数字地球研究所 | Remote sensing image reconstruction method based on reference image structure constraint and non-convex low rank constraint |
CN105678715A (en) * | 2016-02-15 | 2016-06-15 | 江苏大学 | Method for removing image noise based on kernel regression total variation |
CN106981052A (en) * | 2017-01-12 | 2017-07-25 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | The uneven variation bearing calibration of adaption brightness based on variation framework |
WO2018022864A1 (en) * | 2016-07-28 | 2018-02-01 | The Texas A&M University System | Method and apparatus for radiometric calibration and mosaicking of aerial images |
CN108268826A (en) * | 2016-12-30 | 2018-07-10 | 广东精点数据科技股份有限公司 | A kind of image enhancement technique and Classifying Method in Remote Sensing Image based on fuzzy set theory |
CN108364261A (en) * | 2017-12-13 | 2018-08-03 | 湖北工业大学 | A kind of TV-Retinex single-frame images defogging methods of gradient guiding |
-
2018
- 2018-10-30 CN CN201811280676.5A patent/CN109658342A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7034963B2 (en) * | 2001-07-11 | 2006-04-25 | Applied Materials, Inc. | Method for adjusting edges of grayscale pixel-map images |
US20070217713A1 (en) * | 2004-12-16 | 2007-09-20 | Peyman Milanfar | Robust reconstruction of high resolution grayscale images from a sequence of low resolution frames |
CN101441765A (en) * | 2008-11-19 | 2009-05-27 | 西安电子科技大学 | Self-adapting regular super resolution image reconstruction method for maintaining edge clear |
CN101408975A (en) * | 2008-11-28 | 2009-04-15 | 哈尔滨工业大学 | Method for smoothing self-regulation totality variation image base on edge confidence degree |
CN103530848A (en) * | 2013-09-27 | 2014-01-22 | 中国人民解放军空军工程大学 | Double exposure implementation method for inhomogeneous illumination image |
CN105551000A (en) * | 2015-12-11 | 2016-05-04 | 中国科学院遥感与数字地球研究所 | Remote sensing image reconstruction method based on reference image structure constraint and non-convex low rank constraint |
CN105678715A (en) * | 2016-02-15 | 2016-06-15 | 江苏大学 | Method for removing image noise based on kernel regression total variation |
WO2018022864A1 (en) * | 2016-07-28 | 2018-02-01 | The Texas A&M University System | Method and apparatus for radiometric calibration and mosaicking of aerial images |
CN108268826A (en) * | 2016-12-30 | 2018-07-10 | 广东精点数据科技股份有限公司 | A kind of image enhancement technique and Classifying Method in Remote Sensing Image based on fuzzy set theory |
CN106981052A (en) * | 2017-01-12 | 2017-07-25 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | The uneven variation bearing calibration of adaption brightness based on variation framework |
CN108364261A (en) * | 2017-12-13 | 2018-08-03 | 湖北工业大学 | A kind of TV-Retinex single-frame images defogging methods of gradient guiding |
Non-Patent Citations (2)
Title |
---|
SHUO LI等: "Variational method based on Retinex with double-norm hybrid constraints for uneven illumination correction", 《JOURNAL OF APPLIED REMOTE SENSING》 * |
李慧芳: "多成因遥感影像亮度不均的变分校正方法研究", 《中国博士学位论文全文数据库(信息科技辑)》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110046673B (en) | No-reference tone mapping image quality evaluation method based on multi-feature fusion | |
CN105474261B (en) | For improving the method, system and computer program product of picture quality | |
Fu et al. | A weighted variational model for simultaneous reflectance and illumination estimation | |
CN105721846B (en) | Polyphaser apparatus and method for synchronous AWB | |
CN105654436B (en) | A kind of backlight image enhancing denoising method based on prospect background separation | |
WO2019242329A1 (en) | Convolutional neural network training method and device | |
CN106971153B (en) | Illumination compensation method for face image | |
US9336578B2 (en) | Interactive tone mapping for high dynamic range video | |
US8310499B2 (en) | Balancing luminance disparity in a display by multiple projectors | |
US20110150358A1 (en) | Method and system for factoring an illumination image | |
CN104063848A (en) | Enhancement method and device for low-illumination image | |
US20130136348A1 (en) | Method and system for generating intrinsic images using a single reflectance technique | |
Anand et al. | Fast GPU-enabled color normalization for digital pathology | |
El Khoury et al. | Color and sharpness assessment of single image dehazing | |
CN108665421A (en) | The high light component removal device of facial image and method, storage medium product | |
US20130114911A1 (en) | Post processing for improved generation of intrinsic images | |
CN109613023A (en) | A kind of fruit surface defect rapid detection method of regional luminance adaptively correcting | |
CN107566689A (en) | System and method for changing input image data | |
WO2013077946A2 (en) | Color analytics for a digital image | |
CN109859132A (en) | A kind of product quality online test method based on enhancing image recognition technology | |
CN106157264A (en) | Large area image uneven illumination bearing calibration based on empirical mode decomposition | |
US9754155B2 (en) | Method and system for generating intrinsic images using a single reflectance technique | |
CN101510306B (en) | Estimation method for video image illumination distribution | |
Meylan | Tone mapping for high dynamic range images | |
Wang et al. | A background correction method for particle image under non-uniform illumination conditions |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190419 |