CN107566821A - A kind of image color moving method based on multi-dimensional association rule - Google Patents
A kind of image color moving method based on multi-dimensional association rule Download PDFInfo
- Publication number
- CN107566821A CN107566821A CN201710746803.5A CN201710746803A CN107566821A CN 107566821 A CN107566821 A CN 107566821A CN 201710746803 A CN201710746803 A CN 201710746803A CN 107566821 A CN107566821 A CN 107566821A
- Authority
- CN
- China
- Prior art keywords
- mrow
- color
- association rule
- brightness
- strong association
- 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.)
- Withdrawn
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a kind of image color moving method based on multi-dimensional association rule, comprise the following steps：Using the Strong association rule collection between the brightness in the Apriori optimized algorithms excavation target image of more attribution rules constraint, color, and introducing class label element is concentrated to ultimately generate luminance levels color Strong association rule collection in Strong association rule；The brightness of pixel, generic in source images are extracted, new R, G, B color component corresponding to each pixel is generated based on the mapping of Strong association rule collection, generates new cromogram.
Description
Technical field
The present invention relates to a kind of image color migrating technology, particularly a kind of pattern colour based on multi-dimensional association rule
Color moving method.
Background technology
Human eye has the characteristics of higher resolution ratio and susceptibility to coloured image, and coloured image can lift people
Cognition to target and scene information, so coloured image is all taken seriously in the every field such as modern military, civilian, and apply
It is more and more extensive.Painted to gray level image, and image color migration is belonged to according to the color of target image change source images
Category.
The content of the invention
It is an object of the invention to provide a kind of image color moving method based on multi-dimensional association rule, including with
Lower step：
Step 1, using more attribution rules constraint Apriori optimized algorithms excavate target image in brightness, color it
Between Strong association rule collection, and concentrated in Strong association rule and introduce class label element and ultimately generate brightness-classification-color and close by force
Join rule set；
Step 2, the brightness of pixel, generic in source images are extracted, each picture is generated based on the mapping of Strong association rule collection
New R, G, B color component corresponding to element, generates new cromogram.
Using the above method, step 1 includes procedure below：
Step 1.1, the object corresponding target image corresponding with each object in source images is found, object corresponds to target figure
As the training sample set BLOCK of composition respective objectsi, i is the index value of object；
Step 1.2, to each training sample set BLOCKiAdd same class label labeli, different training sample sets
BLOCKiDifferent labels are added, form tally set trainlabel=[label1,label2,...labeli,...,labeln], n
For the quantity of object；
Step 1.3, each training sample set BLOCK is extractediMiddle all objects correspond to target image feature set hog,
Average, variance, homogeney, entropy }；
Step 1.4, most suitable set of image characteristics T is chosen as grouped data traindata=[T (BLOCK1),T
(BLOCK2),...,T(BLOCKi),...T(BLOCKn)]；
(selection can allow the best set of image characteristics of classifying quality)
Step 1.5, SVM classifier model is generated
Model=svmtrain (trainlabel, traindata, '-s 0-t 2')
Step 1.6, training sample set BLOCK is extractediIn each pixel brightness y and color component R, G, B, by y, R, G,
B component carries out section, transaction database database corresponding to foundation；
Step 1.7, the Apriori optimized algorithms based on the constraint of more attribution rules realize the uniqueness of color transfer.
Step 1.7 comprises the following steps：
Step 1.7.1, set support support and confidence level confidence；
Step 1.7.2, constantly filter out by connection, beta pruning using formula (4) (5) and be unsatisfactory for support parameter request
Useless item collection, generate frequent 4- items item collection
Step 1.7.3, using formula (6) (7), filter out the correlation rule that confidence level is less than Confidence, remainder association
Rule forms Strong association rule collection to be constrained
Step 1.7.4, according to the section of y, R, B, G component of determination, convolution (8) is treated constraint Strong association rule collection and done
Further constraint, generate qualified " brightness-color " initial Strong association rule collection
A∈[a1,a2],B∈[b1,b2] (8)
Step 1.7.5, " brightness-color " initial Strong association rule concentrate add class label label formed " brightness-
Classification-color " Strong association rule collection；
Step 1.7.6, it is to be searched strong that different objects is corresponded into Strong association rule collection composition generation corresponding to target image
Association Rules.
Using the above method, the detailed process of step 2 is：
Step 2.1, moving window is selected in source images, based on the SVM classifier model trained in moving window
Image classified, and class label is assigned to top left corner pixel；
Step 2.2, each pixel in view picture source images is traveled through, according to the classification of each pixel in obtained source images
Label and corresponding brightness value generation " brightness-classification " collection to be mapped；
Step 2.3, collection to be mapped concentrates search mapping one by one, R, G, B face corresponding to imparting in Strong association rule to be searched
Colouring component value, ultimately produce the color transfer design sketch of source images.
The present invention compared with prior art, has advantages below：(1) it is easy to hardware conversion, it is possible to achieve the color of real-time
Color transition process：When carrying out color transfer, source images need to only be classified and color maps, substantially reduce color
The time of migration, thus the real-time of color transfer can be realized；(2) color transfer effect is good, is sufficiently close to target image：When
When the interval number that brightness and color component are divided into is enough, the color transfer design sketch essentially identical with target image will be obtained.
With reference to Figure of description, the invention will be further described.
Brief description of the drawings
Fig. 1 is the schematic diagram of image color moving method of the present invention based on multi-dimensional association rule.
Fig. 2 is the flow chart of step 1 in image color moving method of the present invention based on multi-dimensional association rule.
Fig. 3 is the inventive method respectively target image used in experiment, wherein (a) is target image, (b) used in inventive algorithm
Target image used by for welsh algorithms.
Fig. 4 is color transfer design sketch corresponding to different colours interval number, wherein (a) is source images, (b) is target figure
Picture, (c) are 20 section design sketch, and (d) is 50 section design sketch.
Fig. 5 is two width different angle achromatic maps in two kinds of different scenes, wherein (a) is a kind of angle in first scene
Achromatic map is spent, (b) is another angle achromatic map in first scene, and (c) is that a kind of angle in second scenario is monochromatic
Figure, (d) are another angle achromatic map in second scenario.
Fig. 6 is the low-light figure under three kinds of different illumination in Same Scene, wherein under the conditions of (a) is the illumination 1 in scene
Twilight image, (b) are the twilight image under the conditions of the illumination 2 in scene, and (c) is the low-light figure under the conditions of the illumination 3 in scene
Picture.
Embodiment
Invention proposes a kind of brand-new, efficient, single channel source images color transfer method based on correlation rule.By
Reinhard et al. propose global color migration algorithm can only the coloured image similar to those overall keynotes have good effect
Fruit, once similitude is poor between image, then and color transfer effect is just less obvious, in some instances it may even be possible to can fail.User mutual
Although image color migration can improve the accuracy rate of color transfer to a certain extent, its execution efficiency reduces
, and man-machine interaction influenceed by subjective factor it is bigger.The inventive method proposes a kind of new side of image color migration
Method, algorithm is simple, and color transfer effect is good.
With reference to Fig. 1~2, it is proposed by the present invention based on multi-dimensional association rule image color moving method the step of such as
Under：
Step S1, the more attribution rules of source images excavate.
The Apriori optimized algorithms constrained using more attribution rules, excavate bright corresponding to different objects in target image
Degree, the Strong association rule collection between color, and introduce class label element, and by different objects for Strong association rule collection group
It is combined, ultimately generates brightness-classification-color Strong association rule collection；
For the Mining Problems of Strong association rule collection, the present invention divides the bright of each pixel in Object Extraction target image first
Spend y and its corresponding color component R, G, B, it is proposed that a kind of Apriori optimized algorithms of more attribution rule constraints, excavate generation
" brightness-color " Strong association rule collection.Found by experiment, " brightness-color " the Strong association rule collection of generation is only to source figure
The more simple image of the luminance level of different objects has preferable color transfer effect as in, and it is complicated to be arranged for brightness,
And the difference in brightness of multiple objects is little or even the problem of color transfer mistake easily occur in identical source images.To understand
Certainly this problem, the present invention concentrate in above-mentioned Strong association rule and introduce " class label " element, and excavation generates " brightness-class
Not-color " Strong association rule collection.Detailed process is described as follows：
Step S101, all " the corresponding target images of object " corresponding with object in source images are found out, respectively to " object
Corresponding target image " carries out piecemeal processing, and generation libsvm classification functions realize the training sample BLOCK of classification；
Step S102, each " object corresponds to target image " BLOCK that the present invention is given in target image enclose class label
Label, such as formula (1)
Trainlabel=[label1,label2,...labeli,...,labeln] (1)
Step S103, extract target image set of image characteristics { hog, average, variance, homogeney, entropy } to be selected accordingly；
Step S104, by observing classifying quality, most suitable set of image characteristics T is selected as grouped data
Traindata, such as formula (2)
Traindata=[T (BLOCK1),T(BLOCK2),...,T(BLOCKi),...T(BLOCKn)] (2)
So-called " most suitable ", which refers to selection, can allow the best set of image characteristics of classifying quality, that is, meeting can be to corresponding picture
Element is enclosed correct class label and is defined；
Step S105, generation SVM classifier (model), such as formula (3)
Model=svmtrain (trainlabel, traindata, '-s 0-t 2') (3)
Svmtrain is a power function that can realize training, and '-the t 2 ' of-s 0 are consolidating for this power function expression
Fix formula.
Step S106, the brightness y of each pixel is extracted in " object corresponds to target image " BLOCK with x pixel
With color component R, G, B, y, R, G, B component are subjected to section, transaction database database corresponding to foundation；
Sectionization is exactly that the tonal range of piece image is compressed into from 0-255 to 0-10,10 or 20 as 0-20
Individual section, result is can be so that efficiency of association below be increased dramatically, but also can ensure that image is clear
It is clear variable.In the present invention, when to brightness y and color component RGB demarcation intervals, because the interval range of y and RGB component is
0-255, it is divided into 10 sections to respective component, this method can be divided into y serial number 0-9 sections, and R is 10-19 sections, G 20-
29 sections, B is 30-39 sections, according to the distributed area corresponding to each component, it is possible to constrained association rules.For example associate
Rule should beSo can retained by screening " left side numeral in 0-9 scopes, digits right difference
In 10-19,20-29,30-39 scopes " realize.
Step S107, based on the Apriori optimized algorithms of more attribution rules constraint, realize the unique of pixel color migration
Property, detailed process is as follows：
Step S1071, set support support and confidence level Confidence；
Step S1072, support parameter is based on using formula (4), formula (5), constantly by connection, two steps of beta pruning,
The useless item collection for being unsatisfactory for support parameter request is filtered out, generates frequent 4- items item collection (4-Items)
Step S1073, confidence level parameter is based on using formula (6), formula (7), confidence level is filtered out and is less than threshold parameter
Confidence useless weak rigidity rule, excavation generate Strong association rule collection to be constrained (strong rules set for
constraint)
Step S1074, the area corresponding to the brightness y determined respectively according to interval number and section distribution, color component R, G, B
Between, convolution (8) treats constraint, and Strong association rule collection is further is constrained, and it is initial to generate qualified " brightness-color "
Strong association rule collection (Ori (strong rules set))
A∈[a1,a2],B∈[b1,b2] (8)
In formula, A uniformly refers to the element on the left side, and B refers to the element on the right,Correlation rule is represented, a1, a2 divide
It is not the section span that the section span of correlation rule left element, b1, b2 are respectively element on the right side of correlation rule；
Step S1705, concentrates in " brightness-color " Strong association rule and adds class label element (labeln '), and generation is each
Individual " object corresponds to target image " corresponding " brightness-classification-color " Strong association rule collection ((strong rules set) n)；
Step S1706, then by Strong association rule collection (strong corresponding to all different " object corresponds to target image "
Rules set) n combines, generate Strong association rule collection (the strong rules to be searched of final whole target image
set)。
Whole process is as shown in Figure 2.
Step S2, the image color migration of rule-based mapping.
The brightness of pixel and generic in source images are extracted, based on the Strong association rule collection for excavating generation, mapping life
Into R, G, B color component corresponding to each pixel, color transfer design sketch is generated.
For image color migration problem, the present invention selected first in source images size for 16 × 16 moving window,
Image in window block is classified based on the SVM classifier trained, and " class label " is assigned to top left corner pixel, time
Each pixel gone through in entire image, according to " class label " and the corresponding brightness of each pixel in obtained source images
It is worth can generation " brightness-classification " collection to be mapped, collection to be mapped will be concentrated in the Strong association rule excavated and searched one by one
Rope maps, R, G, B color component value corresponding to imparting, ultimately produces the color transfer design sketch of source images.Detailed process describes
It is as follows：
Step S201, selects moving window in source images, based on the SVM classifier model trained to scheming in window
As block is classified, and class label is assigned to top left corner pixel；Such as formula (9), based on the SVM classifier trained
(model) each pixel in source images is classified, class label value is L, generates class label matrix F
Step S202, by expanded images border, each pixel in entire image is traveled through, according to obtained source images
In each pixel class label and corresponding brightness value generation brightness-classification collection to be mapped；With reference to the source images extracted
In each pixel brightness y, collection D to be mapped is generated, such as formula (10)
Step S203, based on the Strong association rule collection excavated, collection to be mapped is by the Strong association rule excavated
Search mapping, R, G, B color component value corresponding to imparting one by one are concentrated, ultimately produces the color transfer design sketch of source images, such as
Formula (11)
Advantage in terms of image color migration to illustrate the invention, because the target image in the inventive method has phase
Like the adaptability of scene, so first giving color transfer of the present invention experiment selected target image, as shown in Figure 3.
Fig. 4 is color transfer design sketch corresponding to different colours interval number, wherein (a) is source images, (b) is target figure
Picture, (c) are 20 section design sketch, and (d) is 50 section design sketch.Fig. 4 (c) is to be divided into brightness of image y and color component R, G, B
Behind 20 sections, obtained color transfer design sketch；After Fig. 4 (d) is divided into 50 sections, obtained color transfer design sketch.
It is apparent that after brightness of image y and color component R, G, B are divided into 50 sections from Fig. 4 (d), removed in design sketch
Extremely indivedual regional areas influence less color transfer failure caused by classification error and intersection color transfer is wrong
Outside by mistake, strip, distortion the color with white dog has essentially become the color of continuous uniform, and the white with dog
With green more the access expansion RGB on meadow.It can be seen that the interval number increase of brightness of image, color component is to a certain extent
When, color transfer effect can be with the targeted color of infinite approach continuous uniform.
Fig. 5 (a)~(d) is two width different angle achromatic maps in two kinds of different scenes.As shown in figure 5, the inventive method
First using the inventive method to the ash corresponding to 4 width coloured pictures of two different angles in two kinds of different factory's scenes being shot with CMOS
Degree figure carries out color transfer, and color transfer design sketch is contrasted with original color image, highlights the effective of the inventive method
Property, it can be found that the present invention from the color transfer design sketch of two width different angle monochrome images in two groups of different scenes of Fig. 5
Algorithm can obtain good color transfer effect, in the design sketch that experiment obtains, have been able to grove, lawn, road
Carry out significantly differentiation and colorization；And according to the color transfer design sketch behind Fig. 4 more sections, it can be seen that,
When the section that brightness and color component are divided into is enough, inventive algorithm can obtain continuous uniform, infinite approach mesh
Mark colored color transfer effect.
Tested with the twilight image colorization for belonging to color transfer category, and by experiment effect and Welsh algorithm comparisons, it is convex
The superiority of aobvious the inventive method, as shown in Figure 6.Fig. 6 experiments are shone from three width in the similar scene of class EMCCD shootings are different
Source images under the conditions of degree.As can be seen that source figure of the algorithm proposed by the present invention under three kinds of illumination conditions from experiment effect
As colorization effect is all to be substantially better than welsh algorithms.Inventive algorithm can be clearly by Fig. 6 (a), (b), (c)
Lawn and trees be distinguished, bright-coloured green is caught to tree portion, corresponding lark, three width can be caught to lawn
Colorization design sketch color contrast is distinct, and visual effect is preferable.And the colorization design sketch that welsh algorithms are showed, not
Have and substantially distinguished on trees and ground, whole visual effect has larger inferior position compared with inventive algorithm.Because welsh is calculated
Method is to realize image colorization based on the statistical distribution of part, and the noise in source images is often more, can be in very great Cheng
The calculating of partial statistics distribution and follow-up matching work are influenceed on degree；And inventive algorithm is realized based on image classification
Color transfer (colorization), first passes through the characteristics of mean of image block, and different objects are substantially distinguished, re-map corresponding face
Color, so color transfer (colorization) effect can be relatively good.
Claims (4)
1. a kind of image color moving method based on multi-dimensional association rule, it is characterised in that comprise the following steps：
Step 1, using between the brightness in the Apriori optimized algorithms excavation target image of more attribution rules constraint, color
Strong association rule collection, and concentrate introducing class label element to ultimately generate brightness-classification-color in Strong association rule and associate rule by force
Then collect；
Step 2, the brightness of pixel, generic in source images are extracted, each pixel institute is generated based on the mapping of Strong association rule collection
Corresponding new R, G, B color component, generates new cromogram.
2. according to the method for claim 1, it is characterised in that step 1 includes procedure below：
Step 1.1, the object corresponding target image corresponding with each object in source images is found, object corresponds to target image group
Into the training sample set BLOCK of respective objectsi, i is the index value of object；
Step 1.2, to each training sample set BLOCKiAdd same class label labeli, different training sample set BLOCKiAdd
Add different labels, form tally set trainlabel=[label1,label2,...labeli,...,labeln], n is object
Quantity；
Step 1.3, each training sample set BLOCK is extractediMiddle all objects correspond to target image feature set hog, average,
Variance, homogeney, entropy }；
Step 1.4, most suitable set of image characteristics T is chosen as grouped data traindata=[T (BLOCK1),T
(BLOCK2),...,T(BLOCKi),...T(BLOCKn)]；
(selection can allow the best set of image characteristics of classifying quality)
Step 1.5, SVM classifier model is generated
Model=svmtrain (trainlabel, traindata, '-t the 2' of-s 0)
Step 1.6, training sample set BLOCK is extractediIn each pixel brightness y and color component R, G, B, by y, R, G, B component
Carry out section, transaction database database corresponding to foundation；
Step 1.7, the Apriori optimized algorithms based on the constraint of more attribution rules realize the uniqueness of color transfer.
3. according to the method for claim 2, it is characterised in that step 1.7 comprises the following steps：
Step 1.7.1, set support support and confidence level confidence；
Step 1.7.2, constantly filter out by connection, beta pruning using formula (4) (5) and be unsatisfactory for the useless of support parameter request
Item collection, generate frequent 4- items item collection
<mrow>
<mi>S</mi>
<mi>u</mi>
<mi>p</mi>
<mi>p</mi>
<mi>o</mi>
<mi>r</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&DoubleRightArrow;</mo>
<mi>B</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&cup;</mo>
<mi>B</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>S</mi>
<mi>u</mi>
<mi>p</mi>
<mi>p</mi>
<mi>o</mi>
<mi>r</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&DoubleRightArrow;</mo>
<mi>B</mi>
<mo>)</mo>
</mrow>
<mo>&GreaterEqual;</mo>
<mi>s</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
Step 1.7.3, using formula (6) (7), filter out the correlation rule that confidence level is less than Confidence, remaining correlation rule
Form Strong association rule collection to be constrained
<mrow>
<mi>C</mi>
<mi>o</mi>
<mi>n</mi>
<mi>f</mi>
<mi>i</mi>
<mi>d</mi>
<mi>e</mi>
<mi>n</mi>
<mi>c</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&DoubleRightArrow;</mo>
<mi>B</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>B</mi>
<mo>/</mo>
<mi>A</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>C</mi>
<mi>o</mi>
<mi>n</mi>
<mi>f</mi>
<mi>i</mi>
<mi>d</mi>
<mi>e</mi>
<mi>n</mi>
<mi>c</mi>
<mi>e</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>&DoubleRightArrow;</mo>
<mi>B</mi>
<mo>)</mo>
</mrow>
<mo>&GreaterEqual;</mo>
<mi>c</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
Step 1.7.4, according to the section of y, R, B, G component of determination, convolution (8) is treated constraint Strong association rule collection and done into one
Step constraint, generate qualified " brightness-color " initial Strong association rule collection
A∈[a1,a2],B∈[b1,b2] (8)
A1, a2 are the area that the section span of correlation rule left element, b1, b2 are respectively element on the right side of correlation rule respectively
Between span；
Step 1.7.5, " brightness-color " initial Strong association rule concentrate add class label label formed " brightness-classification-
Color " Strong association rule collection；
Step 1.7.6, different objects is corresponded into Strong association rule collection composition generation strong association to be searched corresponding to target image
Rule set.
4. according to the method for claim 3, it is characterised in that the detailed process of step 2 is：
Step 2.1, moving window is selected in source images, based on the SVM classifier model trained to the figure in moving window
As being classified, and class label is assigned to top left corner pixel；
Step 2.2, each pixel in view picture source images is traveled through, according to the class label of each pixel in obtained source images
" brightness-classification " collection to be mapped is generated with corresponding brightness value；
Step 2.3, collection to be mapped concentrates search mapping one by one, R, G, B color corresponding to imparting point in Strong association rule to be searched
Value, ultimately produce the color transfer design sketch of source images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710746803.5A CN107566821A (en) | 2017-08-27 | 2017-08-27 | A kind of image color moving method based on multi-dimensional association rule |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710746803.5A CN107566821A (en) | 2017-08-27 | 2017-08-27 | A kind of image color moving method based on multi-dimensional association rule |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107566821A true CN107566821A (en) | 2018-01-09 |
Family
ID=60975998
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710746803.5A Withdrawn CN107566821A (en) | 2017-08-27 | 2017-08-27 | A kind of image color moving method based on multi-dimensional association rule |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107566821A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110348064A (en) * | 2019-06-17 | 2019-10-18 | 南京航空航天大学 | Aircraft cockpit interior design method based on stylistic constraint |
| CN110648375A (en) * | 2018-06-26 | 2020-01-03 | 微软技术许可有限责任公司 | Image colorization based on reference information |
| CN114596372A (en) * | 2022-05-07 | 2022-06-07 | 武汉天际航信息科技股份有限公司 | Image color migration method, image consistency improvement method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103606134A (en) * | 2013-11-26 | 2014-02-26 | 国网上海市电力公司 | Enhancing method of low-light video images |
| CN103955902A (en) * | 2014-05-08 | 2014-07-30 | 国网上海市电力公司 | Weak light image enhancing method based on Retinex and Reinhard color migration |
| CN107481183A (en) * | 2017-07-14 | 2017-12-15 | 南京理工大学 | A kind of twilight image colorization method based on multi-dimensional association rule |
-
2017
- 2017-08-27 CN CN201710746803.5A patent/CN107566821A/en not_active Withdrawn
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103606134A (en) * | 2013-11-26 | 2014-02-26 | 国网上海市电力公司 | Enhancing method of low-light video images |
| CN103955902A (en) * | 2014-05-08 | 2014-07-30 | 国网上海市电力公司 | Weak light image enhancing method based on Retinex and Reinhard color migration |
| CN107481183A (en) * | 2017-07-14 | 2017-12-15 | 南京理工大学 | A kind of twilight image colorization method based on multi-dimensional association rule |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648375A (en) * | 2018-06-26 | 2020-01-03 | 微软技术许可有限责任公司 | Image colorization based on reference information |
| CN110648375B (en) * | 2018-06-26 | 2023-07-07 | 微软技术许可有限责任公司 | Image colorization based on reference information |
| CN110348064A (en) * | 2019-06-17 | 2019-10-18 | 南京航空航天大学 | Aircraft cockpit interior design method based on stylistic constraint |
| CN110348064B (en) * | 2019-06-17 | 2023-09-29 | 南京航空航天大学 | Design method for aircraft cockpit inner decoration based on style constraint |
| CN114596372A (en) * | 2022-05-07 | 2022-06-07 | 武汉天际航信息科技股份有限公司 | Image color migration method, image consistency improvement method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104809187B (en) | A kind of indoor scene semanteme marking method based on RGB D data | |
| CN106778604B (en) | Pedestrian re-identification method based on matching convolutional neural network | |
| CN100573523C (en) | A kind of image inquiry method based on marking area | |
| CN109255334A (en) | Remote sensing image terrain classification method based on deep learning semantic segmentation network | |
| CN106384117B (en) | A kind of vehicle color identification method and device | |
| CN108388905B (en) | A kind of Illuminant estimation method based on convolutional neural networks and neighbourhood context | |
| CN106023259B (en) | A kind of moving target frequency detecting method and device | |
| CN107481183B (en) | Low-light-level image colorization method based on multi-dimensional data association rule | |
| CN106650690A (en) | Night vision image scene identification method based on deep convolution-deconvolution neural network | |
| Kawamura et al. | Discriminating crops/weeds in an upland rice field from UAV images with the SLIC-RF algorithm | |
| CN107566821A (en) | A kind of image color moving method based on multi-dimensional association rule | |
| CN103914699A (en) | Automatic lip gloss image enhancement method based on color space | |
| CN105205804A (en) | Caryon-cytolymph separation method and apparatus of white blood cells in blood cell image, and classification method and apparatus of white blood cells in blood cell image | |
| CN104966099B (en) | A kind of foot type sorting technique based on people's pin image | |
| CN107705254A (en) | A kind of urban environment appraisal procedure based on streetscape figure | |
| CN107516103A (en) | A kind of image classification method and system | |
| CN103714181A (en) | Stratification specific figure search method | |
| CN102208099A (en) | Illumination-change-resistant crop color image segmentation method | |
| CN107346413A (en) | Traffic sign recognition method and system in a kind of streetscape image | |
| CN109120919A (en) | A kind of automatic analysis system and method for the evaluation and test of picture quality subjectivity | |
| CN108230340A (en) | A kind of SLIC super-pixel extraction Weighting and super-pixel extracting method based on MMTD | |
| CN103065282A (en) | Image fusion method based on sparse linear system | |
| CN109726669A (en) | Pedestrian identifies data creation method again under different illumination conditions based on confrontation network | |
| CN110298893A (en) | A kind of pedestrian wears the generation method and device of color identification model clothes | |
| CN103020198B (en) | One searches for mesh calibration method in video |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180109 |
|
| WW01 | Invention patent application withdrawn after publication |