CN105931208B - Enhancement algorithm for low-illumination image based on physical model - Google Patents
Enhancement algorithm for low-illumination image based on physical model Download PDFInfo
- Publication number
- CN105931208B CN105931208B CN201610393584.2A CN201610393584A CN105931208B CN 105931208 B CN105931208 B CN 105931208B CN 201610393584 A CN201610393584 A CN 201610393584A CN 105931208 B CN105931208 B CN 105931208B
- Authority
- CN
- China
- Prior art keywords
- image
- low
- algorithm
- light
- level
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005286 illumination Methods 0.000 title claims abstract description 14
- 230000002708 enhancing effect Effects 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims description 16
- 230000000644 propagated effect Effects 0.000 claims description 4
- 238000002474 experimental method Methods 0.000 abstract description 2
- 238000000034 method Methods 0.000 description 16
- 230000006870 function Effects 0.000 description 12
- 230000000694 effects Effects 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000003014 reinforcing effect Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 230000003321 amplification Effects 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 125000001475 halogen functional group Chemical group 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000012958 reprocessing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000000007 visual 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
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/92—Dynamic range modification of images or parts thereof based on global image properties
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of enhancement algorithm for low-illumination image based on physical model, it is extremely difficult to balance between raising brightness and contrast, removal noise and real-time primarily directed to enhancement algorithm for low-illumination image, it is proposed the innovatory algorithm for combining Retinex algorithm and dark channel prior theory, BM3D algorithm is improved in efficiency first, and it is applied in YCbCr space, then go out brightness propagation figure in the rough estimate of the space HSI, the atmospheric physics model under low light conditions is improved, brightness propagation figure is refined in conjunction with Retinex algorithm；Experiment shows that the algorithm can reach the comprehensive enhancing to low-light (level) brightness of image and contrast, and arithmetic speed is also improved significantly.
Description
Technical field
The present invention relates to a kind of algorithm for image enhancement more particularly to a kind of low-light (level) image enhancement calculations based on physical model
Method.
Background technique
Under low light conditions, such as rainy days, night or in mine etc., if without fill-in light, when acquiring image,
Obtained picture quality is often very poor.Not only human eye seems that visual effect is bad to these images, and many key elements are difficult to see
Clearly, and to machine recognition and monitoring tracking aspect very big difficulty is also brought, causes the practical application of low-light (level) image
Substantially reduce, it is desirable to the low-light (level) image bad to these image quality is applied, low-light (level) image must just be carried out enhancing and
Denoising, to improve the quality of low-light (level) image.Traditional images Enhancement Method does not account for the own characteristic of low-light (level) image,
The effect is unsatisfactory.In recent years, specifically for the algorithm research of low-light (level) image enhancement, there has also been very big progress, these methods
Following a few classes can be divided into: one kind is the improvement based on algorithm of histogram equalization, since low-light (level) image is in histogram
Information all concentrates on the lower one end of brightness ratio, and algorithm of histogram equalization is to the merging of gray level and intensity profile range
Expand so that the processing result of this kind of algorithm is often unsatisfactory；Algorithm based on Retinex is most widely paid close attention to, however
Biggish calculation amount and the Halo effect being difficult to avoid that limit its application, or even will appear ash to the image of extremely low illumination
Change；The method of multiple image fusion needs the high-quality image information under Same Scene, and applicable situation is limited；In addition, there are also one
Class is the algorithm of color space enhancing, this kind of algorithm generally all because to extremely low illumination image it is less practical due to be restricted, and count
It calculates big very big.
2011, Tsinghua University scholar proposed a kind of enhancement algorithm for low-illumination image based on defogging technology, by low
Enhanced after illumination image reversion using defogging method, this thought is relatively new, but deep in bright spot presence or scene
Spending discontinuous place will appear hot spot, and treatment of details is also not fine enough.
Summary of the invention
In view of the above problems, the present invention provides a kind of enhancement algorithm for low-illumination image based on physical model.The calculation
Method can reach the comprehensive enhancing to low-light (level) brightness of image and contrast, and arithmetic speed is also improved significantly.
To reach above-mentioned technical purpose, present invention employs a kind of enhancement algorithm for low-illumination image based on physical model,
Specific algorithm includes as follows:
(1) brightness of the grayscale image as rough estimate after taking the luminance component to low-light (level) image to be finely adjusted first
Figure is propagated, as shown in formula (3):
Wherein, k value is the maximum pixel value of brightness histogram in low-light (level) image, and C is constant；
(2) luminance component is isolated, centered on Gaussian function to luminance component processing using Retinex algorithm
For surround function to luminance component processing, formula is as follows:
Lm(x, y)=G (x, y) * L (x, y),
T (x)=C (255-k) Lm,
Wherein, G (x, y) is Gaussian function, and * indicates convolution algorithm；
(3) according to the thought of reversion defogging, low-light (level) image is enhanced, first has to invert original image I:
Rc(x)=255-Ic(x),
Wherein, c indicates RGB color channel, Rc(x) be low-light (level) image reversion, IcIt (x) is original low-light (level) image；
According to physical model formula (1) to RcIt is handled, the formula of restored image is as follows:
Wherein t (x) is the brightness propagation figure estimated in Section 2, and the value of atmosphere light A can be estimated by dark primary priori
Meter；
(4) it according to dark primary priori theoretical, is described with formula, for arbitrary image J, is defined as follows formula:
Wherein, JdarkImage is indicated in the dark of regional area Ω (x), Ω (x) indicates the regional area centered on x
Block, c indicate a certain Color Channel.According to dark primary priori theoretical, if J is outdoor fog free images, JdarkIntensity always
It is very low and level off to 0, the pixel of brightness maximum 0.1% in dark primary is chosen, is found out in the corresponding original image of these pixels
The maximum value of pixel just obtains the estimated value of atmosphere light A.
The present invention is proposed to substitute atmospheric transmissivity in greasy weather physical model with brightness propagation figure, be compensated in master mould thoroughly
It is insufficient to penetrate theory of the rate when being applied to low-light (level) image, keeps the reinforcing effect of low-light (level) image more true to nature.Finally combine
Retinex algorithm refines brightness propagation figure, can further improve the brightness of image after enhancing, and image is thin
Section aspect, which is also more clear, reaches better reinforcing effect.
Detailed description of the invention
Shown in FIG. 1 is BM3D Denoising Algorithm flow chart in the present invention；
Shown in Fig. 2 is that HSI and YCbCr component compares in the present invention；
Shown in Fig. 3 is low-light (level) image enhancement effects figure in the present invention；
Shown in Fig. 4 is that the arithmetic result in the present invention compares that (from top to bottom image is followed successively by building, plant, vehicle, room
It is interior)；
Shown in fig. 5 is the local detail comparison diagram in the present invention.
Specific embodiment
Present invention will be explained in further detail with reference to the accompanying drawings and detailed description.
Analyze the noise of low-light (level) image, the main impulsive noise generated including transimission and storage process, various devices and
The poisson noise generated under Gaussian noise and the very small situation of illumination that transmission channel generates, wherein poisson noise is to low photograph
The image for spending image is maximum.3D denoising is the denoising method for being best suited for low-light (level) image, wherein Dabov K, Foi A etc.[9]It mentions
Three-dimensional Block- matching Denoising Algorithm (BM3D) out is considered as denoising method best at present, it is a kind of enhancing based on transform domain
The denoising method of type rarefaction representation, main thought are as follows:
It being grouped first, this process is that the two dimensional image block of similar structure is grouped together into three-dimensional array,
Optimal estimation is obtained using collaboration filtering later.The estimated value for considering that collaboration filtering obtains may repeat, and estimate repetition
Estimated value should be weighted and averaged, and to image block carry out integration complete basis estimation.Finally basis is estimated as input
Signal is grouped again, and Wiener filtering is carried out together with original signal, obtains final result.BM3D Denoising Algorithm process
Figure is as shown in Figure 1.
Since BM3D algorithm in entire image due to that will match similar modular blocks to each the reference module, operand is often
It is bigger, in order to improve arithmetic speed, the overall situation be limited in region of search apart from calculating in matching module
In certain neighborhood, arithmetic speed is thus substantially increased.
The brightness of low-light (level) image is secretly its most intuitive feature different from normal picture, therefore, to low-light (level) figure
When the processing of picture, it should consider to carry out in the color space of brightness and color separated as far as possible.The face of common color and brightness separation
The colour space has HSI color space and YCbCr color space.Fig. 2 be low-light (level) image and normal picture in HSI color space and
Each histogram of component of YCbCr color space.In the space HSI, the noise that the H component and S component of low-light (level) image include compares
It is more；From the point of view of each histogram of component, the I component of low-light (level) image is due to whole relatively low so that noise is unobvious.Therefore exist
The space HSI will all denoise three components, but the variation of H and S component is bigger to Color influences, during denoising
It will cause color serious distortion.It is smaller in the noise of YCbCr space, Cb the and Cr component of low-light (level) image, Y-component be due to
It is whole relatively low so that noise is unobvious, but actually noise is most strong.Use YCbCr space as image denoising field experiment herein
Scape, and only need to carry out denoising to Y-component that most noise can be inhibited, to substantially increase processing effect
Rate.
The estimation of brightness propagation figure
Use classical Koschmieder atmospherical scattering model as physical model, the parameter in estimation Atmospheric models is also
Original image, to complete the enhancing processing to low-light (level) image.
The rough estimate of figure is propagated brightness in the space HSI
Koschmieder classics atmospherical scattering model formula is as follows:
I (x)=J (x) t (x)+A (1-t (x))
T (x)=e-βd(x)
Wherein, A is sky brightness, and x is spatial position, and I (x) is the intensity i.e. foggy image of observed image, J (x)
For the clear image after defogging, t (x) is propagation in atmosphere function, and d (x) is the scene depth of field.The right first item is straight in formula (1)
Attenuation term is connect, is as atmospheric particles scattering process during scene reflectivity light propagation to imaging device and caused by decaying；The
Binomial is atmosphere light ingredient, caused by being scattered as natural light.Work of t (x) item under greasy weather mode in this atmospheric physics model
With being to measure impurity in atmosphere to the impact factor of light, and the most important influence factor of image quality at low illumination level difference is light
According to condition, what this parameter should reflect be scenery under the influence of illumination can the ability seen of observed person, under this mode
Influence of the weather condition to imaging just seems insignificant, is applied to atmospheric physics model using brightness propagation figure to improve
Imaging model under low light conditions.
Brightness propagation figure is closely related with the luminance component of piece image, therefore same for the extraction of luminance component
It is considered as carrying out on the color space that color and brightness separate.HSI color space is considered as providing the face of most accurate tone
The colour space can keep color constancy, and brightness is handled and can't be changed to color in this color space, because
This, the present invention does the basis of brightness propagation figure using the luminance component in the space HSI.In this paper algorithm, to low-light (level) image
When reverse image is handled, figure alternatively position of the transmissivity in atmospheric physics model is propagated in brightness.It is bright in order to guarantee
Degree propagates the functional similarity of figure with transmissivity, after we take the luminance component to low-light (level) image to be finely adjusted first
Grayscale image propagates figure as the brightness of rough estimate, as shown in formula (3):
Wherein, k value is the maximum pixel value of brightness histogram in low-light (level) image, and C is constant.
Retinex algorithm propagates brightness the reprocessing of figure
It is found when practical application, directly uses the grayscale image obtained by the fine tuning to luminance component as brightness propagation figure
To low-light (level) image handled to result be not it is very satisfactory, reason is mainly manifested in the brightness propagation figure
Scene profile is not changed, almost consistent with original image, is easy instead to thin when being handled with such propagation function
Save information cover, it is therefore desirable to propagation figure in edge contour generally than more visible.In addition, directly with luminance component come
Assessment brightness propagation function also has certain irrationality, because our brightness propagation function is one to scenery light itself
A reflection, and according to Retinex theory, actual luminance component is the synthesis of reflecting component and luminance component, so we need
Micronization processes are carried out to brightness propagation function, using Retinex algorithm to luminance component processing, isolate luminance component, make
With surround function centered on Gaussian function to luminance component processing, can also be obscured while the method estimates luminance component
Profile meets the requirement to brightness propagation function just.It should be noted that the selection of filter radius, filter radius selected
Small enhanced image lacks three-dimensional sense.
Lm(x, y)=G (x, y) * L (x, y)
T (x)=C (255-k) Lm
Wherein, G (x, y) is Gaussian function, and * indicates convolution algorithm.
Low-light (level) image restores
According to the thought of reversion defogging, low-light (level) image is enhanced, first has to invert original image I:
Rc(x)=255-Ic(x)
Wherein, c indicates RGB color channel, Rc(x) be low-light (level) image reversion, IcIt (x) is original low-light (level) image.
According to physical model formula (1) to RcIt is handled, the formula of restored image is as follows:
Wherein t (x) is the brightness propagation figure estimated in Section 2, and the value of atmosphere light A can be estimated by dark primary priori
Meter.
Researcher observes 5000 width images and statistics obtains dark primary priori theoretical, which has started defogging
Theoretical frontier.It is described with formula, for arbitrary image J, we are defined as follows formula:
Wherein, JdarkImage is indicated in the dark of regional area Ω (x), Ω (x) indicates the regional area centered on x
Block, c indicate a certain Color Channel.According to dark primary priori theoretical, if J is outdoor fog free images, JdarkIntensity always
Very low and level off to 0, researcher eliminates the correct of the image authentication of the sky areas part theory with about 5000 width
Property.The pixel for choosing brightness maximum 0.1% in dark primary finds out in the corresponding original image of these pixels the maximum value of pixel just
Obtain the estimated value of atmosphere light A.
The image J finally restored, which carries out once inverting again, obtains the enhancing image E of low-light (level) image.Use this paper algorithm
Cell street image processing effect to 8 points or so shootings at night is as shown in figure 3, the result figure after processing can be very
Clearly see the pedestrian and Che on street.
In order to verify the actual effect of this paper algorithm, we with mobile phone at night 9 points shooting different scenes under several groups
Low-light (level) image, the building on campus tennis court side, the plant in road roadside, vehicle and indoor tables and chairs are as original low-light (level) figure
Picture.At respectively using Retinex algorithm, the enhancing algorithm based on defogging technology and this paper algorithm to this four sub-picture
Reason, reinforcing effect are as shown in Figure 4.It can be seen that Retinex algorithm is ashed when handling low-light (level) image from whole reinforcing effect
Situation is very serious, as there is the overcover of one layer of grey to be covered on above image, can only show the general outline of scenery；Base
In defogging technology enhancing algorithm to low-light (level) image enhancement after, promotion is truly had on brightness and contrast, but overall brightness is still
It is so partially dark, and with cross-color；And this paper algorithm it is apparent improve brightness and contrast, will be some important in image
Information all displays.
To treatment of details effect is observed again after the same regional area amplification of image after three kinds of algorithm process, such as Fig. 5 institute
Show, is the treatment effect amplification to three branch, window, street lamp details respectively.Therefrom we can observe that, inventive algorithm
The branch of very little also can be displayed clearly after processing, and the details of branch is by mould in the enhancing algorithm based on defogging technology
Paste has fallen many information, illustrates that inventive algorithm has apparent advantage in terms for the treatment of of details.To the edge contour of window into
Row observation, we can clearly be seen that the profile of other two kinds of algorithms with the presence of shade, and algorithm of the invention can be fine
Reduction marginal information.
In addition, in the target processing to this localized high intensity of street lamp, the road enhancing algorithm Hui based on defogging technology
The edge of lamp forms Crape ring, the light of street lamp has been truncated, it appears very unnatural, Retinex algorithm has then directly obscured street lamp
Information, and inventive algorithm then can be good at solving the problems, such as this.
Table 1 list the average brightness MeanG of three kinds of algorithms, image information entropy Entropy, Y-PSNR PSNR and
Running time T ime, parameter definition are as follows:
Wherein, f (i, j) indicates the pixel value of point (i, j)；M and N respectively indicates the length and width of image.Being averaged for image is bright
The bright-dark degree of degree description image, if average brightness value is small, image frame is whole partially dark, conversely, then image is whole brighter.
Wherein, piRepresent the probability of i-th of gray level appearance.Entropy be reflect amount of image information standard, entropy size with
The information content that image includes is directly proportional, and comentropy is bigger, and amount of image information is more.
Wherein, f (i, j) indicates original image；G (i, j) indicates the image after denoising；M and N respectively indicate image length and
It is wide.Y-PSNR is bigger, indicates that denoising effect is better.
The experimental situation of present invention calculating arithmetic speed are as follows: program, which is realized, uses MATLAB R2013a, and host is configured to
Inter (R) Core (TM) i5-3470CPU@3.20GHz, 4GB memory, 7 operating system of Window, picture size 2592 ×
1944。
The objective parameter comparing result of table 1 Fig. 4, tetra- groups of images
Observe data, in addition to Retinex algorithm due to whole misty influence show relatively high average brightness it
Outside, the treatment effect of this paper will be better than other two kinds of algorithms, in processing speed than based on defogging technology enhancing algorithm it is fast on
Very much, but it is slower than Retinex algorithm upper.Illustrate to have in balance reinforcing effect and processing speed herein certain excellent
Gesture.
Claims (1)
1. a kind of enhancement algorithm for low-illumination image based on physical model, which is characterized in that specific algorithm includes as follows:
(1) is isolated by luminance component to low-light (level) image, uses Gaussian function for luminance component processing using Retinex algorithm
Centered on surround function to luminance component processing, reuse luminance component estimated brightness propagation figure, formula is as follows:
Lm(x, y)=G (x, y) * L (x, y),
T (x)=C (255-k) Lm,
Wherein, G (x, y) is Gaussian function, and * indicates convolution algorithm, and k value is the maximum pixel of brightness histogram in low-light (level) image
Value, C is constant；
(2) according to the thought of reversion defogging, low-light (level) image is enhanced, first has to invert original image:
Rc(x)=255-Ic(x),
Wherein, c indicates RGB color channel, Rc(x) be low-light (level) image reversion, IcIt (x) is original low-light (level) image,
According to the formula of restored image to Rc(x) it is handled, specific as follows:
Wherein t (x) is that figure is propagated in the brightness of estimation, and the value of atmosphere light A can be estimated by dark primary priori；
Wherein, it according to dark primary priori theoretical, is described with formula, for arbitrary image J, is defined as follows formula:
Wherein, Jdark(x) image is indicated in the dark of regional area Ω (x), and Ω (x) indicates the regional area centered on x
Block, c indicate a certain Color Channel, according to dark primary priori theoretical, if J is outdoor fog free images, Jdark(x) intensity is total
It is very low and levels off to 0, chooses the pixel of brightness maximum 0.1% in dark primary, find out the corresponding original image of these pixels
The middle maximum value of pixel just obtains the estimated value of atmosphere light A；
(3) after using defogging algorithm enhancing reverse image, output result J (x) is inverted again, obtains enhanced low-light (level) figure
As E.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2015105206965 | 2015-08-22 | ||
| CN201510520696 | 2015-08-22 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105931208A CN105931208A (en) | 2016-09-07 |
| CN105931208B true CN105931208B (en) | 2019-02-01 |
Family
ID=56833564
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610393584.2A Active CN105931208B (en) | 2015-08-22 | 2016-06-06 | Enhancement algorithm for low-illumination image based on physical model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105931208B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292830B (en) * | 2016-12-28 | 2020-09-25 | 首都师范大学 | Low-illumination image enhancement and evaluation method |
| CN107451971A (en) * | 2017-07-30 | 2017-12-08 | 湖南鸣腾智能科技有限公司 | The blind convolved image restoring method of low-light (level) of priori is combined based on dark and Gauss |
| CN108269244B (en) * | 2018-01-24 | 2021-07-06 | 东北大学 | Image defogging system based on deep learning and prior constraint |
| CN109358063B (en) * | 2018-04-22 | 2019-10-01 | 河北金特力斯体育设施有限公司 | It hurdles the detection method of integrity degree |
| CN110111268B (en) * | 2019-04-18 | 2021-08-03 | 上海师范大学 | Single image rain removing method and device based on dark channel and fuzzy width learning |
| CN111311503A (en) * | 2019-11-13 | 2020-06-19 | 深圳大学 | Night low-brightness image enhancement system |
| CN111724332B (en) * | 2020-06-09 | 2023-10-31 | 四川大学 | Image enhancement method and system suitable for closed cavity detection |
| CN112488957A (en) * | 2020-12-15 | 2021-03-12 | 南京理工大学 | Low-illumination color image real-time enhancement method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020920A (en) * | 2013-01-10 | 2013-04-03 | 厦门大学 | Method for enhancing low-illumination images |
| CN103177424A (en) * | 2012-12-07 | 2013-06-26 | 西安电子科技大学 | Low-luminance image reinforcing and denoising method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104346774B (en) * | 2013-07-29 | 2018-04-03 | 诺基亚技术有限公司 | Method and apparatus for image enhaucament |
-
2016
- 2016-06-06 CN CN201610393584.2A patent/CN105931208B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177424A (en) * | 2012-12-07 | 2013-06-26 | 西安电子科技大学 | Low-luminance image reinforcing and denoising method |
| CN103020920A (en) * | 2013-01-10 | 2013-04-03 | 厦门大学 | Method for enhancing low-illumination images |
Non-Patent Citations (1)
| Title |
|---|
| 基于物理模型的低照度图像增强算法;王小元 等;《计算机应用》;20150810;第35卷(第8期);2301-2304,2310 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105931208A (en) | 2016-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105931208B (en) | Enhancement algorithm for low-illumination image based on physical model | |
| Zhang et al. | Fast haze removal for nighttime image using maximum reflectance prior | |
| He et al. | Single image haze removal using dark channel prior | |
| Yu et al. | Fast single image fog removal using edge-preserving smoothing | |
| CN108734670B (en) | Method for restoring single night weak-illumination haze image | |
| Chao et al. | Removal of water scattering | |
| CN102930514B (en) | Rapid image defogging method based on atmospheric physical scattering model | |
| CN103020920B (en) | Method for enhancing low-illumination images | |
| Wang et al. | Fast single haze image enhancement | |
| Tripathi et al. | Single image fog removal using bilateral filter | |
| CN107103591B (en) | Single image defogging method based on image haze concentration estimation | |
| Wang et al. | Improved single image dehazing using dark channel prior | |
| Yuan et al. | A region-wised medium transmission based image dehazing method | |
| CN102663697A (en) | Enhancement method of underwater color video image | |
| CN103218778A (en) | Image and video processing method and device | |
| CN105959510B (en) | A kind of video rapid defogging method | |
| JP2013141210A (en) | Image defogging apparatus, image defogging method, and image processing system | |
| CN105096278A (en) | Image enhancement method based on illumination adjustment and equipment thereof | |
| CN103034983A (en) | Defogging method based on anisotropic filtering | |
| Gao et al. | Haze filtering with aerial perspective | |
| CN105023246B (en) | A kind of image enchancing method based on contrast and structural similarity | |
| CN107451975B (en) | A kind of view-based access control model weights similar picture quality clarification method | |
| Park et al. | Single image haze removal using novel estimation of atmospheric light and transmission | |
| CN108550124B (en) | Illumination compensation and image enhancement method based on bionic spiral | |
| Gadnayak et al. | A survey on image dehazing methods |
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 |