CN108133462B - Single image restoration method based on gradient field region segmentation - Google Patents
Single image restoration method based on gradient field region segmentation Download PDFInfo
- Publication number
- CN108133462B CN108133462B CN201711295623.6A CN201711295623A CN108133462B CN 108133462 B CN108133462 B CN 108133462B CN 201711295623 A CN201711295623 A CN 201711295623A CN 108133462 B CN108133462 B CN 108133462B
- Authority
- CN
- China
- Prior art keywords
- image
- transmittance
- segmentation
- region
- background light
- 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
- 230000011218 segmentation Effects 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000002834 transmittance Methods 0.000 claims abstract description 45
- 238000009499 grossing Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims description 32
- 238000009792 diffusion process Methods 0.000 claims description 29
- 238000001914 filtration Methods 0.000 claims description 24
- 230000007797 corrosion Effects 0.000 claims description 7
- 238000005260 corrosion Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 230000007704 transition Effects 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims 1
- 125000001475 halogen functional group Chemical group 0.000 abstract description 6
- 230000000694 effects Effects 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
Abstract
The invention relates to a single image restoration method based on gradient field area segmentation, which respectively converts a zoomed image into an HSI color space and a gray scale space, automatically performs threshold segmentation on a brightness space, performs gradient field segmentation on the gray scale space at the same time to obtain a bright area and a non-bright area of the image, and obtains a value representing global atmosphere background light according to the bright area; acquiring a dark channel image from the zoomed image, smoothing the image by using the constructed NSRAD model to obtain an optimized transmittance image, introducing a multi-region theory, fusing a bright region and a non-bright region, and further correcting the transmittance image; and on the basis of a dark primary color theoretical model, obtaining a defogged image according to the value representing the global atmosphere background light and the corrected transmittance image. The method has strong adaptability to road environment and good algorithm stability, can effectively eliminate the halo phenomenon, and can effectively recover the image in foggy days, so that a monitor can obtain more effective information.
Description
Technical Field
The invention belongs to the field of image processing, and particularly relates to a single image restoration method based on a gradient field region field.
Background
At present, road traffic monitoring is mainly composed of units such as acquisition equipment, transmission equipment, processing equipment and the like. The collecting device is mainly a camera and aims to change the information of roads and pedestrians into image information through a sensor. Current supervisory equipment is applicable to under the better natural environment that does not shelter from usually, when meetting bad weather, like during the fog day, the image that the camera was gathered will be sheltered from, and road monitoring information just can have the loss. Therefore, the method has very important significance in better monitoring the road information of various natural environments.
At present, two general types of image restoration methods exist: a dark primary color based method, an image enhancement based method. The method based on the dark primary color theory generally adopts a method of directly estimating the atmospheric background light to obtain the value of the global atmospheric background light, carries out simple smoothing operation on the transmissivity image, and then carries out restoration on the foggy day image through a dark channel model. Although the method can achieve the aim of restoration, the rough acquisition of the atmospheric background light directly influences the image brightness of the image restoration; the depth of the smoothing operation degree on the transmittance directly affects the visual effect of the restored image. In most cases, the processed image will show halo in the sky area. Based on an enhancement method, a Retinex theory is generally adopted for processing, and although the multiscale Gaussian convolution filtering adopted by the multiscale method has a good effect on dense fog, an over-enhancement phenomenon can occur on the dense fog.
The uniformity of background light in the environment influences the quality of a video acquisition image, a brighter area in the image, namely an area where the background light in the environment influences the image, is not considered when an acquired foggy day image is processed by a traditional dark primary color theory, global processing is adopted when the transmittance of the image is estimated, the estimated transmittance is small, after the dark primary color model is processed and recovered, the overall brightness of the image is dark, and a halo phenomenon to a certain degree appears at the edge of the brighter area.
Disclosure of Invention
The invention aims to provide a single image restoration method based on gradient field region segmentation, which aims at restoring a foggy image, avoids a halo phenomenon, stabilizes the restoration quality of the image, enables a monitor to acquire more useful information and can more effectively monitor the road condition.
The invention relates to a restoration method of a single image based on gradient field region segmentation, which is used for zooming an input single image; on one hand, the zoomed image is respectively converted into an HSI color space and a gray scale space, a brightness space of the image is obtained in the HSI color space, automatic threshold value segmentation is carried out on the brightness space, meanwhile, gradient field segmentation is carried out on the gray scale space, the two segmentation areas are integrated to obtain a bright area and a non-bright area of the image, and a value representing global atmosphere background light is obtained according to the bright area; on the other hand, a dark channel image is obtained from the zoomed image, a constructed NSRAD model is used for smoothing to obtain an optimized transmittance image, the transmittance image is enlarged to the size of an original image, a multi-region theory is introduced, the obtained bright region and the obtained non-bright region are fused, and the transmittance image is further corrected; and finally, on the basis of a dark primary color theoretical model, obtaining a defogged image according to the obtained value representing the global atmosphere background light and the corrected transmittance image.
The method specifically comprises the following steps:
step 1, obtaining a value representing global atmosphere background light
(1) Zooming the input single image to half of the size of the original image;
(2) converting the zoomed image into an HSI color space, acquiring a brightness space of the image, performing automatic threshold processing on the brightness space, and finally obtaining a segmentation area through median filtering processing and morphological minimum area corrosion;
(3) simultaneously converting the zoomed image into a gray space, converting the image into a gray image, acquiring the gradient of the gray image, performing binary segmentation by using the threshold value of the gradient, and finally obtaining a segmentation area through median filtering processing and morphological minimum area corrosion;
(4) and operating the two obtained segmentation areas to obtain a sky bright area in the image, and calculating a value A representing global atmosphere background light in the bright area:
[Am，As，Is]＝Th(I(x))
where Th (x) is the thresholding of the gradient domain and luminance space, I (x) is the input image, Am、AsAs the mean of the pixels in the bright areaAnd variance, Is an image subjected to binarization after segmentation;
constructing an adaptive function fitting a value A representing global atmospheric background light:
in which Δ A is a preset correction factor and A is AmAnd AsThe value range of k [0,1 ]]θ is a threshold for determination;
step 2, constructing NSRAD filtering algorithm
The expression for constructing the NSRAD model is as follows:
in the formula (I), the compound is shown in the specification,
representing the edge of Ω, n is
F (t) is a diffusion equation, t is an input preliminary estimated transmittance image, c (q) is a diffusion coefficient;
wherein α is an adjustable integer greater than 0, controls the descending speed of diffusion coefficient,
wherein Z is a natural number, the diffusion speed is adjusted, the larger the diffusion speed is, the faster the diffusion coefficient is reduced, the control coefficient β of the homogeneous region is increased, the detail contrast of the transition region is improved, q is a factor for controlling the diffusion coefficient obtained by calculating local variance, and q is a natural number0Is the defined noise variance coefficient, β is the control parameter of the homogeneous region;
step 3, optimizing transmittance image
Obtaining a dark channel image J from an input single image after scalingd(x) Obtaining an optimized transmittance image t through NSRAD model filtering processing2(x):
t2(x)＝f1(Jd(x))
In the formula (f)1(x) Representation of dark channel image Jd(x) Performing NSRAD model filtering processing, t2(x) The transmittance image after the filtering processing is obtained;
step 4, optimizing the transmittance image t2(x) After the image is enlarged to the size of an original image, a multi-region theory is introduced, aiming at the step 1, after the value representing the global atmosphere background light is obtained, the optimized transmissivity image t is obtained2(x) And dark channel image Jd(x) Performing multi-region fusion to obtain accurately corrected transmittance image t3(x) The formula is as follows:
t3(x)＝t2(x)·(1-Is)+IS·(λ·Jd(x)+(1-λ)·t2(x))
in the formula, t2(x) Is a transmittance image after being filtered by an NSRAD model, lambda Is a preset fusion coefficient, Is an image which Is divided and binarized, Jd(x) Is a dark channel image;
and 5, on the basis of a dark primary color theoretical model, obtaining a value A representing the global atmosphere background light and a corrected transmissivity image t according to the value A and the corrected transmissivity image3(x) Obtaining a defogged image:
for formulas I (x) ═ J (x) t (x) + A (1-t (x)) and
go on simplification to order I1(x) When the simplified model is applied, the restored image obtained by the simplified model is:
wherein I (x) is an input image, J (x) is an image in the case of no fog, t (x) represents the transmittance of the medium, A is a value representing the global atmospheric background light, Ic(y) an image of one of the three RGB channels representing an input image, t1(x) Is a preliminarily estimated transmittance image, ω is a tuning parameter;
the restoration image obtained by the simplified model is dark as a whole, and white balance processing is carried out on the restoration image to obtain a final restored image:
J2(x)＝f2(J1(x))
in the formula (f)2(x) For automatic white balance processing function, J1(x) For images obtained using simplified models, J2(x) Is the final restored image.
Through the analysis of experimental statistical characteristics of a large number of foggy day images, the images have the following characteristics: in the foggy area, the pixels of the image block are uniform, the gradient is small, the color tends to be milky white, and the brightness is basically the maximum in the whole image. Based on the characteristics, the invention divides the bright area and the non-bright area of the foggy day image into the gradient areas to obtain the bright area, thereby accurately estimating the brightness value of the sky. The method comprises the steps of segmenting bright areas in an image by using different color spaces, obtaining a value representing global atmosphere background light, fusing different areas to optimize a transmissivity image, optimizing a traditional NSRD model and improving the processing effect of the traditional NSRD model; the method has strong adaptability to the road environment, good algorithm stability and better processing effect, can effectively eliminate the halo phenomenon, can greatly improve the color tone and the brightness distortion of the foggy day image, obviously improve the visual effect of human eyes, can effectively recover the foggy day image for road monitoring, and enables a monitor to obtain more effective information.
Drawings
FIG. 1 is a flow chart of the present invention.
The invention is described in further detail below with reference to the figures and specific examples.
Detailed Description
As shown in fig. 1, the restoration method of a single image based on gradient field area segmentation according to the present invention is to scale an input single image to a half of the size of an original image, on one hand, convert the scaled image into an HSI color space and a gray scale space, acquire a luminance space of the image in the HSI color space, perform automatic threshold segmentation on the luminance space, perform gradient field segmentation on the gray scale space at the same time, integrate the two segmented areas to obtain a bright area and a non-bright area of the image, and acquire a value representing global atmosphere background light from the bright area; on the other hand, a dark channel image is obtained from the zoomed image, a constructed NSRAD model is used for smoothing to obtain an optimized transmittance image, the transmittance image is enlarged to the size of an original image, a multi-region theory is introduced, the obtained bright region and the obtained non-bright region are fused, and the transmittance image is further corrected; finally, on the basis of a dark primary color theoretical model, obtaining a defogged image according to the obtained value representing the global atmosphere background light and the corrected transmittance image; the method specifically comprises the following steps:
step 1, obtaining a value representing global atmosphere background light
(1) Zooming the input single image to half of the size of the original image;
(2) converting the zoomed image into an HSI color space, obtaining a brightness space of the image, performing automatic threshold processing on the brightness space, and finally obtaining a segmentation area through median filtering processing and morphological minimum area corrosion, wherein the minimum corrosion area is 1/1000 of the image through experiments;
(3) simultaneously converting the zoomed image into a gray space, converting the image into a gray image, acquiring the gradient of the gray image, performing binary segmentation by using the threshold value of the gradient, and finally obtaining a segmentation area through median filtering processing and morphological minimum area corrosion;
(4) and operating the two obtained segmentation areas to obtain a sky bright area in the image, and calculating a value A representing global atmosphere background light in the bright area:
[Am，As，Is]＝Th(I(x))
where Th (x) is the thresholding of the gradient domain and luminance space, I (x) is the input image, Am、AsThe mean value and the variance of the pixels in the bright area are obtained, and Is an image subjected to binarization after segmentation;
the method is characterized in that the brightness value A of the sky in the traditional dark primary color theory is selected from the average value of 0.1% of pixels with the maximum brightness in dark channel images, and a large number of foggy day images are processed to find that the brightness of the images is dark due to the estimation method, and for foggy scenes with different brightness, the estimation method has great effect due to inaccurate estimation of the brightness value of the sky. On the basis of experiments, a large number of foggy day images with different brightness are counted and compared with the brightness value of the actual sky, and an adaptive function which represents the value A of the global atmosphere background light is constructed and fitted:
where Δ A is a correction factor, which is set to 0.04 in this embodiment, and A is AmAnd AsThe value range of k [0,1 ]]In the embodiment, 0.4 is set, θ is a threshold for determination, and a value of 0.8 is obtained through a large number of experimental statistics;
step 2, constructing NSRAD filtering algorithm
When the synthetic aperture radar image is processed, the speckle noise is reduced, and the detail characteristics of the image are well maintained, but the current algorithms cannot have the excellent characteristics. Perona and Malik have proposed P-M model with 1990, utilize the anisotropic diffusion equation to calculate the diffusion coefficient in each direction, have kept the detail characteristic well while removing the noise, P-M model has better inhibiting effect to the additive noise in the picture, but not very good to the multiplicative noise containing the speckle in the picture. For this purpose SRAD models are proposed. Proposed in 2002 by Yongjian Yu et al, the model expression is:
in the formula (I), the compound is shown in the specification,
representing the edge of Ω, n is
F (t) is a diffusion equation, t is an input preliminary estimated transmittance image, c (q) is a diffusion coefficient, and the mathematical expression is as follows:
wherein q is a diffusion coefficient control factor calculated from the local variance, and q is0Is a defined noise variance coefficient, expressed as:
in the formula, ENL is the equivalent vision of multiplicative noise and is characterized in that the ENL is kept unchanged in a flat area; by comparing q of the sliding window with q of the flat region0To distinguish flat areas from edge areas and thus set the corresponding diffusion coefficients, q is expressed as:
the mathematical model can generate the problems of block effect, fuzzy boundary, detail loss and the like while removing noise. The main reason for this is found by mathematical analysis to be insufficient sensitivity of the diffusion function to diffusion in image homogeneity and inhomogeneity. Based on this, the invention provides a new filtering algorithm, namely, an NSRAD filtering algorithm, through mathematical experiment comparison, and the new diffusion function is as follows:
where c (q) is the diffusion coefficient, α is an adjustable integer greater than 0, controlling the descending speed of the diffusion coefficient, β is the control parameter of the homogeneous region, β is set to 10 in this embodiment,
wherein Z is a natural number, the diffusion rate is adjusted, the larger the value is, the faster the diffusion coefficient is reduced, and the control coefficient β of the homogeneous region is increased to improve the detail contrast of the transition region, in this embodiment, Z is set to 40, q is a factor for controlling the diffusion coefficient obtained by calculating the local variance, and q is a factor for controlling the diffusion coefficient obtained by calculating the local variance0Is a defined noise variance coefficient;
step 3, optimizing transmittance image
By acquiring a dark channel image J for the zoomed imaged(x) Obtaining an optimized transmittance image t through NSRAD model filtering processing2(x) (ii) a Although the effect of edge smoothing is obvious, for an image with high brightness and obvious color contrast, a halo phenomenon occurs after the filtering processing. At present, many scholars judge through tolerance, although good effects are obtained, because the scholars do not have self-adaptation, when the brightness of an image changes, the processing effects of the image have great difference;
step 4, after the optimized transmittance image is enlarged to the size of an original image, introducing a multi-region theory, fusing a bright region and a non-bright region which are obtained by dividing the image when global atmosphere background light is obtained in the step 1, respectively filtering different regions, and filtering the transmittance image t after filtering2(x) And dark channel image Jd(x) And fusing to obtain an accurately corrected transmittance image, wherein the formula is as follows:
t2(x)＝f1(Jd(x))
t3(x)＝t2(x)·(1-Is)+Is·(λ·Jd(x)+(1-λ)·t2(x))
in the formula (f)1(x) Representation of dark channel image Jd(x) Performing a filtering process, t2(x) For the filtered transmittance image, t3(x) In order to obtain a transmission image accurately corrected after fusion, λ Is a fusion coefficient, which Is set to 0.1 in this embodiment, and Is an image binarized after segmentation;
and 5, on the basis of a dark primary color theoretical model, obtaining a value A representing the global atmosphere background light and a corrected transmissivity image t according to the value A and the corrected transmissivity image3(x) Obtaining a defogged image:
the mathematical model of the dark primary color theory can be divided into two terms in theoretical simplification: the core idea of the original image item and the smoothing item is to smooth the dark channel image to obtain low-frequency information through smoothing processing without influencing the edge details of the image, namely without influencing the high-frequency information, thereby achieving the purpose of enhancing restoration.
For formulas I (x) ═ J (x) t (x) + A (1-t (x)) and
go on simplification to order I1(x) Where I (x) is the input image, j (x) is the image without fog, t (x) represents the transmittance of the medium, a is a value representing the global atmospheric background light, I (x) -ac(y) an image of one of the three RGB channels representing an input image, t1(x) Is a preliminarily estimated transmittance image, ω is a tuning parameter, generally set to 0.98, and the whole equation can be simplified as:
because the restored image obtained by the simplified model is dark in whole and needs automatic exposure processing, white balance processing is carried out on the defogged image, and the formula is as follows:
J2(x)＝f2(J1(x))
in the formula (f)2(x) For automatic white balance processing function, J1(x) For images obtained using simplified models, J2(x) Is the final restored image.
Claims (1)
1. A single image restoration method based on gradient field region segmentation is characterized in that: zooming the input single image; on one hand, the zoomed image is respectively converted into an HSI color space and a gray scale space, a brightness space of the image is obtained in the HSI color space, automatic threshold value segmentation is carried out on the brightness space, meanwhile, gradient field segmentation is carried out on the gray scale space, two segmentation areas are obtained, the two segmentation areas are integrated, a bright area and a non-bright area of the image are obtained, and a value representing global atmosphere background light is obtained according to the bright area; on the other hand, a dark channel image is obtained from the zoomed image, a constructed NSRAD model is used for smoothing to obtain an optimized transmittance image, the transmittance image is enlarged to the size of an original image, a multi-region theory is introduced, the obtained bright region and the obtained non-bright region are fused, and the transmittance image is further corrected; finally, on the basis of a dark primary color theoretical model, obtaining a defogged image according to the obtained value representing the global atmosphere background light and the corrected transmittance image;
the restoration method specifically comprises the following steps:
step 1, obtaining a value representing global atmosphere background light
(1) Zooming the input single image to half of the size of the original image;
(2) converting the zoomed image into an HSI color space, acquiring a brightness space of the image, performing automatic threshold processing on the brightness space, and finally obtaining a segmentation area through median filtering processing and morphological minimum area corrosion;
(3) simultaneously converting the zoomed image into a gray space, converting the image into a gray image, acquiring the gradient of the gray image, performing binary segmentation by using the threshold value of the gradient, and finally obtaining another segmentation region by median filtering processing and morphological minimum region corrosion;
(4) and operating the two obtained segmentation areas to obtain a sky bright area in the image, and calculating a value A representing global atmosphere background light in the bright area:
[Am，As，Is]＝Th(I(x))
where Th (x) is the thresholding of the gradient domain and luminance space, I (x) is the input image, Am、AsThe mean value and the variance of the pixels in the bright area are obtained, and Is an image subjected to binarization after segmentation;
constructing an adaptive function fitting a value A representing global atmospheric background light:
in which Δ A is a preset correction factor and A is AmAnd AsThe value range of k [0,1 ]]θ is a threshold for determination;
step 2, constructing NSRAD filtering algorithm
The expression for constructing the NSRAD model is as follows:
in the formula (I), the compound is shown in the specification,
representing the edge of Ω, n is
F (t) is a diffusion equation, t is an input preliminary estimated transmittance image, c (q) is a diffusion coefficient;
wherein the content of the first and second substances,α is an adjustable integer greater than 0, controls the rate of decrease of the diffusion coefficient,
wherein Z is a natural number, the diffusion speed is adjusted, the larger the diffusion speed is, the faster the diffusion coefficient is reduced, the control coefficient β of the homogeneous region is increased, the detail contrast of the transition region is improved, q is a factor for controlling the diffusion coefficient obtained by calculating local variance, and q is a natural number0Is the defined noise variance coefficient, β is the control parameter of the homogeneous region;
step 3, optimizing transmittance image
Obtaining a dark channel image J from an input single image after scalingd(x) Obtaining an optimized transmittance image t through NSRAD model filtering processing2(x)：
t2(x)＝f1(Jd(x))
In the formula (f)1(x) Representation of dark channel image Jd(x) Performing NSRAD model filtering processing, t2(x) The transmittance image after the filtering processing is obtained;
step 4, optimizing the transmittance image t2(x) After the image is enlarged to the size of an original image, a multi-region theory is introduced, aiming at the step 1, after the value representing the global atmosphere background light is obtained, the optimized transmissivity image t is obtained2(x) And dark channel image Jd(x) Performing multi-region fusion to obtain accurately corrected transmittance image t3(x) The formula is as follows:
t3(x)＝t2(x)·(1-Is)+Is·(λ·Jd(x)+(1-λ)·t2(x))
in the formula, t2(x) Is a transmittance image after being filtered by an NSRAD model, lambda Is a preset fusion coefficient, Is an image which Is divided and binarized, Jd(x) Is a dark channel image;
and 5, on the basis of a dark primary color theoretical model, obtaining a value A representing the global atmosphere background light and a corrected transmissivity image t according to the value A and the corrected transmissivity image3(x) Obtaining a defogged image:
for formulas I (x) ═ J (x) t (x) + A (1-t (x)) and
go on simplification to order I1(x) When the simplified model is applied, the restored image obtained by the simplified model is:
wherein I (x) is an input image, J (x) is an image in the case of no fog, t (x) represents the transmittance of the medium, A is a value representing the global atmospheric background light, Ic(y) an image of one of the three RGB channels representing an input image, t1(x) Is a preliminarily estimated transmittance image, ω is a tuning parameter;
the restoration image obtained by the simplified model is dark as a whole, and white balance processing is carried out on the restoration image to obtain a final restored image:
J2(x)＝f2(J1(x))
in the formula (f)2(x) For automatic white balance processing function, J1(x) For images obtained using simplified models, J2(x) Is the final restored image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711295623.6A CN108133462B (en) | 2017-12-08 | 2017-12-08 | Single image restoration method based on gradient field region segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711295623.6A CN108133462B (en) | 2017-12-08 | 2017-12-08 | Single image restoration method based on gradient field region segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108133462A CN108133462A (en) | 2018-06-08 |
| CN108133462B true CN108133462B (en) | 2020-06-19 |
Family
ID=62389321
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711295623.6A Active CN108133462B (en) | 2017-12-08 | 2017-12-08 | Single image restoration method based on gradient field region segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108133462B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118440B (en) * | 2018-07-06 | 2021-06-08 | 天津大学 | Single image defogging method based on transmissivity fusion and adaptive atmospheric light estimation |
| CN109348206A (en) * | 2018-11-19 | 2019-02-15 | Oppo广东移动通信有限公司 | Image white balancing treatment method, device, storage medium and mobile terminal |
| CN110175967B (en) * | 2019-06-05 | 2020-07-17 | 邓诗雨 | Image defogging processing method, system, computer device and storage medium |
| CN110298809B (en) * | 2019-07-08 | 2021-03-30 | 广东工业大学 | Image defogging method and device |
| CN116310276B (en) * | 2023-05-24 | 2023-08-08 | 泉州装备制造研究所 | Target detection method, target detection device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102243758A (en) * | 2011-07-14 | 2011-11-16 | 浙江大学 | Fog-degraded image restoration and fusion based image defogging method |
| CN103065285A (en) * | 2012-12-30 | 2013-04-24 | 信帧电子技术（北京）有限公司 | Defogging method and device for image data |
| CN103955905A (en) * | 2014-05-13 | 2014-07-30 | 北京邮电大学 | Rapid wavelet transformation and weighted image fusion single-image defogging method |
| CN106251301A (en) * | 2016-07-26 | 2016-12-21 | 北京工业大学 | A kind of single image defogging method based on dark primary priori |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI405147B (en) * | 2010-03-16 | 2013-08-11 | Novatek Microelectronics Corp | Hierarchical motion deblurring method for single image |
-
2017
- 2017-12-08 CN CN201711295623.6A patent/CN108133462B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102243758A (en) * | 2011-07-14 | 2011-11-16 | 浙江大学 | Fog-degraded image restoration and fusion based image defogging method |
| CN103065285A (en) * | 2012-12-30 | 2013-04-24 | 信帧电子技术（北京）有限公司 | Defogging method and device for image data |
| CN103955905A (en) * | 2014-05-13 | 2014-07-30 | 北京邮电大学 | Rapid wavelet transformation and weighted image fusion single-image defogging method |
| CN106251301A (en) * | 2016-07-26 | 2016-12-21 | 北京工业大学 | A kind of single image defogging method based on dark primary priori |
Non-Patent Citations (2)
| Title |
|---|
| Single image haze removal using dark channel prior;He K M等;《IEEE Transactions on Pattern Analysis and Machine Intelligence》;20111231;第33卷(第12期);第2341-2353页 * |
| 雾天交通场景中单幅图像去雾;邢晓敏等;《中国图象图形学报》;20161116;第21卷(第11期);第1440-1447页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108133462A (en) | 2018-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108133462B (en) | Single image restoration method based on gradient field region segmentation | |
| CN108876743B (en) | Image rapid defogging method, system, terminal and storage medium | |
| CN107301623B (en) | Traffic image defogging method and system based on dark channel and image segmentation | |
| Gao et al. | Sand-dust image restoration based on reversing the blue channel prior | |
| CN108765336B (en) | Image defogging method based on dark and bright primary color prior and adaptive parameter optimization | |
| CN112561804A (en) | Low-illumination underwater image enhancement method based on multi-scale detail enhancement | |
| CN110211070B (en) | Low-illumination color image enhancement method based on local extreme value | |
| CN109272475B (en) | Method for rapidly and effectively repairing and strengthening underwater image color | |
| CN112053298B (en) | Image defogging method | |
| CN111598886B (en) | Pixel-level transmittance estimation method based on single image | |
| CN108805826B (en) | Method for improving defogging effect | |
| CN105023246B (en) | A kind of image enchancing method based on contrast and structural similarity | |
| CN108765337B (en) | Single color image defogging processing method based on dark channel prior and non-local MTV model | |
| CN108898561B (en) | Defogging method, server and system for foggy image containing sky area | |
| CN110852971B (en) | Video defogging method based on dark channel prior and Retinex and storage medium | |
| CN111091501A (en) | Parameter estimation method of atmosphere scattering defogging model | |
| CN112825189B (en) | Image defogging method and related equipment | |
| CN109949239B (en) | Self-adaptive sharpening method suitable for multi-concentration multi-scene haze image | |
| CN110349113B (en) | Adaptive image defogging method based on dark primary color priori improvement | |
| CN113298730B (en) | Defogging restoration method based on image decomposition | |
| Tao et al. | Improved Retinex for low illumination image enhancement of nighttime traffic | |
| Negru et al. | Exponential image enhancement in daytime fog conditions | |
| Xie et al. | Image defogging method combining light field depth estimation and dark channel | |
| Nan et al. | Image defogging algorithm based on Fisher criterion function and dark channel prior | |
| Chandpa et al. | Comparative study of linear and non-linear contrast enhancement techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |