CN104809709A - Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition - Google Patents
Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition Download PDFInfo
- Publication number
- CN104809709A CN104809709A CN201510242413.5A CN201510242413A CN104809709A CN 104809709 A CN104809709 A CN 104809709A CN 201510242413 A CN201510242413 A CN 201510242413A CN 104809709 A CN104809709 A CN 104809709A
- Authority
- CN
- China
- Prior art keywords
- image
- mist
- transmissivity
- value
- weighted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention discloses a single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition. The method comprises the following concrete implementation steps that a foggy image is obtained, the foggy image is subjected to RGB (red green and blue) three-channel minimum value calculation to obtain a dark primary color image of the foggy image, the brightness average values of the foggy image and the dark primary color image are weighted, and the fog concentration of the foggy image is judged according to the weighted value range; a weighted quadtree decomposition method is used for determining the atmosphere light vector value through selecting the region with the most concentrated fog from the dark primary color image; a coarse transmission rate is obtained according to the dark primary color image, and the transmission rate value in a similar part to the atmosphere light value is corrected; a filter based on domain transformation is used for filtering the corrected transmission rate function, and in addition, a primary defogging image is obtained according to an atmosphere scattering model; the foggy image and the primary defogging image are transformed to an L alpha beta space, L passages of the two images are subjected to weighted fusion and are transformed back to the RGB space to obtain the final defogging image. The defogging method provided by the invention has the advantages that the image definition in foggy weather is improved, and the processing speed is high.
Description
Technical field
The present invention relates to digital image processing field, particularly relate to image restoration field, specifically, relate to a kind of based on the single image self-adaptation defogging method capable of territory conversion with cum rights quadtree decomposition.
Background technology
Image is the basis of computer vision, is main path and means that the mankind obtained and utilized visual information.And in real life, because environmental pollution is day by day serious, the greasy weather takes place frequently.By the impact of fog, it is low to there is contrast in the video that collecting device obtains and image, the problems such as color serious distortion, the many features contained in image are made to be capped or to thicken, cause video monitoring product can not collect image scene clearly, the reliability of the all-weather outdoor computer vision systems such as road traffic, outdoor monitoring system, all kinds of place safety strick precautions is produced serious influence.Therefore, research, how to having mist image and video to carry out sharpening process, reducing the negative effect that atrocious weather phenomenon is brought, having become the key issue that computer vision system is in the urgent need to address, had wide market application foreground.
At present, image mist elimination technology has become the study hotspot of image procossing and computer vision field, and researchist both domestic and external achieves certain achievement in research.Image mist elimination technology is mainly divided into two large classes: based on the method for image enhaucament and the method for physically based deformation model.
Based on the defogging method capable of image enhaucament, do not consider the degrading cause of image, but only strengthen for part interested in image.As traditional technical method, be widely used.Space difference for image place can be divided into: frequency field and spatial domain two class methods; Common method mainly contains: histogram equalization, warp wavelet, Fourier transform, wavelet transformation etc.
The method of physically based deformation model, mainly in the theoretical foundation of atmospherical scattering model, in conjunction with priori, realizes the sharpening of fog-degraded image.Early stage mist elimination algorithm mainly solves the depth of field by multiple image, and then realizes image restoration, obtains image clearly.But due to condition restriction, be difficult to the image and the video that obtain different weather situation under Same Scene.Therefore the mist elimination technology based on single image just has more practical value.
In recent years, the defogging method capable based on single image has been made significant headway.Tan (R.T.Tan, " Visibility in bad weather from asingle image; " IEEE Conference on Computer Vision and Pattern Recognition, pp.1 – 8,2008.) contrast observed without mist image is higher than the contrast with or without image, has the local contrast of mist image to realize mist elimination by maximizing.He, by using brightness value the highest in air as overall atmosphere light photograph, removes the particle in mist by the mode of the maximum local contrast of repairing image.This method visual effect is fine, but has deviation with real physics scene.Fattl (R.Fattal, " Single image dehazing; " International Conference onComputer Graphics and Interactive Techniques, pp.1 – 9,2008.) be the incoherent transfer rate estimating Misty Image by supposition transmitance and surface projection at regional area.Because the method is based on colouring information, be not suitable for gray level image, and also by distortion under thick fog condition.He Kaiming (K.He, J.Sun, and X.Tang, " Single image haze removal using dark channel prior; " IEEE Conference onComputer Vision and Pattern Recognition, pp.1957 – 1963,2009.) etc. people has delivered one section of paper about dark primary priori theoretical in CVPR meeting in 2009, propose a kind of mist elimination algorithm based on this theory, and obtain splendid mist elimination effect, create great impact at computer picture visual field.Liu Xuan (Xuan Liu, Fanxiang Zeng, Zhitong Huang, Yuefeng Ji, " Single color imagedehazing based on digital total variation filter with color transfer, " IEEE International Conference on ImageProcessing, pp.909-913, etc. Sep.2013.) people proposes a kind of realtime graphic mist elimination algorithm based on digital total variation filter, the phenomenon simultaneously enriched not for the color of image after mist elimination, color transformation model is incorporated into greasy weather coloured image sharpening field, Misty Image is made to have the colouring information with picture rich in detail, but the speed of algorithm is slower.
In a word, because most computers vision system needs to carry out real-time image procossing, said method all can not requirement of real time, therefore, in the validity not only ensureing image mist elimination but also the real-time improving mist elimination algorithm, is still a difficult problem urgently to be resolved hurrily.
Summary of the invention
The invention provides a kind of based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, this method can realize there is the sharpening process of mist image to the greasy weather fast and effectively.
In order to solve the problems of the technologies described above, specific embodiment of the invention step is:
(1) acquisition has mist image, the dark primary image that minimum value obtains mist image is asked for its RGB triple channel, to there being the average brightness value of mist image and dark primary image to be weighted, judge there is mist image fog concentration and sets mist elimination controling parameters according to weighted value scope;
(2) use cum rights quadtree decomposition method, from dark primary image, the selected the denseest region of fog, determines atmosphere light vector value;
(3) obtain rough transmissivity according to dark primary image, revise the transmittance values with air light value resemblance, obtain the transmissivity revised;
(4) use the wave filter based on territory conversion to carry out to the transmissivity revised the transmissivity that filtering is optimized, according to atmospherical scattering model, utilize the atmosphere light vector value that obtains and transmissivity to obtain preliminary mist elimination image;
(5) mist image and preliminary mist elimination image conversion will be had to L α β space, and the L passage of two width images will be weighted the L passage merging and replace preliminary mist elimination image, results conversion be returned rgb space, obtains final mist elimination image;
Further, the dark primary figure described in step (1) refers to: according to
obtain and have dark primary image D (x) of mist image, wherein I originally has mist image, and r, g, b are respectively red, green, blue three Color Channels, and c is the set of three Color Channels.
Further, described in step (1) to there being the average brightness value of mist image and dark primary image to be weighted, judge have mist image fog concentration to refer to the mean value calculating respectively and have mist image and all pixels of dark primary image, then according to S=δ * I according to weighted value scope
avg+ (1-δ) * D
avgcalculate the weighted value of fog concentration, wherein S is weighted value, I
avgfor the average brightness value of original image, D
avgfor the average brightness value of dark primary image, δ is weighting coefficient, and δ is set to 0.5.
Further, judge that having mist image fog concentration to refer to is categorized as there being the fog concentration in mist image without mist, mist, middle mist and thick fog according to formula (1) according to weighted value scope described in step (1).And according to the parameter ω of formula (1) setup control mist elimination degree.
Further, utilization cum rights quadtree decomposition method described in step (2), the selected the denseest region of fog from dark primary image also determines that atmosphere light vector value refers to the dark primary image quartern, from being followed successively by sub-block 1 ~ sub-block 4 left to bottom right, according to θ
i=α
n(D
iavg-σ
i 2), i=1,2,3,4 Hes
Calculate the fractional value of each sub-block; Wherein i is the sequence number of sub-block, θ
ithe mark of each block,
the average brightness value of each sub-block,
it is the variance of each sub-block.α
nbe used to the parameter controlling each sub-block weight, n is recurrence number of times.Then choose to be selected piece that the highest sub-block of fractional value is done as segmenting next time, repeat above-mentioned steps until the size of sub-block is less than the threshold value preset.
Further, the determination atmosphere light vector value described in step (2) refers to there is the average of each passage of R, G, B at mist picture position place as atmosphere light vector value A corresponding to the sub-block region choosing and finally determine
c.
Further, obtaining rough transmissivity according to dark primary image and refer to basis described in step (3)
calculate rough transmissivity.Wherein
for rough transmissivity figure, ω are the parameter value controlling mist elimination degree.
Further, the transmittance values of the correction described in step (3) and air light value resemblance, the transmissivity obtaining revising refers to according to formula
calculate the transmissivity revised.Wherein t ' (x) is the transmittance function revised, and K is the parameter relevant to the mean flow rate of mini-components figure.When fog concentration be mist and middle mist time, we make K=0.1D
avg, when fog concentration is thick fog, we make K=0.3D
avg.
Further, the wave filter that the utilization described in step (4) converts based on territory carries out to the transmissivity revised the transmissivity that filtering is optimized and refers to according to formula
calculate the distance between the rear neighbor pixel of conversion, wherein σ
sthe standard deviation of spatial dimension, σ
rcodomain standard deviation.σ
sbe set as 20, σ
rbe set as 0.4.Use formula J [n]=(1-a
d) I [n]+a
dthe monodimensional iterative wave filter that J [n-1] provides each row to the transmissivity figure revised carries out horizontal direction filtering, and the Output rusults of its horizontal direction filtering as input, then carries out the filtering of vertical direction to each row of the transmissivity figure revised.Transmissivity figure t (x) be finally optimized.Wherein d=ct (x
n)-ct (x
n-1) be adjacent two sampled point x in the transform domain as illustrated
nand x
n-1between distance.Equally spaced to the sampling of I [n].
it is a feedback factor.
Further, described in step (4) according to atmospherical scattering model, utilize the atmosphere light vector value that obtains and transmissivity to obtain preliminary mist elimination image and refer to basis
derive and obtain preliminary mist elimination image J
1x (), wherein, I is for there being mist image, and t is transmissivity, and A is atmosphere light vector value, t
0for constant coefficient.
Further, mist image and preliminary mist elimination image conversion will be had to refer to L α β space obtain I according to formula (2) described in step (5)
lab(x) with
Further, being weighted by the L passage of two width images described in step (5) is merged the L passage replacing preliminary mist elimination image and is referred to extract in L α β space have mist image I
lab(x) and preliminary mist elimination image
l channel components, with parameter b for weighting parameters, according to formula J
l 2(x)=b*I
l(x)+(1-b) * J
l 1x () is to preliminary restored image
keep α, β passage constant, L passage is carried out to the Weighted Fusion of image.Wherein I
lx () is for having the L channel components of mist image, J
l 1x L channel components that () is preliminary restored image, J
l 2x L channel components that () is final restored image.Finally by J
l 2(x) component and J
a 1(x) component and J
b 1x () component merges, obtain
Further, described in step (5), results conversion is returned rgb space, obtain final mist elimination image and refer to, according to formula (2), image is returned rgb space from L α β spatial alternation, obtain final mist elimination image J
2(x).
The present invention has following advantage: the present invention adopts based on the single image self-adaptation defogging method capable of territory conversion with cum rights quadtree decomposition, and the method can identify fog concentration in mist image, realizes self-adaptation mist elimination; Restored image contrast is high, clear picture, and details is enriched; This method computation complexity is low simultaneously, and fast operation, can meet current most computer vision system to the requirement of real-time.
Accompanying drawing explanation
Fig. 1 is the implementing procedure schematic diagram of the embodiment of the present invention;
Fig. 2 be input have mist image;
Fig. 3 is the dark primary image of Fig. 2;
Fig. 4 is quadtree decomposition method schematic diagram;
Fig. 5 is the rough transmissivity image of Fig. 2;
Fig. 6 is the correction transmissivity image of Fig. 2;
Fig. 7 is the optimization transmissivity image of Fig. 2;
Fig. 8 is the preliminary mist elimination image of Fig. 2;
Fig. 9 is the final mist elimination image of Fig. 2;
Embodiment
In order to better the present invention is described, referring to drawings and Examples, further detailed description is done to specific embodiment of the invention.
As shown in Figure 1, specific embodiment of the invention step is:
(1) acquisition has mist image, the dark primary image that minimum value obtains mist image is asked for its RGB triple channel, to there being the average brightness value of mist image and dark primary image to be weighted, judge there is mist image fog concentration and sets mist elimination controling parameters according to weighted value scope;
Read in a width have mist image and be transformed into RGB color space, be designated as I.Fig. 2 is for there being mist image in the present embodiment, and pixel size is 600 × 400.
According to
obtain and have dark primary image D (x) of mist image, wherein I originally has mist image, and r, g, b are respectively red, green, blue three Color Channels, and c is the set of three Color Channels.Fig. 3 is the dark primary image having mist image graph 2.
Calculate the mean value having mist image graph 2 and all pixels of dark primary image graph 3 respectively, then according to S=δ * I
avg+ (1-δ) * D
avgcalculate the weighted value of fog concentration, wherein S is weighted value, I
avgfor the average brightness value of original image, D
avgfor the average brightness value of dark primary image, δ is weighting coefficient, and δ is set to 0.5.I in the present embodiment
avg=168, D
avg=156, S=164.
Judge that the fog concentration of Fig. 2 is thick fog according to formula (1), and according to parameter ω=0.95 of formula (1) setup control mist elimination degree.
(2) use cum rights quadtree decomposition method, from dark primary image, the selected the denseest region of fog, determines atmosphere light vector value;
By the dark primary image quartern, from being followed successively by sub-block 1 ~ sub-block 4 left to bottom right, according to formula
and formula
Calculate the fractional value of each sub-block; Wherein i is the sequence number of sub-block, θ
ithe mark of each block,
the average brightness value of each sub-block,
it is the variance of each sub-block.α
nbe used to the parameter controlling each sub-block weight, n is recurrence number of times.Then choose to be selected piece that the highest sub-block of fractional value is done as segmenting next time, repeat above-mentioned steps until the size of sub-block is less than the threshold value preset.In the present embodiment, Fig. 3 is cum rights quadtree decomposition schematic diagram, and the sequence number number of sub-block when numeral 1 ~ 4 is for carrying out ground floor quadtree decomposition in figure, the sequence number number of each layer omits in figure 3 afterwards.
Choosing has the average of each passage of R, G, B at mist picture position place as atmosphere light vector value A corresponding to the sub-block region finally determined
c.In the present embodiment, Fig. 3 elliptic region is that atmosphere light vector value A is chosen in the denseest region of fog finally determined namely
cregion.
(3) obtain rough transmissivity according to dark primary image, revise the transmittance values with air light value resemblance, obtain the transmissivity revised;
According to
derive and calculate rough transmissivity.Wherein
for rough transmissivity figure, ω are the parameter value controlling mist elimination degree.Fig. 5 is rough transmissivity image in the present embodiment.
Then basis
calculate the transmissivity revised.Wherein t ' (x) is the transmittance function revised, and K is the parameter relevant to the mean flow rate of mini-components figure.When fog concentration be mist and middle mist time, we make K=0.1D
avg, when fog concentration is thick fog, we make K=0.3D
avg.In the present embodiment, K=0.1D
avg, Fig. 6 is for revising transmissivity image.
(4) use the wave filter based on territory conversion to carry out to the transmissivity revised the transmissivity that filtering is optimized, according to atmospherical scattering model, utilize the atmosphere light vector value that obtains and transmissivity to obtain preliminary mist elimination image;
According to
calculate the distance between the rear neighbor pixel of conversion, wherein σ
sthe standard deviation of spatial dimension, σ
rcodomain standard deviation.σ
sbe set as 20, σ
rbe set as 0.4.Use formula J [n]=(1-a
d) I [n]+a
dthe monodimensional iterative wave filter that J [n-1] provides each row to the transmissivity figure revised carries out horizontal direction filtering, and the Output rusults of its horizontal direction filtering as input, then carries out the filtering of vertical direction to each row of the transmissivity figure revised.Transmissivity figure t (x) be finally optimized.Wherein d=ct (x
n)-ct (x
n-1) be adjacent two sampled point x in the transform domain as illustrated
nand x
n-1between distance.Equally spaced to the sampling of I [n].
it is a feedback factor.In the present embodiment, a=0.932, Fig. 7 are for optimizing transmissivity image.
According to
derive and obtain preliminary mist elimination image J
1x (), wherein, I is for there being mist image, and t is transmissivity, and A is atmosphere light vector value, t
0for constant coefficient.In the present embodiment, t
0=0.1 is constant, and the preliminary mist elimination image of Fig. 2 as shown in Figure 8.
(5) mist image and preliminary mist elimination image conversion will be had to L α β space, and the L passage of two width images will be weighted the L passage merging and replace preliminary mist elimination image, results conversion be returned rgb space, obtains final mist elimination image;
Mist image I (x) and preliminary mist elimination image J has been obtained according to formula (2)
1(x) value I in L α β space
lab(x) with
extract in L α β space afterwards and have mist image I
lab(x) and preliminary mist elimination image
l channel components, with parameter b for going weight, according to formula J
l 2(x)=b*I
l(x)+(1-b) * J
l 1x () is to preliminary restored image
keep α, β passage constant, L passage is carried out to the Weighted Fusion of image.Wherein I
lx () is for having the L channel components of mist image, J
l 1x L channel components that () is preliminary restored image, J
l 2x L channel components that () is final restored image.Finally by J
l 2(x) component and J
a 1(x) component and J
b 1x () component merges, obtain
again according to formula (2), image after merging is returned rgb space from L α β spatial alternation, obtain final mist elimination image J
2(x).In the present embodiment, weighting parameters b=0.3, final mist elimination image as shown in Figure 9.
Mist image is had compared with the result images after mist elimination through method process of the present invention:
Original mist picture contrast is low, cross-color, and scenery details composition less as shown in Figure 2; The dark primary image obtained by mini-value filtering as shown in Figure 3; Cum rights quadtree decomposition method is adopted to find the schematic diagram in skylight region as shown in Figure 4; Rough transmissivity figure as shown in Figure 5; The transmissivity figure repaired the scene close to atmosphere light value place as shown in Figure 6; Adopt territory conversion Edge preservation wave filter to the optimization transmissivity figure obtained after the transmissivity figure smoothing details preserving edge filtering revised as shown in Figure 7; Derive the preliminary mist elimination image that obtains as shown in Figure 8 according to atmospherical scattering model, and as can be seen from the figure part scene color there will be distortion phenomenon, and preliminary restored image is slightly dark, therefore needs to adopt the method for image co-registration to make up; Final mist elimination image as shown in Figure 9;
The hardware environment that the present embodiment runs is the desktop computer of configuration 3.60GHz Intel (R) Xeon (R) E5-1620CPU and 8G internal memory, and by software MATLA programming realization, the Color Image Processing time for this embodiment is 0.49s.Can find out that from this example can obtain details fast and effectively based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition enriches, contrast improves, the natural mist elimination figure of color recovery.
Claims (6)
1., based on territory conversion and a single image self-adaptation defogging method capable for cum rights quadtree decomposition, it is characterized in that the concrete steps of the method are:
Step (1), acquisition has mist image, the dark primary image that minimum value obtains mist image being asked for its RGB triple channel, to there being the average brightness value of mist image and dark primary image to be weighted, judging there is mist image fog concentration and sets mist elimination controling parameters according to weighted value scope;
Step (2), uses cum rights quadtree decomposition method, and atmosphere light vector value is also determined in the selected the denseest region of fog from dark primary image;
Step (3), obtains rough transmissivity according to dark primary image, revises the transmittance values with air light value resemblance, obtains the transmissivity revised;
Step (4), uses the wave filter based on territory conversion to carry out to the transmissivity revised the transmissivity that filtering is optimized, and according to atmospherical scattering model, utilizes the transmissivity of atmosphere light vector value and the optimization obtained to obtain preliminary mist elimination image;
Step (5), will have mist image and preliminary mist elimination image conversion to L α β space, the L passage of two width images will be weighted the L passage merging and replace preliminary mist elimination image, and convert back rgb space, obtain final mist elimination image.
2. according to right 1 based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, it is characterized in that obtaining in described step (1) have mist image, the dark primary image that minimum value obtains mist image is asked for its RGB triple channel, be weighted there being the average brightness value of mist image and dark primary image, judge there is mist image fog concentration according to weighted value scope, concrete steps are as follows:
S1.1: the dark primary image D obtaining Misty Image according to formula (1).
Wherein I originally has mist image, and r, g, b are respectively red, green, blue three Color Channels, and c is the set of three Color Channels.
S1.2: calculate the average brightness value having mist image and dark primary image according to formula (2).
S＝δ*I
avg+(1-δ)*D
avg(2)
Wherein S is weighted value, I
avgfor the average brightness value of original image, D
avgfor the average brightness value of dark primary image, δ is weighting coefficient, and δ is set to 0.5.
S1.3: judge the fog concentration had in mist image according to formula (3), fog concentration is set as successively without mist, mist, middle mist and thick fog.And setting can control the parameter ω of mist elimination concentration.
。
3. according to right 1 based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, it is characterized in that described step (2) uses cum rights quadtree decomposition method, atmosphere light vector value is also determined in the selected the denseest region of fog from dark primary image, and concrete steps are as follows:
S2.1: by the dark primary image quartern, from top to bottom, be from left to right followed successively by sub-block 1 ~ sub-block 4, calculates the fractional value of each sub-block according to formula (4) and formula (5);
θ
i＝α
n(D
iavg-σ
i 2)i＝1,2,3,4 (4)
Wherein i is the sequence number of sub-block, θ
ithe mark of each block,
the average brightness value of each sub-block,
it is the variance of each sub-block.α
nbe used to the parameter controlling each sub-block weight, n is recurrence number of times.
S2.2: choose to be selected piece that the highest sub-block of fractional value is done as segmenting next time, repeats step S2.1, until the size of sub-block is less than the threshold value preset.
S2.3: choose the sub-block finally determined corresponding have the three-channel average of RGB at mist picture position place as atmosphere light vector value A
c.
4. according to right 1 based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, it is characterized in that described step (3) obtains rough transmissivity according to dark primary image, revise the transmittance values with air light value resemblance, obtain the transmissivity revised, concrete steps are as follows:
S3.1: calculate rough transmissivity figure according to formula (6)
Wherein
for the parameter value that rough transmissivity figure, ω are the control mist elimination degree set in step S1.3.
S3.2: calculate the transmittance function revised according to formula (7)
Wherein t ' (x) is the transmittance function revised, and K is the parameter relevant to the mean flow rate of mini-components figure.When fog concentration be mist and middle mist time, we make K=0.1D
avg, when fog concentration is thick fog, we make K=0.3D
avg.
5. according to right 1 based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, it is characterized in that described step (4) uses the wave filter based on territory conversion to carry out to the transmissivity revised the transmissivity that filtering is optimized, according to atmospherical scattering model, utilize the transmissivity of atmosphere light vector value and the optimization obtained to obtain preliminary mist elimination image, concrete steps are as follows:
S4.1: calculate the distance between the rear neighbor pixel of conversion according to formula (8), monodimensional iterative wave filter each row to the transmissivity figure revised using formula (9) to provide carries out horizontal direction filtering, the Output rusults of its horizontal direction filtering as input, then carries out the filtering of vertical direction to each row of the transmissivity figure revised.Finally, our transmissivity figure t (x) of being optimized.
Wherein σ
sthe standard deviation of spatial dimension, σ
rcodomain standard deviation.σ
sbe set as 20, σ
rbe set as 0.4.
J[n]＝(1-a
d)I[n]+a
dJ[n-1], (9)
Wherein d=ct (x
n)-ct (x
n-1) be adjacent two sampled point x in the transform domain as illustrated
nand x
n-1between distance.Equally spaced to the sampling of I [n].
it is a feedback factor.
S4.2: according to atmospherical scattering model, derives and obtains formula (10), obtain preliminary mist elimination image J
1.
Wherein, I is for there being mist image, and t is transmissivity, and A is atmosphere light vector value, t
0for constant coefficient.
6. according to right 1 based on territory conversion and the single image self-adaptation defogging method capable of cum rights quadtree decomposition, it is characterized in that described step (5) will have mist image and preliminary mist elimination image conversion to L α β space, the L passage of two width images is weighted the L passage merging and replace preliminary mist elimination image, and convert back rgb space, obtain final mist elimination image, concrete steps are as follows:
S5.1: mist image and preliminary mist elimination image conversion will be had to L α β space, and obtain I
lab(x) with
S5.2: extract the L channel components having mist image and preliminary mist elimination image, according to formula (11), with parameter b for weight, to preliminary restored image J
1x () keeps α, β passage constant, L passage is carried out to the Weighted Fusion of image.
J
l 2(x)＝b*I
l(x)+(1-b)*J
l 1(x) (11)
Wherein I
lx () is for having the L channel components of mist image, J
l 1x L channel components that () is preliminary restored image, J
l 2x L channel components that () is final restored image.
S5.3: by J
l 2(x) component and J
a 1(x) component and J
b 1x () component merges, switch back to rgb space, obtain final mist elimination image J
2(x).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510242413.5A CN104809709A (en) | 2015-05-13 | 2015-05-13 | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510242413.5A CN104809709A (en) | 2015-05-13 | 2015-05-13 | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104809709A true CN104809709A (en) | 2015-07-29 |
Family
ID=53694514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510242413.5A Pending CN104809709A (en) | 2015-05-13 | 2015-05-13 | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809709A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105096272A (en) * | 2015-08-19 | 2015-11-25 | 常州工学院 | De-hazing method based on dual-tree complex wavelet |
| CN105872373A (en) * | 2016-03-31 | 2016-08-17 | 北京奇虎科技有限公司 | Automatic defogging photographing method, device and equipment |
| CN105913390A (en) * | 2016-04-07 | 2016-08-31 | 潍坊学院 | Image defogging method and system |
| CN106355560A (en) * | 2016-08-30 | 2017-01-25 | 潍坊学院 | Method and system for extracting atmospheric light value in haze image |
| CN107451962A (en) * | 2017-07-03 | 2017-12-08 | 山东财经大学 | A kind of image defogging method and device |
| CN108109129A (en) * | 2017-12-15 | 2018-06-01 | 四川大学 | A kind of rapid image defogging method based on near-infrared |
| CN108154482A (en) * | 2017-10-24 | 2018-06-12 | 天津大学 | Image Blind deblurring method based on dark channel prior and multi-direction Weighted T V |
| CN108717686A (en) * | 2018-04-04 | 2018-10-30 | 华南理工大学 | A kind of real-time video defogging method based on dark channel prior |
| CN110310223A (en) * | 2019-07-03 | 2019-10-08 | 云南电网有限责任公司电力科学研究院 | A kind of fusion method of ultraviolet light and visible images |
| CN111091501A (en) * | 2018-10-24 | 2020-05-01 | 天津工业大学 | Parameter estimation method of atmosphere scattering defogging model |
| CN113298732A (en) * | 2021-06-08 | 2021-08-24 | 北京联合大学 | Image defogging method and system based on regional similarity |
| CN114494084A (en) * | 2022-04-14 | 2022-05-13 | 广东欧谱曼迪科技有限公司 | Image color homogenizing method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
-
2015
- 2015-05-13 CN CN201510242413.5A patent/CN104809709A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
Non-Patent Citations (3)
| Title |
|---|
| ZHENWEI SHI等: "Single image dehazing in inhomogeneous atmosphere", 《OPTIK》 * |
| 王永虎: "基于大气散射模型的单幅图像去雾方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 许丽红: "单幅雾天图像清晰化技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105096272B (en) * | 2015-08-19 | 2018-02-02 | 常州工学院 | A kind of defogging method based on dual-tree complex wavelet |
| CN105096272A (en) * | 2015-08-19 | 2015-11-25 | 常州工学院 | De-hazing method based on dual-tree complex wavelet |
| CN105872373A (en) * | 2016-03-31 | 2016-08-17 | 北京奇虎科技有限公司 | Automatic defogging photographing method, device and equipment |
| CN105913390B (en) * | 2016-04-07 | 2019-11-22 | 潍坊学院 | A kind of image defogging method and system |
| CN105913390A (en) * | 2016-04-07 | 2016-08-31 | 潍坊学院 | Image defogging method and system |
| CN106355560A (en) * | 2016-08-30 | 2017-01-25 | 潍坊学院 | Method and system for extracting atmospheric light value in haze image |
| CN106355560B (en) * | 2016-08-30 | 2019-03-01 | 潍坊学院 | The extracting method and system of air light value in a kind of haze image |
| CN107451962A (en) * | 2017-07-03 | 2017-12-08 | 山东财经大学 | A kind of image defogging method and device |
| CN107451962B (en) * | 2017-07-03 | 2020-05-05 | 山东财经大学 | Image defogging method and device |
| CN108154482A (en) * | 2017-10-24 | 2018-06-12 | 天津大学 | Image Blind deblurring method based on dark channel prior and multi-direction Weighted T V |
| CN108109129A (en) * | 2017-12-15 | 2018-06-01 | 四川大学 | A kind of rapid image defogging method based on near-infrared |
| CN108717686A (en) * | 2018-04-04 | 2018-10-30 | 华南理工大学 | A kind of real-time video defogging method based on dark channel prior |
| CN108717686B (en) * | 2018-04-04 | 2022-02-01 | 华南理工大学 | Real-time video defogging method based on dark channel prior |
| CN111091501A (en) * | 2018-10-24 | 2020-05-01 | 天津工业大学 | Parameter estimation method of atmosphere scattering defogging model |
| CN110310223A (en) * | 2019-07-03 | 2019-10-08 | 云南电网有限责任公司电力科学研究院 | A kind of fusion method of ultraviolet light and visible images |
| CN110310223B (en) * | 2019-07-03 | 2023-04-07 | 云南电网有限责任公司电力科学研究院 | Fusion method of ultraviolet light and visible light image |
| CN113298732A (en) * | 2021-06-08 | 2021-08-24 | 北京联合大学 | Image defogging method and system based on regional similarity |
| CN114494084A (en) * | 2022-04-14 | 2022-05-13 | 广东欧谱曼迪科技有限公司 | Image color homogenizing method and device, electronic equipment and storage medium |
| CN114494084B (en) * | 2022-04-14 | 2022-07-26 | 广东欧谱曼迪科技有限公司 | Image color homogenizing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104809709A (en) | Single-image self-adaptation defogging method based on domain transformation and weighted quadtree decomposition | |
| CN102831591B (en) | Gaussian filter-based real-time defogging method for single image | |
| CN107301623B (en) | Traffic image defogging method and system based on dark channel and image segmentation | |
| CN102063706B (en) | Rapid defogging method | |
| CN103218778B (en) | The disposal route of a kind of image and video and device | |
| CN106846263A (en) | The image defogging method being immunized based on fusion passage and to sky | |
| Gao et al. | Sand-dust image restoration based on reversing the blue channel prior | |
| CN105574830A (en) | Low-quality image enhancement method under extreme weather conditions | |
| CN105787904B (en) | For the image defogging method of the adaptive global dark primary priori of bright areas | |
| CN107358585B (en) | Foggy day image enhancement method based on fractional order differential and dark channel prior | |
| CN106157267A (en) | A kind of image mist elimination absorbance optimization method based on dark channel prior | |
| CN106251301A (en) | A kind of single image defogging method based on dark primary priori | |
| CN103020917B (en) | Method for restoring ancient Chinese calligraphy and painting images on basis of conspicuousness detection | |
| CN105023256B (en) | A kind of image defogging method and system | |
| CN105913390B (en) | A kind of image defogging method and system | |
| CN103955905A (en) | Rapid wavelet transformation and weighted image fusion single-image defogging method | |
| CN104318524A (en) | Method, device and system for image enhancement based on YCbCr color space | |
| CN107330871A (en) | The image enchancing method of insulator automatic identification is run under bad weather condition | |
| CN103985091A (en) | Single image defogging method based on luminance dark priori method and bilateral filtering | |
| CN104253930A (en) | Real-time video defogging method | |
| CN104036466A (en) | Video defogging method and system | |
| CN103198459A (en) | Haze image rapid haze removal method | |
| CN103914820A (en) | Image haze removal method and system based on image layer enhancement | |
| CN104867121A (en) | Fast image defogging method based on dark channel prior and Retinex theory | |
| CN108447068A (en) | Ternary diagram automatic generation method and the foreground extracting method for utilizing the ternary diagram |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150729 |