CN103020921A - Single image defogging method based on local statistical information - Google Patents
Single image defogging method based on local statistical information Download PDFInfo
- Publication number
- CN103020921A CN103020921A CN2013100084869A CN201310008486A CN103020921A CN 103020921 A CN103020921 A CN 103020921A CN 2013100084869 A CN2013100084869 A CN 2013100084869A CN 201310008486 A CN201310008486 A CN 201310008486A CN 103020921 A CN103020921 A CN 103020921A
- Authority
- CN
- China
- Prior art keywords
- image
- value
- secretly
- window
- dark
- 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
Images
Abstract
The invention relates to image processing and discloses a single image defogging method based on local statistical information. The single image defogging method based on the local statistical information is capable of rapidly defogging single-frame images and improving visual effects of images shot under severe weather conditions such as foggy days. The method includes: using a minimum value of three RGB channels of an acquired color image as a pixel value of a current point so that a gray level image which is called as 'dark image' is obtained; performing mathematical modeling for the dark image, and assuming that a pixel gray level value is accordant with Gaussian distribution and is also accordant with the same statistical law within a large rectangular window; in a filtering window, subtracting triple variance from a mean value to be statistic into a minimum value in the window so as to serve as a value of a current pixel point, and filtering so that a dark channel image is obtained; and restoring a scene image according to the atmosphere illumination intensity and the dark channel image. The single image defogging method based on the local statistical information solves the problem of incapability of defogging at a dense staggered position of a distant view and a nearby view, improves dark channel getting precision and improves image processing speed by getting the dark channel from the perspective of statistics.
Description
Technical field
The present invention relates to image and process, especially relate to the single image defogging method capable based on local statistic information.
Background technology
In safety-security area, abominable extreme weather causes image quality decrease to make troubles to monitoring all the time.Utilize computer vision technique, it is the focus of current image scientific research field that single frames is had mist image Fast Restoration.
The mist elimination problem is a challenging problem in the image information field, and the range information that collects the concentration of mist in the image and scene is relevant.In the mist elimination algorithm of classics, Mr.'s He Kaiming the single image mist elimination algorithm based on helping priori secretly utilizes the characteristic of helping secretly to estimate accurately scenery to the range information between the image acquisition device well, but pilot process has adopted the method for stingy figure, has consumed widely computer processing time.
Chinese patent 201010139441.1 discloses a kind of automated graphics defogging method capable based on dark primary, the method utilizes dark primary priori to ask for transmitting image, multiple dimensioned Retinex asks for the luminance component image, its processing speed is slow, the lower threshold of transition function can not dynamic self-adapting, the sky dummy section after the processing have halation.
Summary of the invention
The object of the present invention is to provide and to realize quick single-frame images mist elimination, improve the single image defogging method capable based on local statistic information of institute's pickup image visual effect under the severe weather conditions such as greasy weather.
The present invention includes following steps:
1) coloured image that collects is got minimum value in three passages of RGB as the pixel value of current point, obtained a width of cloth gray level image, be referred to as " dark image ";
2) should dark image mathematical modeling, suppose that its grey scale pixel value meets Gaussian distribution, and in larger rectangular window, also meet same statistical law;
3) in filter window, subtract three times of variances with average and count on minimum value in the window as the value of current pixel point, obtain helping secretly image after the filtering;
4) according to the atmosphere intensity of illumination with help the image restoration scene image secretly.
In step 1), the described minimum value that the coloured image that collects is got in three passages of RGB can be as the concrete grammar of the pixel value of current point:
Ask for the most current pixel value of minimum value of three passages of RGB in the coloured image, obtain " dark image ".
Wherein, I
c(x, y) is input color image, " the dark image " of D (x, y) for obtaining.
In step 2) in, described should dark image mathematical modeling, suppose that its grey scale pixel value meets Gaussian distribution, and the concrete grammar that also meets same statistical law in larger rectangular window can be:
Approach with quick two-sided filter and to ask for sample point expectation and variance, effectively protect the marginal information of image simultaneously, prevent at far and near scape intersection owing to edge fog produces " white sideband " effect;
Wherein, Bilateral is quick bilateral filtering operator, and μ (x, y) and σ (x, y) are respectively Mean Matrix and variance matrix.
In step 3), described in filter window, subtract three times of variances with average and count on minimum value in the window as the value of current pixel point, the concrete grammar that obtains helping secretly image after the filtering can be:
According to the probability statistics principle, estimate helping secretly of image
I
min(x,y)=μ(x,y)-3σ(x,y)
I
DarkChannel(x,y)≈min(I
min(x,y),D(x,y))
Wherein, I
DarkChannel(x, y) is the image of helping secretly that finally estimates.
In step 4), describedly can be according to atmosphere intensity of illumination and the concrete grammar of helping the image restoration scene image secretly:
Estimation atmosphere light intensity, image restoration atmosphere light intensity A
cIt is the maximum pixel value of being helped secretly the coordinate in the front 0.1% corresponding original image of bright spot in the image by statistics.
Wherein, t (x, y) is the atmospheric dissipation function, and it is only relevant with the depth information of scene, R (x, y) be after restoring without the mist image.
The present invention sets up the Gaussian distribution mathematical model with the gray-scale value of image pixel, utilizing brand-new method to estimate faster and more accurately helps secretly, mist elimination for nearly distant view intersection has good effect, adopt simultaneously the method for quick bilateral filtering greatly to improve computing speed, make real-time mist elimination become possibility.
The present invention improved in the classical mist elimination algorithm the intensive staggered place of far and near scape can't mist elimination problem, improved and helped the precision of asking for secretly, abandon simultaneously scratching the method for figure and using instead to ask for from statistical angle and help secretly with image, improved image processing speed.
Description of drawings
Fig. 1 is the process flow diagram of the embodiment of the invention.
Fig. 2 is one of experimental result of the embodiment of the invention.
Fig. 3 be the embodiment of the invention experimental result two.
In Fig. 2 and 3, (a) for the former figure of mist is arranged, (b) be the result after the present invention's processing.
Embodiment
Following examples, the present invention is further illustrated in connection with accompanying drawing.
Present embodiment has the mist image for colour, and its recuperation comprises following 4 steps:
Step 1, the coloured image that collects is got minimum value in three passages of RGB as the pixel value of current point, obtains a width of cloth gray level image, be referred to as " dark image ":
Coloured image has three passages of RGB, and each passage has respectively a gray-scale value, goes the minimum value of three passages of RGB as current some pixel value, obtains dark image D (x, y).
Step 2, should dark image mathematical modeling, suppose that its grey scale pixel value meets Gaussian distribution, and in larger rectangular window, also meet same statistical law:
Grey scale pixel value in window is met Gaussian distribution as a priori conditions.
Step 3, in filter window, subtract three times of variances with average and count on minimum value in the window as the value of current pixel point, obtain helping secretly image after the filtering:
Use quick two-sided filter, approach fast and effectively the process of asking sample point expectation and variance, obtain respectively expected matrix and variance matrix, then subtract three times variance matrix with expected matrix and obtain helping secretly image I
DarkChannel(x, y).
Step 4, according to the atmosphere intensity of illumination with help the image restoration scene image secretly:
At first calculate dissipative function t (x, y) according to the image of helping secretly that obtained, statistics is helped in the image maximum pixel value of the coordinate in the front 0.1% corresponding original image of bright spot secretly as atmosphere light intensity value A.Threshold value t is set again
0To prevent denominator as zero situation, the while can effectively be controlled the mist elimination degree of distant view, keeps sky to get the validity of infinite distance scenery.
Claims (5)
1. based on the single image defogging method capable of local statistic information, it is characterized in that may further comprise the steps:
1) coloured image that collects is got minimum value in three passages of RGB as the pixel value of current point, obtained a width of cloth gray level image, be referred to as " dark image ";
2) should dark image mathematical modeling, suppose that its grey scale pixel value meets Gaussian distribution, and in larger rectangular window, also meet same statistical law;
3) in filter window, subtract three times of variances with average and count on minimum value in the window as the value of current pixel point, obtain helping secretly image after the filtering;
4) according to the atmosphere intensity of illumination with help the image restoration scene image secretly.
2. as claimed in claim 1 based on the single image defogging method capable of local statistic information, it is characterized in that in step 1) that the described minimum value that the coloured image that collects is got in three passages of RGB as the concrete grammar of the pixel value of current point is:
Ask for the most current pixel value of minimum value of three passages of RGB in the coloured image, obtain " dark image " D (x, y) and be:
Wherein, I
c(x, y) is input color image.
3. as claimed in claim 1 based on the single image defogging method capable of local statistic information, it is characterized in that in step 2) in, described should dark image mathematical modeling, suppose that its grey scale pixel value meets Gaussian distribution, and the concrete grammar that also meets same statistical law in larger rectangular window is:
Approach with quick two-sided filter and to ask for sample point expectation and variance, effectively protect the marginal information of image simultaneously, prevent at far and near scape intersection owing to edge fog produces " white sideband " effect;
Wherein, Bilateral is quick bilateral filtering operator, and μ (x, y) and σ (x, y) are respectively Mean Matrix and variance matrix.
4. as claimed in claim 1 based on the single image defogging method capable of local statistic information, it is characterized in that in step 3), described in filter window, subtract three times of variances with average and count on minimum value in the window as the value of current pixel point, the concrete grammar that obtains helping secretly image after the filtering is:
According to the probability statistics principle, estimate helping secretly of image:
I
min(x,y)=μ(x,y)-3σ(x,y)
I
DarkChannel(x,y)≈min(I
min(x,y),D(x,y))
Wherein, I
DarkChannel(x, y) is the image of helping secretly that finally estimates.
5. as claimed in claim 1 based on the single image defogging method capable of local statistic information, it is characterized in that in step 4), describedly according to atmosphere intensity of illumination and the concrete grammar of helping the image restoration scene image secretly be:
Estimation atmosphere light intensity, image restoration atmosphere light intensity A
cThe maximum pixel value of being helped secretly the coordinate in the front 0.1% corresponding original image of bright spot in the image by statistics,
Wherein, t (x, y) is the atmospheric dissipation function, and it is only relevant with the depth information of scene, R (x, y) be after restoring without the mist image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013100084869A CN103020921A (en) | 2013-01-10 | 2013-01-10 | Single image defogging method based on local statistical information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013100084869A CN103020921A (en) | 2013-01-10 | 2013-01-10 | Single image defogging method based on local statistical information |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103020921A true CN103020921A (en) | 2013-04-03 |
Family
ID=47969490
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013100084869A Pending CN103020921A (en) | 2013-01-10 | 2013-01-10 | Single image defogging method based on local statistical information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103020921A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103198459A (en) * | 2013-04-10 | 2013-07-10 | 成都国腾电子技术股份有限公司 | Haze image rapid haze removal method |
CN103489166A (en) * | 2013-10-12 | 2014-01-01 | 大连理工大学 | Bilateral filter-based single image defogging method |
CN104217404A (en) * | 2014-08-27 | 2014-12-17 | 华南农业大学 | Video image sharpness processing method in fog and haze day and device thereof |
CN104299198A (en) * | 2014-10-14 | 2015-01-21 | 嘉应学院 | Fast image defogging method based on dark channels of pixels |
CN107133927A (en) * | 2017-04-21 | 2017-09-05 | 汪云飞 | Single image to the fog method based on average mean square deviation dark under super-pixel framework |
CN108537737A (en) * | 2017-03-03 | 2018-09-14 | 防城港市港口区思达电子科技有限公司 | A kind of improved image defogging method |
CN109685725A (en) * | 2018-11-21 | 2019-04-26 | 南京航空航天大学 | A kind of car surface image based on dark channel prior removes dust collecting method |
CN114881896A (en) * | 2022-07-12 | 2022-08-09 | 广东欧谱曼迪科技有限公司 | Endoscope image real-time defogging method and device, electronic equipment and storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101783012A (en) * | 2010-04-06 | 2010-07-21 | 中南大学 | Automatic image defogging method based on dark primary colour |
-
2013
- 2013-01-10 CN CN2013100084869A patent/CN103020921A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101783012A (en) * | 2010-04-06 | 2010-07-21 | 中南大学 | Automatic image defogging method based on dark primary colour |
Non-Patent Citations (1)
Title |
---|
WENBO JIN等: "《Computer Science and Automation Engineering (CSAE), 2012 IEEE International Conference on 》", 25 May 2012, article "《Single image de-haze based on a new dark channel estimation method》", pages: 791-795 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103198459A (en) * | 2013-04-10 | 2013-07-10 | 成都国腾电子技术股份有限公司 | Haze image rapid haze removal method |
CN103198459B (en) * | 2013-04-10 | 2016-02-03 | 成都国翼电子技术有限公司 | Haze image rapid haze removal method |
CN103489166A (en) * | 2013-10-12 | 2014-01-01 | 大连理工大学 | Bilateral filter-based single image defogging method |
CN104217404A (en) * | 2014-08-27 | 2014-12-17 | 华南农业大学 | Video image sharpness processing method in fog and haze day and device thereof |
CN104217404B (en) * | 2014-08-27 | 2017-06-20 | 华南农业大学 | Haze sky video image clearness processing method and its device |
CN104299198A (en) * | 2014-10-14 | 2015-01-21 | 嘉应学院 | Fast image defogging method based on dark channels of pixels |
CN108537737A (en) * | 2017-03-03 | 2018-09-14 | 防城港市港口区思达电子科技有限公司 | A kind of improved image defogging method |
CN107133927A (en) * | 2017-04-21 | 2017-09-05 | 汪云飞 | Single image to the fog method based on average mean square deviation dark under super-pixel framework |
CN107133927B (en) * | 2017-04-21 | 2020-03-17 | 汪云飞 | Single image defogging method based on mean-square error dark channel under super-pixel frame |
CN109685725A (en) * | 2018-11-21 | 2019-04-26 | 南京航空航天大学 | A kind of car surface image based on dark channel prior removes dust collecting method |
CN114881896A (en) * | 2022-07-12 | 2022-08-09 | 广东欧谱曼迪科技有限公司 | Endoscope image real-time defogging method and device, electronic equipment and storage medium |
CN114881896B (en) * | 2022-07-12 | 2022-10-04 | 广东欧谱曼迪科技有限公司 | Endoscope image real-time defogging method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103020921A (en) | Single image defogging method based on local statistical information | |
KR101448164B1 (en) | Method for Image Haze Removal Using Parameter Optimization | |
CN103218778B (en) | The disposal route of a kind of image and video and device | |
CN101950416B (en) | Bidirectional filtration-based real-time image de-hazing and enhancing method | |
CN102063706B (en) | Rapid defogging method | |
CN102930514B (en) | Rapid image defogging method based on atmospheric physical scattering model | |
CN103955905A (en) | Rapid wavelet transformation and weighted image fusion single-image defogging method | |
CN107767354A (en) | A kind of image defogging algorithm based on dark primary priori | |
CN102982513B (en) | A kind of adapting to image defogging method capable based on texture | |
CN106846263A (en) | The image defogging method being immunized based on fusion passage and to sky | |
CN103198459B (en) | Haze image rapid haze removal method | |
CN103985091A (en) | Single image defogging method based on luminance dark priori method and bilateral filtering | |
CN102768760A (en) | Quick image dehazing method on basis of image textures | |
CN103049888A (en) | Image/video demisting method based on combination of dark primary color of atmospheric scattered light | |
CN104318524A (en) | Method, device and system for image enhancement based on YCbCr color space | |
CN102609909B (en) | Method and device for defogging single image | |
CN102831591A (en) | Gaussian filter-based real-time defogging method for single image | |
CN107301624A (en) | The convolutional neural networks defogging algorithm pre-processed based on region division and thick fog | |
CN105550999A (en) | Video image enhancement processing method based on background reuse | |
CN103489166A (en) | Bilateral filter-based single image defogging method | |
CN107451966A (en) | A kind of real-time video defogging method realized using gray-scale map guiding filtering | |
CN103578083A (en) | Single image defogging method based on joint mean shift | |
CN110288539A (en) | A kind of mobile clear method of underwater picture with dark channel prior in color combining space | |
CN104574325A (en) | Skylight estimation method and system as well as image defogging method thereof | |
CN104504722A (en) | Method for correcting image colors through gray points |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130403 |