CN109886901A - A kind of nighttime image enhancing method decomposed based on multi-path - Google Patents
A kind of nighttime image enhancing method decomposed based on multi-path Download PDFInfo
- Publication number
- CN109886901A CN109886901A CN201910218341.9A CN201910218341A CN109886901A CN 109886901 A CN109886901 A CN 109886901A CN 201910218341 A CN201910218341 A CN 201910218341A CN 109886901 A CN109886901 A CN 109886901A
- Authority
- CN
- China
- Prior art keywords
- image
- channel
- luminance
- nighttime
- tomographic image
- 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.)
- Granted
Links
Landscapes
- Image Processing (AREA)
- Facsimile Image Signal Circuits (AREA)
Abstract
The present invention discloses a kind of nighttime image enhancing method decomposed based on multi-path, applied to technical field of image processing, aiming at the problem that noise cannot be removed well when the prior art is in nighttime image processing, nighttime image to be processed according to the estimation of nighttime image global noise level to be processed, is decomposed into substrate tomographic image and details tomographic image first by the present invention；Then luminance adaptation calculating is carried out to substrate tomographic image；Secondly the color of basal layer luminance adaptation image is corrected；Edge-protected and noise suppressed is carried out to details tomographic image again；The substrate tomographic image after color correction is merged with the edge-protected details tomographic image with after noise suppressed finally, obtains enhanced nighttime image；Method of the invention can remove the noise jamming of nighttime image well.
Description
Technical field
The invention belongs to field of image processing, in particular to the enhancing technology of a kind of nighttime image.
Background technique
In image procossing, often have that brightness is low, uneven illumination is even and noise jamming in the image that night scenes obtain
The problems such as big.Low-quality nighttime image often will affect the working performance of the computing system based on image information, such as night
Monitoring system etc..Therefore, the pretreatment of vision enhancement is carried out (such as denoising, brightness to the collected low quality nighttime image of equipment
Enhancing, details enhancing etc.) it is of great significance.Have than more typical night or low-luminosity picture Enhancement Method and is estimated based on light source
Image enchancing method (LIME), referring to document " X.Guo, Y.Li, and H.Ling, " LIME:Low-light image
enhancement via illumination map estimation,”IEEE Trans.Image Processing,
vol.26,no.2,pp.982–993,2017.".This method is preferable to brightness reinforcing effect.And the processing for noise, the party
The scheme of use in method is after image enhancement, in addition plus an image de-noising method as post-processing.But due to figure
As noise is amplified during brightness enhances, therefore existing Denoising Algorithm can not remove noise well.
Summary of the invention
In order to solve the above technical problems, the present invention proposes a kind of nighttime image enhancing method decomposed based on multi-path, energy
It is enough not amplify noise in brightness enhancing.
A kind of the technical solution adopted by the present invention are as follows: nighttime image enhancing method decomposed based on multi-path, comprising:
S1, according to the estimation of nighttime image global noise level to be processed, nighttime image to be processed is decomposed into basal layer
Image and details tomographic image；Specifically: three Color Channels of red, green, blue for extracting nighttime image to be processed are logical to each color
The image in road carries out noise estimation, obtains the estimated value for reflecting each channel global noise level；With the global noise in each channel
Horizontal estimated value decomposes the color channel image as regularization parameter, based on Total Variation, and it is logical to obtain each color
Substrate tomographic image and details tomographic image after road decomposition.
S2, luminance adaptation calculating is carried out to substrate tomographic image；Specifically: substrate tomographic image is extracted using color notation conversion space
Luminance channel, luminance adaptation calculating then is carried out to luminance channel image, luminance channel image after obtaining luminance adaptation:
Wherein,ωg(x, y)=Lin(x,y)k, ωl
(x, y)=1- ωg(x, y), n=exp (σg)；(x, y) indicates image slices vegetarian refreshments coordinate, Lin(x, y) is from substrate tomographic image
The luminance channel image of extraction, MgFor the pixel mean value of luminance channel image, SgFor the standard deviation of luminance channel image pixel, Sl
(x, y) is the Local standard deviation in luminance channel image at the position pixel (x, y), Lout(x, y) is the brightness after luminance adaptation
Channel image, k are the parameter of setting, wsFor contrast weight coefficient.
S3, color correction is carried out to the substrate tomographic image after luminance adaptation；Specifically: some face obtained for step S1
The substrate tomographic image of chrominance channel, in conjunction with before each Color Channel luminance adaptation with the luminance channel image after luminance adaptation, to the face
The substrate tomographic image of chrominance channel carries out color correction processing, the substrate tomographic image after obtaining color correction:
Wherein,For three Color Channels of original substrate tomographic image, Lin(x, y) is brightness before luminance adaptation
Channel image, Lout(x, y) is the luminance channel image after adapting to, parameter s value range are as follows: [0,1].
S4, edge-protected and noise suppressed is carried out to details tomographic image；Specifically: to the details tomographic image of each Color Channel
The processing that takes absolute value is carried out, obtains the corresponding absolute value images of each Color Channel, then to the corresponding absolute value figure of each Color Channel
The local energy of each pixel is obtained as carrying out gaussian filtering process；Successively by the local energy value of each pixel and the picture
Gray value of the vegetarian refreshments in corresponding details tomographic image is multiplied, and obtains each pixel and carries out the value after noise suppressed；It is final to obtain
The details tomographic image of each Color Channel after edge-protected and noise suppressed.
S5, the substrate tomographic image that step S3 is obtained is merged with the details tomographic image that step S4 is obtained, is enhanced
Nighttime image afterwards.
Beneficial effects of the present invention: method of the invention first makes an uproar to three channels of the red, green, blue of image to be processed
Sound estimation, and using this group of noise level estimated value as parameter to each channel carry out picture breakdown, obtain substrate tomographic image with
Details tomographic image；Luminance channel is extracted from substrate tomographic image by color notation conversion space, and it is suitable to carry out brightness to luminance channel
It should calculate, color correction be carried out to basal layer luminance adaptation image, the substrate tomographic image after obtaining luminance adaptation, in levels of detail figure
Local energy estimation is carried out as in and realizes adaptive noise suppressed；Finally by the substrate tomographic image and edge after luminance adaptation
Details tomographic image after protection and noise suppressed is weighted fusion, obtains enhanced nighttime image；What the present invention was shown
Method can not amplify noise by picture breakdown when brightness enhances, and by can be very after edge-protected and noise suppressed
Night noise interference is removed well, is very suitable for the enhancing processing of low quality nighttime image.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of the nighttime image enhancing method based on more pathways for vision of the present invention.
Fig. 2 is the comparison diagram group enhanced using the method for the present invention a width nighttime image in embodiment；
Wherein, Fig. 2 (a) is original nighttime image, and Fig. 2 (b) is the enhancing image being calculated by control methods (LIME),
The enhancing image that Fig. 2 (c) is calculated by image enchancing method of the invention.
Specific embodiment
For convenient for those skilled in the art understand that technology contents of the invention, with reference to the accompanying drawing to the content of present invention into one
Step is illustrated.
The present invention is further elaborated with specific embodiment with reference to the accompanying drawing.
From presently disclosed low-quality image library PKU-EAQA (download address: https: //pkuml.org/
Resources/pku-eaqa.html select a width nighttime image (number: N-002-0) as objective for implementation, image size in)
It is 400 × 300 × 3, format is 24 bmp format-patterns.The process of circular is as shown in Figure 1, detailed process is as follows:
Picture breakdown is basal layer and levels of detail: extracting figure to be processed by the estimation S1. based on image overall noise level
Red (R) of picture, green (G), blue (B) three Color Channels, and noise estimation carried out to the image in each channel, in the present embodiment,
The noise level estimated value for obtaining three channels of red, green, blue is respectively as follows: 0.0086,0.0087,0.0086；With each channel
Noise level estimated value is regularization parameter, is decomposed respectively to the image in each channel based on Total Variation, is obtained every
Substrate tomographic image and details tomographic image after a channel decomposition；By taking pixel (100,100) as an example, the pixel is in input picture
The gray value in middle three channels of red, green, blue is respectively 0.0784,0.0745,0.0588.After picture breakdown, the pixel
The gray value in three channels is respectively 0.0900,0.0865,0.0705 in substrate tomographic image, and logical at details tomographic image three
The value in road is respectively -0.0116, -0.0120, -0.0116.
Picture breakdown method employed in step S1 is the method based on Total Variation, i.e., by solving following optimization
Problem:C ∈ { R, G, B }, image is basal layer after being decomposed
ImageAnd it corresponds to details tomographic image and isWherein λc,c∈{R,G,B}
For regularization parameter, ▽ indicates that gradient operator, (x, y) indicate image slices vegetarian refreshments coordinate.The estimated value table of global noise level
Up to formula are as follows:
Wherein,C ∈ { R, G, B }, IcFor image to be processed, (x, y) indicates that image slices vegetarian refreshments is sat
Mark,Indicate that convolution algorithm, W indicate that the width of image, H indicate the height of image.
S2. luminance adaptation calculating is carried out for substrate tomographic image；In the present embodiment, the basal layer figure that will be obtained in step S1
Picture (RGB image) is transformed into hsv color space, and extracts the channel V as luminance channel.By taking pixel (100,100) as an example, become
Value of this in the channel H, S, V is respectively as follows: 0.136,0.2169,0.0900 after changing to HSV space, extracts the channel V as brightness
Channel, i.e. pixel (100,100) are 0.0900 in the value of luminance channel；In the present embodiment, parameter k setting value is 0.01, ginseng
Number wsIt is 5.0 for setting value, the luminance channel image after luminance adaptation is calculated in luminance adaptation is carried out to luminance channel image
It is 0.8148 in the value of pixel (100,100)；Specific calculating formula is as follows:
Wherein,ωg(x, y)=Lin(x,y)k, ωl
(x, y)=1- ωg(x, y), n=exp (σg)；(x, y) indicates image slices vegetarian refreshments coordinate, Lin(x, y) is from substrate tomographic image
The luminance channel image of extraction, MgFor the pixel mean value of luminance channel image, SgFor the standard deviation of luminance channel image pixel, Sl
(x, y) is the Local standard deviation in luminance channel image at the position pixel (x, y), Lout(x, y) is the brightness after luminance adaptation
Channel image, k are the parameter of setting, value range are as follows: [0,1], wsFor contrast weight coefficient, value range are as follows: [0, ∞).
S3. the color correction of substrate tomographic image；Setup parameter value s is 0.6, after obtaining luminance adaptation after color correction
Substrate tomographic image (RGB image).By taking pixel (100,100) as an example, basal layer of this before original brightness adapts to calculate
Brightness of image channel intermediate value is 0.0900, and luminance channel image value of this after luminance adaptation is 0.8148, which exists
The gray value in three channels of red, green, blue is respectively 0.0784,0.0745,0.0588 in input picture, therefore color correction calculates
Afterwards, the gray value in three channels of substrate tomographic image of this after color correction be respectively as follows: 0.8148,0.7957,
0.7037；Specific calculating formula are as follows:
Wherein,For three Color Channels of original substrate tomographic image, Lin(x, y) is logical for brightness before luminance adaptation
Road image, Lout(x, y) is the luminance channel image after adapting to, parameter s value range are as follows: [0,1], ∈ are to belong to, and indicate element
Relationship between set, i.e. c are the element gathered in { R, G, B }.
S4. it is directed to the edge-protected and noise suppressed of details tomographic image；Specifically: to each channel of details tomographic image, each picture
Plain value, which takes absolute value, obtains absolute value images, then carries out gaussian filtering process to each channel of absolute value images and obtain each picture
The local energy of vegetarian refreshments；To channel each in image, each pixel, successively by the local energy value of each point with the point thin
The gray value saved in figure is multiplied, and obtains each pixel and carries out the value after noise suppressed；Final obtain passes through edge-protected and makes an uproar
Details tomographic image after sound inhibition；By taking pixel (100,100) as an example, the point in three Color Channels of details tomographic image
Gray value is respectively as follows: -0.0116, -0.0120, -0.0116, after taking absolute value are as follows: 0.0116,0.0120,0.0116；Again to exhausted
Gaussian filtering process is carried out to each channel of value image, obtains the local energy of pixel (100,100) in three channels
It is respectively as follows: 0.6942,0.6841,0.5739；By each channel, the local energy value of each point and ash of this in detail view
Angle value is multiplied, and the value after obtaining each pixel progress noise suppressed is respectively as follows: -0.0080, -0.0082, -0.0067；
S5. the substrate tomographic image after merging luminance adaptation and the details tomographic image after edge-protected and noise suppressed；
Weight coefficient value range are as follows: [0, ∞), in the present embodiment, setting weighting coefficient is 5.0；By taking pixel (100,100) as an example,
The value in substrate tomographic image after the luminance adaptation that the pixel obtains in step s3 are as follows: 0.8148,0.7957,0.7037,
The details tomographic image intermediate value after edge-protected and noise suppressed that the pixel obtains in step s 4 be -0.0080, -
0.0082, -0.0067, with the product of weighting coefficient are as follows: 0.6942,0.6841,0.5739；Finally obtain enhanced night
Value in image at pixel (100,100) are as follows: 0.7748,0.7547,0.6703；
Test results are shown in figure 2, and Fig. 2 (a) is original nighttime image, and Fig. 2 (b) is to be calculated by prior art LIME
Enhancing image, Fig. 2 (c) is the enhancing image being calculated by image enchancing method of the invention.It can understand from Fig. 2
Out, being enhanced using the method for the present invention nighttime image can obtain preferably denoising in effect, such as Fig. 2 (b) compared with dark space
There are obvious spot block distortions in domain (ground region), and ground region is more uniform in Fig. 2 (c), i.e., method noise of the invention
Inhibitory effect is more significant.
Those of ordinary skill in the art will understand that the embodiments described herein, which is to help reader, understands this hair
Bright principle, it should be understood that protection scope of the present invention is not limited to such specific embodiments and embodiments.For ability
For the technical staff in domain, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made
Any modification, equivalent substitution, improvement and etc. should be included within scope of the presently claimed invention.
Claims (8)
1. a kind of nighttime image enhancing method decomposed based on multi-path characterized by comprising
S1, according to the estimation of nighttime image global noise level to be processed, nighttime image to be processed is decomposed into substrate tomographic image
With details tomographic image；
S2, luminance adaptation calculating is carried out to substrate tomographic image；
S3, color correction is carried out to the substrate tomographic image after luminance adaptation；
S4, edge-protected and noise suppressed is carried out to details tomographic image；
S5, the substrate tomographic image that step S3 is obtained is merged with the details tomographic image that step S4 is obtained, is obtained enhanced
Nighttime image.
2. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 1, which is characterized in that step
S1 specifically: three Color Channels of red, green, blue for extracting nighttime image to be processed make an uproar to the image of each Color Channel
Sound estimation, obtains the estimated value for reflecting each channel global noise level；Using the global noise horizontal estimated value in each channel as
Regularization parameter decomposes the color channel image, obtains the substrate tomographic image after each Color Channel decomposes and levels of detail
Image.
3. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 2, which is characterized in that channel
Image is based on Total Variation and is decomposed.
4. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 3, which is characterized in that step
S2 specifically: extract the luminance channel of substrate tomographic image using color notation conversion space, brightness then is carried out to luminance channel image
It adapts to calculate, the luminance channel image after obtaining luminance adaptation.
5. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 4, which is characterized in that brightness
The expression formula of luminance channel image after adaptation are as follows:
Wherein,ωg(x, y)=Lin(x,y)k, ωl(x,y)
=1- ωg(x, y), n=exp (σg)；(x, y) indicates image slices vegetarian refreshments coordinate, Lin(x, y) is extracted from substrate tomographic image
Luminance channel image, MgFor the pixel mean value of luminance channel image, SgFor the standard deviation of luminance channel image pixel, Sl(x, y) is
Local standard deviation in luminance channel image at the position pixel (x, y), Lout(x, y) is the luminance channel figure after luminance adaptation
Picture, k are the parameter of setting, wsFor contrast weight coefficient.
6. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 5, which is characterized in that step
S3 specifically: for the substrate tomographic image of some obtained Color Channel of step S1, in conjunction with before each Color Channel luminance adaptation with
Luminance channel image after luminance adaptation carries out color correction processing to the substrate tomographic image of the Color Channel, obtains color and rectify
Substrate tomographic image after just.
7. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 6, which is characterized in that color
Basal layer image expression formula after correction are as follows:
Wherein,For three Color Channels of original substrate tomographic image, Lin(x, y) is luminance channel figure before luminance adaptation
Picture, Lout(x, y) is the luminance channel image after adapting to, parameter s value range are as follows: [0,1].
8. a kind of nighttime image enhancing method decomposed based on multi-path according to claim 7, which is characterized in that step
S4 specifically: the processing that takes absolute value is carried out to the details tomographic image of each Color Channel, obtains the corresponding absolute value of each Color Channel
Image, then the local energy that gaussian filtering process obtains each pixel is carried out to the corresponding absolute value images of each Color Channel；
The local energy value of each pixel is multiplied with gray value of the pixel in corresponding details tomographic image successively, is obtained each
Pixel carries out the value after noise suppressed；The final levels of detail for obtaining each Color Channel after edge-protected and noise suppressed
Image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910218341.9A CN109886901B (en) | 2019-03-21 | 2019-03-21 | Night image enhancement method based on multi-channel decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910218341.9A CN109886901B (en) | 2019-03-21 | 2019-03-21 | Night image enhancement method based on multi-channel decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109886901A true CN109886901A (en) | 2019-06-14 |
| CN109886901B CN109886901B (en) | 2021-07-02 |
Family
ID=66933543
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910218341.9A Active CN109886901B (en) | 2019-03-21 | 2019-03-21 | Night image enhancement method based on multi-channel decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109886901B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023138273A1 (en) * | 2022-01-18 | 2023-07-27 | 武汉联影医疗科技有限公司 | Image enhancement method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070003156A1 (en) * | 2005-07-01 | 2007-01-04 | Ali Corporation | Image enhancing system |
| CN102129673A (en) * | 2011-04-19 | 2011-07-20 | 大连理工大学 | Color digital image enhancing and denoising method under random illumination |
| CN103034986A (en) * | 2012-11-29 | 2013-04-10 | 奇瑞汽车股份有限公司 | Night vision image enhancement method based on exposure fusion |
| CN103413275A (en) * | 2013-07-26 | 2013-11-27 | 北京工业大学 | Retinex night image enhancement method based on gradient zero norm minimum |
| US20130329972A1 (en) * | 2012-06-08 | 2013-12-12 | Advanced Micro Devices, Inc. | Biomedical data analysis on heterogeneous platform |
| CN106327443A (en) * | 2016-08-24 | 2017-01-11 | 电子科技大学 | Night image enhancement method based on improved genetic algorithm |
-
2019
- 2019-03-21 CN CN201910218341.9A patent/CN109886901B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070003156A1 (en) * | 2005-07-01 | 2007-01-04 | Ali Corporation | Image enhancing system |
| CN102129673A (en) * | 2011-04-19 | 2011-07-20 | 大连理工大学 | Color digital image enhancing and denoising method under random illumination |
| US20130329972A1 (en) * | 2012-06-08 | 2013-12-12 | Advanced Micro Devices, Inc. | Biomedical data analysis on heterogeneous platform |
| CN103034986A (en) * | 2012-11-29 | 2013-04-10 | 奇瑞汽车股份有限公司 | Night vision image enhancement method based on exposure fusion |
| CN103413275A (en) * | 2013-07-26 | 2013-11-27 | 北京工业大学 | Retinex night image enhancement method based on gradient zero norm minimum |
| CN106327443A (en) * | 2016-08-24 | 2017-01-11 | 电子科技大学 | Night image enhancement method based on improved genetic algorithm |
Non-Patent Citations (2)
| Title |
|---|
| KAI YAN等: "Total Variation Models Based Algorithm of Illumination Normalization for Face Recognition", 《IEEE》 * |
| 涂清华: "低照度图像增强与去噪研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023138273A1 (en) * | 2022-01-18 | 2023-07-27 | 武汉联影医疗科技有限公司 | Image enhancement method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109886901B (en) | 2021-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Auto color correction of underwater images utilizing depth information | |
| CN105761227B (en) | Underwater picture Enhancement Method based on dark channel prior and white balance | |
| Su et al. | Shadow detection and removal for occluded object information recovery in urban high-resolution panchromatic satellite images | |
| CN106157267B (en) | Image defogging transmissivity optimization method based on dark channel prior | |
| Kim et al. | Optimized contrast enhancement for real-time image and video dehazing | |
| Nithyananda et al. | Review on histogram equalization based image enhancement techniques | |
| Gao et al. | Sand-dust image restoration based on reversing the blue channel prior | |
| Singh et al. | A novel dehazing model for remote sensing images | |
| Wang et al. | Variational single nighttime image haze removal with a gray haze-line prior | |
| CN111429370B (en) | Underground coal mine image enhancement method, system and computer storage medium | |
| Chaudhry et al. | A framework for outdoor RGB image enhancement and dehazing | |
| US8406559B2 (en) | Method and system for enhancing image sharpness based on local features of image | |
| Yu et al. | Image and video dehazing using view-based cluster segmentation | |
| CN109118440B (en) | Single image defogging method based on transmissivity fusion and adaptive atmospheric light estimation | |
| CN112435184B (en) | Image recognition method for haze days based on Retinex and quaternion | |
| CN111861896A (en) | UUV-oriented underwater image color compensation and recovery method | |
| Kim et al. | Single image haze removal using hazy particle maps | |
| Song et al. | An improved image dehazing and enhancing method using dark channel prior | |
| CN111311503A (en) | Night low-brightness image enhancement system | |
| CN109886901A (en) | A kind of nighttime image enhancing method decomposed based on multi-path | |
| JPWO2011021518A1 (en) | Image processing method, image processing apparatus, and program | |
| CN109636735B (en) | Rapid video defogging method based on space-time consistency constraint | |
| Wang et al. | Rapid nighttime haze removal with color-gray layer decomposition | |
| Abbaspour et al. | A new fast method for foggy image enhancement | |
| CN105741248A (en) | Method for removing hazed and degraded part in image |
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 |