CN108320272A - The method that image delusters - Google Patents
The method that image delusters Download PDFInfo
- Publication number
- CN108320272A CN108320272A CN201810111774.XA CN201810111774A CN108320272A CN 108320272 A CN108320272 A CN 108320272A CN 201810111774 A CN201810111774 A CN 201810111774A CN 108320272 A CN108320272 A CN 108320272A
- Authority
- CN
- China
- Prior art keywords
- image
- delusters
- original graph
- light
- pixel
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The present invention relates to the methods that image delusters, including：A. judge otherwise to enter step C if it is, entering step B with the presence or absence of strong light in the original graph of input by histogram；B. by intensity ratio and pixel cluster, diffusing reflection ingredient and the specular reflection component in original graph are isolated；C. the half-light in original graph or step B diffusing reflection ingredients is eliminated by histogram equalization；D. target image is exported.The method that the image of the present invention delusters, highlight detection can be carried out to the picture that electronic eyes is shot, and it is rapidly performed by bloom and/or unitary of illumination according to actual conditions, significantly enhance the local contrast of picture, to eliminate the influence of bloom and half-light to picture, picture recognition degree is effectively increased.
Description
Technical field
The present invention relates to the methods of image procossing, are concretely the methods that image delusters, and are particularly suitable for but are not limited only to
The method that image in terms of for public safety delusters.
Background technology
The system towards public safety applications can be real by spreading over the electronic eyes of each road in each city at present
When obtain road surface conditions.But the picture of electronic eyes shooting can be influenced by environment, such as strong light and half-light are deposited
The details for being likely to result in picture is smudgy, to being interfered to the feature recognition on subsequent picture.In order to improve
The accuracy rate of identification, should first be removed picture the pretreatment work of bloom, then carry out the normalization of illumination again to dark
Light carries out a degree of removal.
The existing algorithm that delusters mostly is to carry out reflecting component by iteration frame or solution linear/non-linear system
Separation, this needs the identification of minute surface pixel and the interaction of a large amount of local pixel, very big so as to cause taking, and can not reach
The effect delustered in real time.
Mobile edge calculations (Mobile Edge Computing, MEC) are the frameworks based on 5G evolution.MEC can utilize nothing
Line access network provides telecommunication user IT required services and high in the clouds computing function nearby, and creates one and have high-performance, low
Delay and the carrier grade service environment of high bandwidth, accelerate every content in network, service and alllication it is quick-downloading, allow consumer
Enjoy continual high network quality experience.It is cached after being handled the picture that electronic eyes is sent by MEC, to slow
Solve the processing pressure of Cloud Server.
Invention content
The present invention provides a kind of methods that image delusters, and highlight detection is carried out by the picture shot to electronic eyes, and
Bloom and/or unitary of illumination are carried out according to actual conditions, to enhance the local contrast of picture, to eliminate bloom and dark
Influence of the light to picture improves picture recognition degree.
The method that the image of the present invention delusters, including：
A. judge to whether there is strong light in the original graph of input by histogram, if it is, enter step B, otherwise into
Enter step C.If there are strong light in the picture of electronic eyes shooting, intensity value is 200~255 in the histogram of gray level image
Probability it is just very big.Therefore a probability threshold value can be preset, if intensity value is 200 in the histogram of its gray level image
~255 probability value has exceeded preset probability threshold value, then it is assumed that it needs to enter high-intensity region step there are strong light
Suddenly；Otherwise it is just not required to carry out the operation of high-intensity region；
B. by intensity ratio and pixel cluster, diffusing reflection ingredient and the specular reflection component in original graph are isolated；
C. the half-light in original graph or step B diffusing reflection ingredients is eliminated by histogram equalization；
D. target image is exported.
In computer vision field, a common hypothesis scene is exactly that body surface only has pure diffusing reflection, still
For the various non-uniform materials in reality scene, the reflection of body surface includes two kinds of diffusing reflection and mirror-reflection.It is unrestrained anti-
It penetrates and mirror-reflection is formed by Physical interaction different between light and body surface.When light is irradiated to uneven object
When the surface of matter, part of it is immediately in the reflected at interfaces of body surface and air, referred to as mirror-reflection or bloom.Part dissipates
It penetrates light to return to body surface and be again introduced into air, referred to as diffusing reflection.The color of mirror-reflection is identical as the color of illumination, and overflows
The color of reflection is the inherent characteristic of object.
Specifically, the intensity ratio and pixel cluster in step B include：The pseudo- coloration image in original graph is first calculated, then
Pixel is clustered according to pseudo- coloration image and specified Chroma threshold, then calculates separately out the intensity ratio of each cluster,
And the intensity ratio is estimated according to percentage threshold.The intensity ratio is referred to since image generally comprises one
Part has pure irreflexive pixel, therefore the ratio of the maximum value and value range (maximum value subtracts minimum value) diffusing reflection pixel
Value is known as the intensity ratio of image.
On this basis, due in image the white pixel of itself can be interfered to delustering, filtering can be passed through
Operation by illumination and image itself white distinguish, be then demultiplex out the diffusing reflection ingredient in original graph and mirror-reflection at
Point.
Specifically, since intensity histogram can show the frequency of occurrences of the different pixel values on different intensity values,
Strength range can independently show three channels between [0~255] for the coloured image of RGB for gray level image
Intensity histogram.Therefore a width gray scale can be worth to by calculating the maximum of the intensity histogram in three channels of image
Then figure is used as the filtering threshold progress filter operation by calculating the average intensity value of the gray-scale map.
Further, step C includes：The grey level histogram of original graph or step B diffusing reflection ingredients from the gray scale of concentration
Section is uniformly distributed in whole tonal ranges.
Specifically, by carrying out Nonlinear extension to original graph or step B diffusing reflection ingredients, the pixel of image is redistributed
Value, keeps the pixel quantity in a tonal range roughly the same, to which the contrast for enhancing image local is whole without influencing image
The contrast of body is finally reached the effect for eliminating half-light.
The method that the image of the present invention delusters can carry out highlight detection to the picture that electronic eyes is shot, and according to reality
Situation is rapidly performed by bloom and/or unitary of illumination, significantly enhances the local contrast of picture, to eliminate bloom
Influence with half-light to picture effectively increases picture recognition degree.
Specific implementation mode with reference to embodiments is described in further detail the above of the present invention again.
But the range that this should not be interpreted as to the above-mentioned theme of the present invention is only limitted to example below.Think not departing from the above-mentioned technology of the present invention
In the case of thinking, the various replacements or change made according to ordinary skill knowledge and customary means should all be included in this hair
In bright range.
Description of the drawings
Fig. 1 is the flow chart for the method that image of the present invention delusters.
Specific implementation mode
Embodiment 1：
In the present embodiment, the image that electronic eyes is shot is uploaded to the Cloud Server of distal end after processing locality, it is then objective
It family end can be by asking Cloud Server come the image that obtains that treated.
The method that image of the present invention as shown in Figure 1 delusters, step include：
A. original graph is shot by electronic eyes, and original graph is kept in advance in local storage card.To described temporary
Original graph carry out highlight detection.A probability threshold value is preset, if intensity value is 200 in the histogram of its gray level image
~255 probability value has exceeded preset probability threshold value, then it is assumed that it enters step B there are strong light；Otherwise without going
High light processing, enters step C.
B. take that minimum value in the intensity histogram in three channels of original graph constitutes a width and original graph is of a size
Then obtained minimum value gray-scale map f1 divided by total pixel number are obtained another width gray-scale map f2 by minimum value gray-scale map f1, then
F1 is subtracted with original image and obtains mirror-reflection figure fc plus f2；Then three channel strengths of each pixel of fc are calculated separately
The corresponding coloration of histogram, and take respectively three channels of each pixel correspond to coloration minimum value and maximum value constitute a width bilateral
The pseudo- coloration image in road.
Then cluster behaviour is carried out to pixel according to obtained twin-channel pseudo- coloration image and preset Chroma threshold
Make, if the chrominance distance between two pixels is less than preset Chroma threshold, then it is assumed that they belong to same cluster, no
Then belong to different clusters；
The intensity ratio of each cluster is calculated separately, and intensity ratio is estimated according to percentage threshold, to be estimated
The intensity ratio of calculation；
Since the white pixel of itself can generate interference to illumination in image, the strong histogram in three channels of artwork is taken
Maximum value constitutes a width gray-scale map, its average light intensity value is then calculated, using this light intensity value as the threshold of filter operation
Value, to avoid the generation of interference.
The largest light intensity angle value of each pixel in original image is finally subtracted to the intensity ratio and range of light intensities of its estimation
The product of value, if specular component is less than the average light intensity value obtained before, uses original image to obtain specular component
Light intensity value subtract specular component and obtain diffusing reflection component, i.e., the final figure that delusters.
C. it in the present embodiment, regardless of whether having carried out the processing for removing strong light of step B, all needs to carry out illumination normalizing to image
The grey level histogram of image is become in whole tonal ranges equal by the processing of change from some gray scale interval for comparing concentration
Even distribution.Method is by carrying out Nonlinear extension to original graph or step B diffusing reflection ingredients, redistributing the pixel of image
Value, keeps the pixel quantity in a tonal range roughly the same, to which the contrast for enhancing image local is whole without influencing image
The contrast of body is finally reached the effect for eliminating half-light.Purpose is in order to enhance local contrast, to eliminate the shadow of half-light
It rings, and more image details can be presented, further promote the accuracy rate of subsequent feature recognition in whole system.
D. after above-mentioned steps, electronic eyes will it is processed after cycle of images the Cloud Server for being uploaded to distal end,
Cloud Server is stored in after obtaining in its database.When client sends the request for obtaining image to the Cloud Server of distal end, cloud
Server is sent to client after obtaining image from database according to the request of reception.
Embodiment 2：
On the basis of embodiment 1, in the step A of the present embodiment, after electronic eyes shoots image, bloom is carried out to image
Detection process is then uploaded to the processing that MEC servers carry out bloom and unitary of illumination, otherwise directly if necessary to remove bloom
Connect the Cloud Server that the image of shooting is uploaded to distal end.
Step B and C is same as Example 1.In step D, the image after having handled is stored in its data by MEC servers
In library.When client request image, client to first give MEC servers send obtain image request, MEC servers according to
Whether the request of reception has corresponding image from data base querying, if there is being then transmitted directly to client, otherwise notifies client
End is obtained from Cloud Server.
Claims (6)
1. the method that image delusters, feature include：
A. judge to whether there is strong light in the original graph of input by histogram, if it is, entering step B, otherwise enter step
Rapid C；
B. by intensity ratio and pixel cluster, diffusing reflection ingredient and the specular reflection component in original graph are isolated；
C. the half-light in original graph or step B diffusing reflection ingredients is eliminated by histogram equalization；
D. target image is exported.
2. the method that image as described in claim 1 delusters, it is characterized in that：Intensity ratio and pixel cluster in step B include：
The pseudo- coloration image in original graph is first calculated, clusters pixel further according to pseudo- coloration image and specified Chroma threshold,
Then the intensity ratio of each cluster is calculated separately out, and the intensity ratio is estimated according to percentage threshold.
3. the method that image as claimed in claim 2 delusters, it is characterized in that：By filter operation by illumination and image itself
White distinguish, diffusing reflection ingredient and specular reflection component in original graph is then demultiplex out.
4. the method that image as claimed in claim 3 delusters, it is characterized in that：It is straight by the intensity for calculating three channels of image
The maximum of square figure is worth to a width gray-scale map, is then used as filtering threshold progress institute by calculating the average intensity value of the gray-scale map
The filter operation stated.
5. the method that image as described in claim 1 delusters, it is characterized in that：Step C includes：Original graph or step B is unrestrained anti-
The grey level histogram for penetrating ingredient is uniformly distributed in whole tonal ranges from the gray scale interval of concentration.
6. the method that image as claimed in claim 5 delusters, it is characterized in that：By to original graph or step B diffusing reflection ingredients
Nonlinear extension is carried out, the pixel value of image is redistributed, keeps the pixel quantity in a tonal range roughly the same, to increase
Contrast of the contrast of strong image local without influencing image entirety, is finally reached the effect for eliminating half-light.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810111774.XA CN108320272A (en) | 2018-02-05 | 2018-02-05 | The method that image delusters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810111774.XA CN108320272A (en) | 2018-02-05 | 2018-02-05 | The method that image delusters |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108320272A true CN108320272A (en) | 2018-07-24 |
Family
ID=62901889
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810111774.XA Pending CN108320272A (en) | 2018-02-05 | 2018-02-05 | The method that image delusters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108320272A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110346116A (en) * | 2019-06-14 | 2019-10-18 | 东南大学 | A kind of scene illumination calculation method based on Image Acquisition |
| CN112465940A (en) * | 2020-11-25 | 2021-03-09 | 北京字跳网络技术有限公司 | Image rendering method and device, electronic equipment and storage medium |
| CN113014912A (en) * | 2019-12-19 | 2021-06-22 | 合肥君正科技有限公司 | Method for detecting direct projection of strong light source of monitoring camera in vehicle |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1945627A (en) * | 2006-10-20 | 2007-04-11 | 清华大学 | Method for correcting digital tongue picture colour cast |
| CN102209246A (en) * | 2011-05-23 | 2011-10-05 | 北京工业大学 | Real-time video white balance processing system |
| CN102289657A (en) * | 2011-05-12 | 2011-12-21 | 西安电子科技大学 | Breast X ray image lump detecting system based on visual attention mechanism |
| CN102903081A (en) * | 2012-09-07 | 2013-01-30 | 西安电子科技大学 | Low-light image enhancement method based on red green blue (RGB) color model |
| WO2016068326A1 (en) * | 2014-10-31 | 2016-05-06 | オムロン株式会社 | Image processing device, character recognition device, image processing method, and program |
| CN105894050A (en) * | 2016-06-01 | 2016-08-24 | 北京联合大学 | Multi-task learning based method for recognizing race and gender through human face image |
| CN107607040A (en) * | 2017-08-11 | 2018-01-19 | 天津大学 | A kind of three-dimensional scanning measurement device and method suitable for High Reflective Surface |
-
2018
- 2018-02-05 CN CN201810111774.XA patent/CN108320272A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1945627A (en) * | 2006-10-20 | 2007-04-11 | 清华大学 | Method for correcting digital tongue picture colour cast |
| CN102289657A (en) * | 2011-05-12 | 2011-12-21 | 西安电子科技大学 | Breast X ray image lump detecting system based on visual attention mechanism |
| CN102209246A (en) * | 2011-05-23 | 2011-10-05 | 北京工业大学 | Real-time video white balance processing system |
| CN102903081A (en) * | 2012-09-07 | 2013-01-30 | 西安电子科技大学 | Low-light image enhancement method based on red green blue (RGB) color model |
| WO2016068326A1 (en) * | 2014-10-31 | 2016-05-06 | オムロン株式会社 | Image processing device, character recognition device, image processing method, and program |
| CN105894050A (en) * | 2016-06-01 | 2016-08-24 | 北京联合大学 | Multi-task learning based method for recognizing race and gender through human face image |
| CN107607040A (en) * | 2017-08-11 | 2018-01-19 | 天津大学 | A kind of three-dimensional scanning measurement device and method suitable for High Reflective Surface |
Non-Patent Citations (3)
| Title |
|---|
| SHEN H等: "Real-time highlight removal using intensity ratio", 《APPLIED OPTICS》 * |
| 李芳: "基于偏振光原理的图像高光消除方法", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王巍: "基于多小波变换的光照鲁棒性人脸特征提取算法研究与设计", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110346116A (en) * | 2019-06-14 | 2019-10-18 | 东南大学 | A kind of scene illumination calculation method based on Image Acquisition |
| CN113014912A (en) * | 2019-12-19 | 2021-06-22 | 合肥君正科技有限公司 | Method for detecting direct projection of strong light source of monitoring camera in vehicle |
| CN112465940A (en) * | 2020-11-25 | 2021-03-09 | 北京字跳网络技术有限公司 | Image rendering method and device, electronic equipment and storage medium |
| CN112465940B (en) * | 2020-11-25 | 2021-10-15 | 北京字跳网络技术有限公司 | Image rendering method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dhara et al. | Color cast dependent image dehazing via adaptive airlight refinement and non-linear color balancing | |
| RU2658874C1 (en) | Fog remover device and method of forming images | |
| Cheng et al. | A simple and effective histogram equalization approach to image enhancement | |
| Xu et al. | Fast image dehazing using improved dark channel prior | |
| Santhi et al. | Adaptive contrast enhancement using modified histogram equalization | |
| CN107103591B (en) | Single image defogging method based on image haze concentration estimation | |
| CN108320272A (en) | The method that image delusters | |
| KR20140142381A (en) | Method and Apparatus for removing haze in a single image | |
| CN105959510B (en) | A kind of video rapid defogging method | |
| US20160104272A1 (en) | Auto-contrast enhancement system | |
| Sanila et al. | Underwater image enhancement using white balance, USM and CLHE | |
| CN108230275A (en) | The method of image defogging | |
| JP2020503587A (en) | Method and apparatus for estimating projection and / or highlight areas in an image | |
| Nair et al. | An effective surround filter for image dehazing | |
| CN112750089B (en) | Optical remote sensing image defogging method based on local block maximum and minimum pixel prior | |
| CN110136085B (en) | Image noise reduction method and device | |
| KR20150002187A (en) | System and method for image defogging based on gaussian filtering | |
| CN108898561A (en) | A kind of defogging method, server and the system of the Misty Image containing sky areas | |
| CN113066077B (en) | Flame detection method and device | |
| Mosny et al. | Cubical gamut mapping colour constancy | |
| CN112699760A (en) | Face target area detection method, device and equipment | |
| Harifi et al. | Efficient iris segmentation based on converting iris images to high dynamic range images | |
| Chandpa et al. | Comparative study of linear and non-linear contrast enhancement techniques | |
| Kaur et al. | Comparative study on various single image defogging techniques | |
| CN104809712B (en) | A kind of image fast repairing method based on rough set |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |