CN109146905A - For the CANNY operator edge detection algorithm of low-light level environment - Google Patents
For the CANNY operator edge detection algorithm of low-light level environment Download PDFInfo
- Publication number
- CN109146905A CN109146905A CN201810999814.9A CN201810999814A CN109146905A CN 109146905 A CN109146905 A CN 109146905A CN 201810999814 A CN201810999814 A CN 201810999814A CN 109146905 A CN109146905 A CN 109146905A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- gradient
- follows
- formula
- 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
- 238000003708 edge detection Methods 0.000 title claims abstract description 12
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000009499 grossing Methods 0.000 claims abstract description 9
- 238000011551 log transformation method Methods 0.000 claims abstract description 5
- 230000035772 mutation Effects 0.000 claims abstract description 4
- 230000008569 process Effects 0.000 claims description 14
- 238000009415 formwork Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 238000009795 derivation Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 230000002708 enhancing effect Effects 0.000 description 5
- 238000001514 detection method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of CANNY operator edge detection algorithms for low-light level environment, comprising the following steps: step 1, carries out Gaussian smoothing filter, LOG transformation, histogram equalization to original image；Step 2, the mutation of sobel operator and sobel operator on tilted direction is respectively adopted on 2 groups of 4 directions in the image obtained to step 1, takes amplitude mean value as treated pixel number evidence；Step 3, on the contrary using high-low threshold value method, the pixel higher than high threshold retains, then give up, and the punishment of reservation, which is told, forms an edge-detected image.
Description
Technical field
The present invention relates to a kind of image processing techniques, especially a kind of CANNY operator edge for low-light level environment is examined
Method of determining and calculating.
Background technique
Image border part refers to that image in the significant part of partial zones surrounding pixel grey scale change, is widely present in not
With between object and between object and background.Edge extraction is exactly to detect these edges, and as far as possible completely and not by it
It extracts with introducing noise edge, forms the profile of primary objects and background.More other arithmetic operators, CANNY are calculated
Son uses the calculus of variations, to find an optimal edge detection algorithm.CANNY operator has detection accuracy very high, energy
The actual edge of image is detected as much as possible；Signal-to-noise ratio is big, and the edge of same object can only identify once, will not will likely
Existing noise token is edge；Accurate positioning, the edge at the edge and real image that identify very close to the advantages that；
The basic step of CANNY operator has 3 steps: image filtering, edge enhancing, edge detection.The algorithm master of edge enhancing
If single order and second dervative based on image, but derivative is very sensitive to noise, thus processing image first can to image into
The a degree of smoothing processing of row, to filter out noise；On the basis of filtering out noise, by determining image each point field intensity
Changing value, the point that image grayscale point field intensity value can be had significant change highlight, and reach the enhancing to image border,
Common practice is the amplitude of calculating gradient to determine；By the image of enhancing, often there is the gradient value much put in field very
Greatly, when detection, these points can be accepted or rejected using some way, common method is thresholding.
But the image under low-light level environment, image pixel gray level value is generally very low, and noise on image influences bigger and practical
Marginal point gradient magnitude is smaller, and image border is difficult to detect；Edge lines detection effect of traditional CANNY operator for tilted direction
It is bad；And the threshold value of final edge detection need to be specified artificially.
Summary of the invention
The purpose of the present invention is to provide a kind of CANNY operator edge detection algorithms for low-light level environment, including with
Lower step:
Step 1, Gaussian smoothing filter, LOG transformation, histogram equalization are carried out to original image；
Step 2, the image obtained to step 1 is respectively adopted sobel operator on 2 groups of 4 directions and sobel operator exists
Mutation on tilted direction takes amplitude mean value as treated pixel number evidence；
Step 3, using high-low threshold value method, the pixel higher than high threshold retains, on the contrary then give up, and the punishment of reservation is told a little
Form edge-detected image.
Compared with prior art, the present invention have the advantage that the present invention propose one kind before pretreatment through LOG transformation,
Histogram equalization is handled to enhance the detailed information of image, gives up image redundancy information；In image enhancement, increase to rectangle
Extraction to gradient information；And in the thresholding of final edge detection, proposes the adaptive method of multi-threshold, reach automatic
With extracting image optimal edge purpose.
The invention will be further described with reference to the accompanying drawings of the specification.
Detailed description of the invention
Fig. 1 is total algorithm flow chart.
Fig. 2 is 5 × 5 Gaussian convolution core template schematic diagrames.
Fig. 3 is the diagonal convolution mask schematic diagram of 3 × 3sobel operator.
Fig. 4 is untreated original image schematic diagram.
Fig. 5 is the image schematic diagram handled through original CANNY operator.
Fig. 6 is the image schematic diagram through this algorithm process.
Specific embodiment
In conjunction with Fig. 1, a kind of CANNY operator edge detection algorithm for low-light level environment, comprising the following steps:
Step 1, Gaussian smoothing filter is carried out to original image, to filter out high-frequency noise:
Since gaussian filtering is largely effective to the noise for inhibiting Normal Distribution, but with height in actual process
The increase of this template, image can cause the boundary between edge gradually to obscure by excessively smoothing, identified image border meeting
It gradually decreases.Therefore it needs to select appropriately sized Gaussian template.On the one hand inhibit some pseudo-edge points while retaining original image
Important edges information.
Gaussian smoothing filter function representation is as follows:
By original image f (x, y) and Gaussian filter convolution, smooth rear image g (x, y), procedural representation are obtained are as follows:
G (x, y)=f (x, y) * H (x, y, σ) (2)
σ appropriate is chosen in real image processing, 5 × 5 obtained Gaussian convolution kernel functions are shown in Fig. 2.
Step 2, LOG transformation is carried out to the image that previous step obtains, improves brightness of image and contrast, enhancing with whole
Image detail information:
It is image after transformation for original image f (x, y), L (x, y).
Transform method:
L (x, y)=c*log (1+v*g (x, y)) (3)
The value for choosing the truth of a matter appropriate and c, v handles each pixel.
Step 3, continue to carry out histogram equalization to image, enhance local contrast, rich image profile information:
What it is due to histogram equalization change is pixel grayscale distribution, and treatment process is indicated using another form.Directly
Side's figure equalization can be by following procedural representation:
Pixel grayscale distribution in image before and after equalization is regarded as F (x) and F (y), and the two random changes
Existence function transformational relation Y=T (X) between amount.Note f (x), f (y) are respectively the probability density function of F (x), F (y), then asking
The process of T (X) is exactly histogram equalization process.
Firstly,
Wherein nkThe pixel for being K for gray scale accounts for the ratios of all pixels, and M and N are distributed as the line number and columns of image；
Secondly
Wherein L is the number of gray level, usual L=256；
Because
F (y)=P (Y≤y)=P (T (x)≤y)=P (X≤T-1(y))=F (x) x=T-1(y) (6)
The y derivation simultaneously on above formula both sides is obtained
Above formula is inserted in part that is known and requiring
Arrangement is
Both sides are simultaneously to x integral (being summation herein)
Step 4, the mutation of sobel operator and sobel operator on tilted direction is respectively adopted on 2 groups of (totally 4) directions,
Take amplitude mean value as treated pixel number evidence:
The fundamental formular of sobel operator are as follows:
The amplitude of gradient is expressed from the next with direction, and wherein G is amplitude, and θ is direction:
2 groups of 4 directions, first group is level and the two vertical directions, i.e. X, Y-coordinate axle.Second group i.e. (just with x-axis
Direction at) it is oblique 45 degree with oblique 135 degree of the two directions.It is richer to do so information of the profile that can make on tilted direction.
The extraction to tilted direction gradient information is described below:
Two 3 × 3 pairs of Angle formworks shown in Fig. 2 are weighted and averaged the pixel value in image, and actually this is classical sobel
Expression of the operator on tilted direction.Correspondence is calculated as the gradient of Angle formwork 1
Correspondence is calculated as the gradient of Angle formwork 2
Root of making even obtains total gradient size
If remembering, the gradient magnitude on the direction x, y is G1, and the gradient magnitude on tilted direction is G2, then the image gradient after integrating
G is
Step 5, using high-low threshold value method, the intensity profile of threshold size image obtained by previous step is determined, threshold value
Be divided into high threshold, Low threshold, the pixel lower than Low threshold is given up, higher than high threshold pixel retain, if between being in its
There is the pixel higher than high threshold in field, then retains, it is on the contrary then give up；Finally obtain edge-detected image；Finally obtain side
Edge detection image:
Determination about high and low threshold value:
The average gray of entire image is found out first
Conventional edge detection algorithm often remains more redundancy, causes information clutter not simplify, this algorithm
Give up the redundancy of original image percent 30, therefore high threshold is set as 1.2g, Low threshold is the half of high threshold, i.e.,
0.6g。
For the pixel between high-low threshold value
Here field is appointed as symmetrical range certain around space of matrices locating for pixel, threshold value 1.2g.
Embodiment
This example has chosen personage's Background under low-light level, and Fig. 4 is untreated original image, and Fig. 5 is classics CANNY
Operator treated image, Fig. 6 are the image after this algorithm process.It can be seen that the image task and background wheel of this algorithm process
Wide more continuous, effective information is more and redundancy is less.
Claims (6)
1. a kind of CANNY operator edge detection algorithm for low-light level environment, which comprises the following steps:
Step 1, Gaussian smoothing filter, LOG transformation, histogram equalization are carried out to original image；
Step 2, sobel operator and sobel operator is respectively adopted in rectangle in the image obtained to step 1 on 2 groups of 4 directions
Upward mutation takes amplitude mean value as treated pixel number evidence；
Step 3, on the contrary using high-low threshold value method, the pixel higher than high threshold retains, then give up, and the punishment of reservation, which is told, to be formed
Edge-detected image.
2. the method according to claim 1, wherein in step 1 Gaussian smoothing filter detailed process are as follows:
By original image f (x, y) and Gaussian filter convolution, smooth rear image g (x, y) is obtained
G (x, y)=f (x, y) * H (x, y, σ) (1)
Wherein, H (x, y, σ) is Gaussian smoothing filter function,
3. according to the method described in claim 2, it is characterized in that, the detailed process that LOG changes in step 1 are as follows: pass through formula (2)
LOG is carried out to smooth rear image g (x, y) to change to obtain L (x, y)
L (x, y)=c*log (1+v*g (x, y)) (2)
Wherein, c, v are constant.
4. according to the method described in claim 3, it is characterized in that, setting picture in the image before and after equalizing in step 1
Plain gray level is F (x) and F (y), and existence function transformational relation y=T (x) between the two stochastic variables, if f (x), f (y)
The probability density function of respectively F (x) and F (y), seeking the process of T (x) is exactly histogram equalization process, detailed process are as follows:
Step S101, acquisition probability density function f (x), f (y)
Wherein, nkThe pixel for being K for gray scale accounts for the ratios of all pixels, and L is the number of gray level, and M, N are respectively the line number of image
And columns；
Step S102 is enabled
F (y)=P (Y≤y)=P (T (x)≤y)=P (X≤T-1(y))=F (x) | x=T-1(y) (6)
Wherein, P is probability, and Y is that original image passes through the transformed stochastic variable of histogram, and X is the stochastic variable before converting；
Step S103 obtains the y derivation simultaneously on formula (6) both sides
Formula (4) (5) are substituted into formula (7) and obtained by step S104
Step S105 integrates formula (9) both sides to x simultaneously
5. according to the method described in claim 4, it is characterized in that, in step 2 sobel operator formula are as follows:
The amplitude of point (x, y) gradient on S (x, y) is expressed from the next with direction, and wherein G is amplitude, and θ is direction:
The pixel value in image is weighted and averaged using two 3 × 3 pairs of Angle formworks, detailed process are as follows:
Step 2.1, it obtains and corresponds to Angle formwork 1 and the gradient of Angle formwork 2 is calculated as
Step 2.2, root of making even obtains total gradient size
Step 2.3, if the gradient magnitude on the note direction x, y is G1, the gradient magnitude on tilted direction is G2, then the image after being averaged
Gradient G is
6. according to the method described in claim 5, it is characterized in that, the detailed process of step 3 are as follows:
Step 5.1, the average gray of entire image is found out
Wherein, f'(x, y) it is the pixel number evidence for taking amplitude mean value；
Step 5.2, if high threshold is 1.2g, Low threshold 0.6g
Step 5.3, the pixel between high-low threshold value
Wherein, field is appointed as symmetrical range certain around space of matrices locating for pixel, threshold value 1.2g.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810999814.9A CN109146905A (en) | 2018-08-30 | 2018-08-30 | For the CANNY operator edge detection algorithm of low-light level environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810999814.9A CN109146905A (en) | 2018-08-30 | 2018-08-30 | For the CANNY operator edge detection algorithm of low-light level environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109146905A true CN109146905A (en) | 2019-01-04 |
Family
ID=64829204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810999814.9A Pending CN109146905A (en) | 2018-08-30 | 2018-08-30 | For the CANNY operator edge detection algorithm of low-light level environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109146905A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2588674A (en) * | 2019-11-01 | 2021-05-05 | Apical Ltd | Image processing |
| CN114140481A (en) * | 2021-11-03 | 2022-03-04 | 中国安全生产科学研究院 | Edge detection method and device based on infrared image |
| CN117557590A (en) * | 2024-01-11 | 2024-02-13 | 杭州汇萃智能科技有限公司 | High anti-noise image edge feature extraction method, system and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104112141A (en) * | 2014-06-29 | 2014-10-22 | 中南大学 | Method for detecting lorry safety belt hanging state based on road monitoring equipment |
| CN104700421A (en) * | 2015-03-27 | 2015-06-10 | 中国科学院光电技术研究所 | Adaptive threshold edge detection algorithm based on canny |
| CN107169979A (en) * | 2017-05-11 | 2017-09-15 | 南宁市正祥科技有限公司 | A kind of method for detecting image edge of improvement Canny operators |
| US20170301081A1 (en) * | 2015-09-30 | 2017-10-19 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
-
2018
- 2018-08-30 CN CN201810999814.9A patent/CN109146905A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104112141A (en) * | 2014-06-29 | 2014-10-22 | 中南大学 | Method for detecting lorry safety belt hanging state based on road monitoring equipment |
| CN104700421A (en) * | 2015-03-27 | 2015-06-10 | 中国科学院光电技术研究所 | Adaptive threshold edge detection algorithm based on canny |
| US20170301081A1 (en) * | 2015-09-30 | 2017-10-19 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| CN107169979A (en) * | 2017-05-11 | 2017-09-15 | 南宁市正祥科技有限公司 | A kind of method for detecting image edge of improvement Canny operators |
Non-Patent Citations (1)
| Title |
|---|
| YI ZHANG 等: "Edge Detection Algorithm of Image Fusion Based on Improved Aobel Operator", 《2017 IEEE 3RD INFORMATION TECHNOLOGY AND MECHATRONICS ENGINEERING CONFERENCE (ITOEC)》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2588674A (en) * | 2019-11-01 | 2021-05-05 | Apical Ltd | Image processing |
| GB2588674B (en) * | 2019-11-01 | 2022-02-02 | Apical Ltd | Image processing |
| US11823361B2 (en) | 2019-11-01 | 2023-11-21 | Arm Limited | Image processing |
| CN114140481A (en) * | 2021-11-03 | 2022-03-04 | 中国安全生产科学研究院 | Edge detection method and device based on infrared image |
| CN117557590A (en) * | 2024-01-11 | 2024-02-13 | 杭州汇萃智能科技有限公司 | High anti-noise image edge feature extraction method, system and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107808378B (en) | Method for detecting potential defects of complex-structure casting based on vertical longitudinal and transverse line profile features | |
| CN107507173B (en) | No-reference definition evaluation method and system for full-slice image | |
| CN107301661B (en) | High-resolution remote sensing image registration method based on edge point features | |
| CN102156996B (en) | Image edge detection method | |
| CN112819772B (en) | High-precision rapid pattern detection and recognition method | |
| CN113034452B (en) | Weldment contour detection method | |
| CN107067382A (en) | A kind of improved method for detecting image edge | |
| CN105913396A (en) | Noise estimation-based image edge preservation mixed de-noising method | |
| CN109785245A (en) | A kind of light spot image dressing method | |
| CN109242870A (en) | A kind of sea horizon detection method divided based on image with textural characteristics | |
| CN109146905A (en) | For the CANNY operator edge detection algorithm of low-light level environment | |
| CN109377450B (en) | Edge protection denoising method | |
| CN116468641A (en) | Infrared blood vessel image enhancement processing method | |
| CN110807763A (en) | Method and system for detecting ceramic tile surface bulge | |
| CN103455985B (en) | Crack on road Enhancement Method based on Hessian structural analysis | |
| CN114943744A (en) | Edge detection method based on local Otsu thresholding | |
| CN114332081B (en) | Textile surface abnormity determination method based on image processing | |
| CN104616259B (en) | A kind of adaptive non-local mean image de-noising method of noise intensity | |
| CN117853510A (en) | Canny edge detection method based on bilateral filtering and self-adaptive threshold | |
| CN112102189B (en) | Line structure light bar center line extraction method | |
| CN109636822A (en) | A kind of improvement Canny adaptive edge extracting method based on new building subordinating degree function | |
| CN110705568B (en) | Optimization method for image feature point extraction | |
| CN114998186B (en) | Method and system for detecting surface scab defect of copper starting sheet based on image processing | |
| CN110633705A (en) | Low-illumination imaging license plate recognition method and device | |
| CN104966271A (en) | Image denoising method based on biological vision receptive field mechanism |
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 |
Application publication date: 20190104 |
|
| RJ01 | Rejection of invention patent application after publication |