CN106157301A - A kind of threshold value for Image Edge-Detection is from determining method and device - Google Patents
A kind of threshold value for Image Edge-Detection is from determining method and device Download PDFInfo
- Publication number
- CN106157301A CN106157301A CN201610466010.3A CN201610466010A CN106157301A CN 106157301 A CN106157301 A CN 106157301A CN 201610466010 A CN201610466010 A CN 201610466010A CN 106157301 A CN106157301 A CN 106157301A
- Authority
- CN
- China
- Prior art keywords
- edge
- marginal point
- threshold value
- chain
- 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 Analysis (AREA)
Abstract
The invention discloses a kind of threshold value for Image Edge-Detection from determining method and device.This threshold value determines method certainly, including: use at least two Alternate thresholds that pending image carries out Image Edge-Detection respectively, and respectively the testing result of at least two Alternate thresholds is evaluated according to the Evaluation Strategy set；At least two evaluation result is compared, to determine target detection threshold value by the comparison criterion according to setting.Utilize this threshold value from determining method, it is possible to reduce strong noise interference during rim detection, additionally it is possible to preferably avoid the loss of important edges point during rim detection, it is ensured that the seriality at edge；Simultaneously, it is also possible to simplify selection course and the process of calculating, the accuracy rate of testing result when improve rim detection of detection threshold value.
Description
Technical field
The present embodiments relate to digital image processing techniques field, particularly relate to a kind of threshold for Image Edge-Detection
Value is from determining method and device.
Background technology
The edge of image is one of most basic feature of image, refers to the significant part of image local grey scale change.Due to
Human eye derives from the region with the strong grey scale change in local to the perception at scenery edge in image, and rim detection is intended to simulate people
The eye perception to scenery edge, extracts the edge that local gray level change is more violent, therefore, is typically based on rim detection to determine figure
The marginal portion of picture.When carrying out rim detection, main operation is namely based on the detection threshold value of setting and determines the edge of image
Point, and determined by the whether accurate of marginal point directly affect the whether accurate of edge detection results, therefore, to detection threshold value
Select and be determined to become the important component part of rim detection.
Usually, when image is carried out rim detection, mainly arranging detection threshold value by artificial selection, artificial selection examines
Survey threshold value and there is following deficiency: (1) subjectivity is strong.The result that different people selects is the most different, the technical staff's choosing lacked experience
The result selected is frequently not optimum, is not even suboptimum, is hereby based on selected detection threshold value when determining marginal point, deposits
Lose important edges point or there is the situation of strong noise interference.(2) process is loaded down with trivial details.Requirement during detection threshold value selection
Technical staff constantly attempts different threshold values evaluation result, expects that approximate optimal solution generally requires the selection course of complexity
And calculating process.
Summary of the invention
The purpose of the present invention is to propose to a kind of threshold value for Image Edge-Detection from determining method and device, to improve limit
The accuracy of edge testing result, thus reduces noise jamming and important edges while simplifying threshold value selection calculating process
The loss of point.
On the one hand, embodiments provide a kind of threshold value for Image Edge-Detection and certainly determine method, including:
Use at least two Alternate thresholds that pending image carries out Image Edge-Detection respectively, and according to the evaluation set
The testing result of at least two Alternate thresholds is evaluated by strategy respectively；
At least two evaluation result is compared, to determine target detection threshold value by the comparison criterion according to setting.
On the other hand, embodiments provide a kind of threshold value for Image Edge-Detection and certainly determine device, including:
Evaluation module, is used for using at least two Alternate thresholds respectively pending image to be carried out image border
Detection, and respectively the testing result of at least two Alternate thresholds is evaluated according to the Evaluation Strategy set；
Targets threshold determines module, at least two evaluation result being compared according to the comparison criterion set, with
Determine target detection threshold value.
The embodiment of the present invention provides a kind of threshold value for Image Edge-Detection from determining method and device.This threshold value
Initially with at least two Alternate thresholds, respectively pending image is carried out Image Edge-Detection from the method that determines, then according to setting
The testing result of at least two Alternate thresholds is evaluated by fixed Evaluation Strategy respectively；Finally according to the comparison criterion pair set
At least two evaluation result compares, and determines target detection threshold value.Utilize this threshold value from determining method, it is possible to reduce edge
Strong noise interference during detection, additionally it is possible to preferably avoid the loss of important edges point during rim detection, it is ensured that the company at edge
Continuous property；Simultaneously, it is also possible to simplify selection course and the process of calculating, the standard of testing result when improve rim detection of detection threshold value
Really rate.
Accompanying drawing explanation
A kind of threshold value for Image Edge-Detection that Fig. 1 provides for the embodiment of the present invention one determines that the flow process of method is shown certainly
It is intended to；
A kind of threshold value for Image Edge-Detection that Fig. 2 provides for the embodiment of the present invention two determines that the flow process of method is shown certainly
It is intended to；
A kind of threshold value for Image Edge-Detection that Fig. 3 provides for the embodiment of the present invention three determines that the flow process of method is shown certainly
It is intended to；
A kind of threshold value for Image Edge-Detection that Fig. 4 a provides for the embodiment of the present invention four determines the preferred of method certainly
Embodiment；
The left adjacent marginal point determining marginal point in the first hunting zone that Fig. 4 b provides for the embodiment of the present invention four and the right side
The exemplary plot of adjacent marginal point；
Fig. 5 determines the structural frames of device certainly for a kind of threshold value for Image Edge-Detection that the embodiment of the present invention five provides
Figure.
Detailed description of the invention
Further illustrate technical scheme below in conjunction with the accompanying drawings and by detailed description of the invention.May be appreciated
It is that specific embodiment described herein is used only for explaining the present invention, rather than limitation of the invention.Further need exist for explanation
, for the ease of describing, accompanying drawing illustrate only part related to the present invention rather than entire infrastructure.
Embodiment one
A kind of threshold value for Image Edge-Detection that Fig. 1 provides for the embodiment of the present invention one determines that the flow process of method is shown certainly
Being intended to, the method can be by the threshold value for Image Edge-Detection from determining that device performs, it is adaptable to carries out pending image
Threshold value during rim detection is from determining.Wherein this device can be realized by software and/or hardware, and is typically integrated at digital picture
In reason system.
As it is shown in figure 1, a kind of threshold value for Image Edge-Detection that the embodiment of the present invention one provides is from determining method, tool
Body includes operating as follows:
S101, employing at least two Alternate thresholds carry out Image Edge-Detection to pending image respectively, and according to setting
Evaluation Strategy respectively the testing result of at least two Alternate thresholds is evaluated.
In the present embodiment, described Alternate thresholds specifically can be regarded as the selected inspection for carrying out Image Edge-Detection
Survey threshold value.It should be noted that the detection threshold value carrying out Image Edge-Detection is generally a gray level, described gray level is concrete
Can be regarded as the different color range values residing for different colours, and generally the gray level that white is corresponding is denoted as 0, by ash corresponding for black
Degree level be denoted as 255, it follows that when color from black to leucismus time, gray level has also changed to 255 from 0, there occurs 256
The change of gray level.
In the present embodiment, described Evaluation Strategy specifically can be regarded as evaluating commenting of Alternate thresholds testing result quality
Price card is accurate.It should be noted that the evaluation index that the setting Main Basis user of described Evaluation Strategy is used, evaluation index is not
Just differ with the corresponding Evaluation Strategy set.Usually, carrying out threshold value can having as evaluation index in time determining: edge
Closeness, the length of boundary chain and the area etc. of marginal area of point.It follows that described Evaluation Strategy can be based on being used
The difference of evaluation index carries out concrete setting.
In the present embodiment, need at least the testing result of two Alternate thresholds to be evaluated based on Evaluation Strategy, and
The target detection threshold value being more suitable for Image Edge-Detection is determined in described at least two Alternate thresholds.In the present embodiment, institute
The MAXIMUM SELECTION scope stating Alternate thresholds is regarded as the excursion of gray level, and i.e. 0 to 255, but in actual applications, generally
One comparison alternative scope of rational threshold value can be set on the basis of MAXIMUM SELECTION scope, thus ensureing that threshold value is from determining standard
Threshold value is shortened from the operating time determined on the premise of exactness.
In the present embodiment, can be that Alternate thresholds presets the alternative scope of rational threshold value, the alternative model of described threshold value
Enclose and can be made up of the upper threshold condition of the minimum initial Alternate thresholds set and setting.Wherein, described minimum is the most alternative
Threshold value and upper threshold condition can be manually set, it is also possible to system default sets, and typically can carry out based on Alternate thresholds
Set before Image Edge-Detection.
Further, described at least two Alternate thresholds is from the beginning of the minimum initial Alternate thresholds set, to set step
Length is stepped up and obtains.
In the present embodiment, can arbitrarily choose an Alternate thresholds in the alternative scope of described threshold value, and based on selected
The Alternate thresholds selected carries out Image Edge-Detection to pending image, but does not limit owing to Alternate thresholds being chosen order,
Choosing of Alternate thresholds is made to there is situation about omitting.Preferably, in order to avoid the omission of Alternate thresholds, can be in described threshold value
Set one for Alternate thresholds to be chosen in alternative scope and choose order, i.e. first from the minimum initial Alternate thresholds set
Start, then carry out threshold value with setting step-length and certainly increase operation, the most progressively carry out choosing of other Alternate thresholds, until it reaches
Upper threshold condition.
S102, according to set comparison criterion at least two evaluation result is compared, to determine target detection threshold value.
In the present embodiment, after the testing result of described Alternate thresholds being evaluated based on setting strategy, can be formed
Corresponding evaluation result, thus can be analyzed comparing at least two evaluation result formed, determine required target
Detection threshold value.In the present embodiment, by the comparison criterion set at least two evaluation result can be analyzed and compare,
Wherein, the setting setting the described Evaluation Strategy of dependence of described comparison criterion, need concrete condition to make a concrete analysis of.Usually, often
The comparison criterion seen is set with: the size of value of calculation corresponding in comparative evaluation result or assay result are to whole edge
The impact etc. of detection performance.
A kind of threshold value for Image Edge-Detection that the embodiment of the present invention one provides is from determining method, initially with at least
Two Alternate thresholds carry out Image Edge-Detection respectively to pending image, then according to the Evaluation Strategy set respectively at least
The testing result of two Alternate thresholds is evaluated；Finally according to the comparison criterion set, at least two evaluation result is compared
Relatively, target detection threshold value is determined.Utilize this threshold value from determining method, it is possible to reduce strong noise interference during rim detection, also
Can preferably avoid the loss of important edges point during rim detection, it is ensured that the seriality at edge；Simultaneously, it is also possible to simplify inspection
Survey selection course and the process of calculating, the accuracy rate of testing result when improve rim detection of threshold value.
Embodiment two
A kind of threshold value for Image Edge-Detection that Fig. 2 provides for the embodiment of the present invention two determines that the flow process of method is shown certainly
It is intended to.The present invention implements two and is optimized based on above-described embodiment, in the present embodiment, is using the alternative threshold of at least two
Before value carries out Image Edge-Detection to pending image respectively, also optimize and include: determine based on the edge detection operator set
The edge gradient image of pending image, wherein, the corresponding gradient information of each pixel in described edge gradient image,
Described gradient information includes gradient magnitude and gradient direction；The alternative threshold of at least two is determined based on the threshold value alternative conditions set
Value, wherein, described threshold value alternative conditions sets based on the gradient information of pixel in described edge gradient image.
As in figure 2 it is shown, a kind of threshold value for Image Edge-Detection that the embodiment of the present invention two provides is from determining method, tool
Body includes operating as follows:
S201, edge detection operator based on setting determine the edge gradient image of pending image.
Usually, carry out during rim detection, needing that pending image is carried out edge gradient image and determine, described edge ladder
Degree image can determine based on edge detection operator, and described edge detection operator is the conventional sensing algorithm carrying out rim detection,
Common edge detection operator has Sobel operator, Kirsch operator, Robot operator and Canny operator etc..
In the present embodiment, the corresponding gradient information of each pixel in the edge gradient image formed, described
Gradient information includes gradient magnitude and gradient direction.Concrete, pending image is carried out the process that edge gradient image determines,
Can be regarded as each pixel in pending image and determine the process of corresponding gradient magnitude and gradient direction.In the present embodiment,
The gradient magnitude of described pixel specifically can be regarded as a functional value determined based on the gray level that described pixel is corresponding；Institute
State gradient direction and specifically can be regarded as the direction vector that the gradient magnitude of described pixel is corresponding.Based on described pixel gradient width
The size of value and gradient direction may be used to determine whether this pixel is the marginal point in pending image.
S202, threshold value alternative conditions based on setting determine at least two Alternate thresholds, wherein, described threshold value alternative conditions
Set based on the gradient information of pixel in described edge gradient image.
In the present embodiment, before the pixel of pending image carrying out marginal point and determines, need first to determine for limit
The Alternate thresholds of edge detection.Understanding based on above-described embodiment one, described at least two Alternate thresholds can be alternative in described threshold value
In the range of determine, and the alternative scope of described threshold value can be by the minimum initial Alternate thresholds set and the upper threshold condition of setting
Composition.Wherein, the determination of the alternative scope of described threshold value is based primarily upon the initial Alternate thresholds of described minimum and upper threshold condition
Set.
In the present embodiment, the initial Alternate thresholds of described minimum and upper threshold condition belong to threshold value alternative conditions, can
To set based on the gradient information of pixel in described edge gradient image.Additionally, after setting described threshold value alternative conditions, can
With in the alternative scope of the threshold value meeting threshold value alternative conditions, choose at least two Alternate thresholds, and described Alternate thresholds can be excellent
Elect the minimum initial Alternate thresholds from setting as to start to choose, and obtain other Alternate thresholds so that setting step-length is stepped up.
Specifically, based on the gradient information of pixel in described edge gradient image, can be by the most alternative for described minimum
Threshold value is set as: the gradient magnitude minimal gray level corresponding to pixel more than 0；Described upper threshold condition can be set
For: after carrying out thresholding based on selected Alternate thresholds, if the marginal point area formed and described edge gradient image surface
Long-pending ratio is less than setup parameter, then it is assumed that described Alternate thresholds reaches upper threshold, and wherein, described marginal point area is equal to threshold
The gradient magnitude sum of each marginal point determined by after value.
S203, employing at least two Alternate thresholds carry out Image Edge-Detection to pending image respectively, and according to setting
Evaluation Strategy respectively the testing result of at least two Alternate thresholds is evaluated.
S204, according to set comparison criterion at least two evaluation result is compared, to determine target detection threshold value.
A kind of threshold value for Image Edge-Detection that the embodiment of the present invention two provides, from determining method, specifically adds limit
The determination operation of edge gradient image and Alternate thresholds, thus for certainly the determining to provide and determine basis of threshold value during rim detection；
Additionally, also add optimal edge chain threshold value determination operation, determined by optimal edge chain threshold value can be at target detection threshold
The noise spot identical with marginal points information intensity is further filtered out, testing result when enhancing rim detection on the basis of value
Accuracy.Utilize this threshold value from determining method, it is possible to preferably reduce noise jamming when rim detection and avoid important limit
The loss of edge point, it is ensured that the seriality at edge and the accuracy rate of testing result.
Embodiment three
A kind of threshold value for Image Edge-Detection that Fig. 3 provides for the embodiment of the present invention three determines that the flow process of method is shown certainly
It is intended to.The embodiment of the present invention is optimized based on above-described embodiment, in the present embodiment, " will use at least two further
Individual Alternate thresholds carries out Image Edge-Detection respectively to pending image, and according to the Evaluation Strategy set respectively at least two
The testing result of Alternate thresholds is evaluated " it is embodied as: determine described limit respectively based at least two Alternate thresholds chosen
Marginal point in edge gradient image, and generate at least one boundary chain based on described marginal point；Add up the generation of described boundary chain
Bar number, and determine the boundary chain length of each of the edges chain；Based on standby described in described generation bar number and each boundary chain length computation
Select the average edge chain length that threshold value is corresponding；Using described average edge chain length as the evaluation result of described Alternate thresholds.
Further, also by " at least two evaluation result is compared, to determine target by the comparison criterion according to setting
Detection threshold value " it is embodied as: compare the length value of at least two average edge chain lengths；The average limit maximum by described length value
Edge chain length is defined as optimum evaluation result, and Alternate thresholds corresponding to described optimum evaluation result is designated as described target detection threshold
Value.
Further, after " determining target detection threshold value ", also comprise determining that described target detection threshold value is corresponding
Excellent boundary chain threshold value.
As it is shown on figure 3, the present invention implements a kind of threshold value for Image Edge-Detection of three offers from determining method, specifically
Including operating as follows:
S301, edge detection operator based on setting determine the edge gradient image of pending image.
S302, threshold value alternative conditions based on setting determine at least two Alternate thresholds.
S303, determine the marginal point in described edge gradient image respectively based at least two Alternate thresholds chosen, and
At least one boundary chain is generated based on described marginal point.
In the present embodiment, in the alternative scope of the threshold value meeting threshold value alternative conditions, choose at least two Alternate thresholds,
And be based respectively on described Alternate thresholds and determine the marginal point in edge gradient image.Concrete, the determination of described marginal point can table
State for: if the gray level that in described edge gradient image, pixel is corresponding is more than or equal to described Alternate thresholds, then it is assumed that institute
Stating pixel is marginal point.Afterwards, can based on determined by marginal point generate at least one boundary chain.
Further, generate at least one boundary chain based on described marginal point, including: according to the search order set and institute
State the gradient direction that in edge gradient image, each marginal point is corresponding, determine the left adjacent marginal point of each marginal point and right adjacent side edge respectively
Point；Each marginal point is attached with corresponding left adjacent marginal point and right adjacent side edge point；Based on consistency check criterion to often
Individual marginal point travel direction consistency check, it is thus achieved that at least one meets the boundary chain of consistency check criterion.
In the present embodiment, generate it is critical only that of boundary chain and determine the left adjacent marginal point of each marginal point and right adjacent side edge
Point.Concrete, the left adjacent marginal point of described marginal point and the determination process of right adjacent marginal point can be expressed as: a, by 0 ° to 360 °
Being divided into eight angular ranges clockwise, if note 0 ° to 45 ° is the first hunting zone, then 315 ° are just designated as the 8th to 360 °
Hunting zone, and set corresponding search order for each hunting zone；B, centered by described marginal point, determine with described
8 pixels that marginal point is adjacent；C, obtain the gradient direction of described marginal point, and determine the search belonging to described gradient direction
Scope；D, the left adjacent marginal point determining described marginal point based on the search order that described hunting zone is corresponding and right adjacent marginal point.
In the present embodiment, described for each hunting zone set corresponding search order, specifically can be expressed as:
1) in described eight hunting zones, centered by marginal point point be defined as the start angle of each hunting zone with
And start angle positive direction (setting with the direction turned clockwise as positive direction).
2) based on start angle positive direction, 8 pixels adjacent with marginal point are divided into 2 pixel point sets.
Concrete, the pixel on the central point start angle positive direction left side and the pixel adjacent with marginal point positive direction will be crossed
Point is divided to left adjacent pixel point set, by the pixel on the right of start angle positive direction and the pixel adjacent with marginal point opposite direction
It is divided to right adjacent pixel point set.
3) concentrate at left adjacent pixel point set and right adjacent pixel respectively, choose vertical with crossing central point start angle positive direction
Pixel as respective first Searching point；Choose and cross the parallel pixel of central point start angle positive direction as respective the
Two Searching point；Choosing and crossing central point start angle positive direction is that the pixel of negative 45° angle is as the 3rd Searching point；Choose and mistake
Central point start angle positive direction is that the pixel of positive 45° angle is as respective 4th Searching point.
In the present embodiment, as a example by the determination process of the left adjacent marginal point of described marginal point: based on residing for marginal point
Hunting zone, obtains the search order that described hunting zone is corresponding；Described first search is first determined whether based on described search order
Whether point is marginal point, if marginal point, is then denoted as the left adjacent marginal point of described marginal point, otherwise determines described second search
O'clock to whether the 4th Searching point exists left adjacent marginal point.If four Searching point are not left adjacent marginal point, then it is assumed that described
There is not left adjacent marginal point in marginal point；In like manner, can start to determine described from the first Searching point set based on identical mode
The right adjacent marginal point of marginal point.
In the present embodiment, after determining left adjacent marginal point and the right adjacent marginal point of each marginal point respectively, by each limit
Edge point is attached with corresponding left adjacent marginal point and right adjacent side edge point, now may form multiple initial edge chain, also need
The initial edge chain travel direction consistency check each marginal point is formed, it is thus achieved that at least one meets consistency check
The boundary chain of criterion.
Further, described consistency check criterion is: if the right adjacent marginal point that the left adjacent marginal point of marginal point is corresponding
It not described marginal point, then disconnect the connection of described marginal point and described left adjacent marginal point；If the right adjacent marginal point of marginal point
Corresponding left adjacent marginal point is not described marginal point, then disconnect the connection of described marginal point and described right adjacent marginal point.
In the present embodiment, after the initial edge chain travel direction consistency check that each marginal point will be formed, can
To obtain at least one boundary chain meeting consistency check criterion.
S304, add up the generation bar number of described boundary chain, and determine the boundary chain length of each of the edges chain.
In the present embodiment, the bar number that generates of described boundary chain specifically can be regarded as determining based on selected Alternate thresholds
The bar number of the generated boundary chain of marginal point；The length of described boundary chain specifically can refer to marginal point present in described boundary chain
Number.
S305, based on average edge chain corresponding to Alternate thresholds described in described generation bar number and each boundary chain length computation
Length.
In the present embodiment, the average edge chain length that described Alternate thresholds is corresponding is represented by: each boundary chain length it
And with the business of described generation bar number.
S306, using described average edge chain length as the evaluation result of described Alternate thresholds.
In the present embodiment, by the evaluation index of the length alternately threshold value of described boundary chain, the most described Alternate thresholds
Evaluation result be expressed as the average edge chain length that described Alternate thresholds is corresponding.
S307, compare the length value of at least two average edge chain lengths.
In the present embodiment, it is that at least two threshold value of putting on record is determined accordingly if based on step S303 to step S306
After evaluation result, then need at least two evaluation result is compared analysis.Concrete, that described Alternate thresholds is corresponding evaluation
Result is average edge chain length, accordingly, it would be desirable to compare the length value of at least two average edge chain lengths.
S308, average edge chain length maximum for described length value is defined as optimum evaluation result, described optimum evaluates
Alternate thresholds corresponding to result is designated as described target detection threshold value.
In the present embodiment, the average edge chain length that length value at least two average edge chain lengths is maximum is determined
For optimum evaluation result, and determining described optimum Alternate thresholds corresponding to evaluation result, described Alternate thresholds is exactly image limit
Target detection threshold value needed for edge detection.
S309, determine the optimal edge chain threshold value that described target detection threshold value is corresponding.
In the present embodiment, when determining target detection threshold value based on above-mentioned steps, and determine based on target detection threshold value
After going out the marginal point of described pending image, the marginal point determined still suffers from a small amount of noise spot.Because described noise
The noise intensity of some correspondence and the signal intensity of marginal point are with in the range of one, it is impossible to be based only upon target detection threshold filtering,
It is thus desirable to further determine that boundary chain detects threshold value, and filtered by the length scale of boundary chain formed to marginal point and make an uproar
Sound point.
Further, the described optimal edge chain threshold value determining that described target detection threshold value is corresponding, specifically include: determine institute
State the maximal margin chain length angle value in the chain of target detection threshold value corresponding edge；Calculate described maximal margin chain length angle value and set hundred
The product value of proportion by subtraction；Using described product value as the optimal edge chain threshold value of described target detection threshold value.
In the present embodiment, may determine that, based on step S303 to S306, the marginal point that described target detection threshold value is corresponding,
It may also be determined that the boundary chain that described target detection threshold value is corresponding, it is possible to length based on described boundary chain determines maximal margin
Chain length angle value, the optimal edge chain threshold value of the most described target detection threshold value be represented by calculating described maximal margin chain length angle value with
Set the product value of percentage ratio.
A kind of threshold value for Image Edge-Detection that the embodiment of the present invention three provides, from determining method, embodies alternative
The determination process of threshold ratings result and the determination process of target detection threshold value, use simple boundary chain length as evaluation
Index so that the determination process of target detection threshold value is simpler convenient.Utilize this threshold value from determining method, it is possible to examine at edge
Preferably reduce noise jamming during survey and avoid the loss of important edges point, it is ensured that the seriality at edge and testing result
Accuracy rate.
Embodiment four
A kind of threshold value for Image Edge-Detection that Fig. 4 a provides for the embodiment of the present invention four determines the preferred of method certainly
Embodiment.The embodiment of the present invention carries out rim detection based on Soble edge detection operator.As shown in fig. 4 a, the embodiment of the present invention
The preferred embodiment provided, specifically includes and operates as follows:
S401, determine the edge gradient image of pending image based on Soble edge detection operator.
Exemplary, in the edge gradient image that can will be determined, the distribution function of the gradient magnitude of each pixel sets
Being set to H (i), wherein, i ∈ [0,255], i represent the gray level corresponding to pixel；The value of the gradient direction of each pixel can
With in the range of [0 °, 360 °].
S402, determine threshold value alternative conditions based on gray level and the gradient information of pixel in edge gradient image.
In the present embodiment, described threshold value alternative conditions specifically can include setting and the threshold of minimum initial Alternate thresholds
Value upper bound condition.The initial Alternate thresholds of described minimum and upper threshold condition can be based on pixels in described edge gradient image
The gradient information of point sets.
Exemplary, connect above-mentioned example, if the distribution function of the gradient magnitude of pixel is set to H (i), wherein, i ∈
[0,255], i represents that the gray level corresponding to pixel, then set minimum initial Alternate thresholds formula are represented by:
Min (i) and H (i) > 0；Additionally, set upper threshold condition formula is represented by:Wherein, t
∈ [0,255], function H (i) represents the gradient magnitude of the marginal point of gray level i, formulaIt is expressed as based on Alternate thresholds t
The marginal point area formed by marginal point after thresholding, constant Area represents the gross area of edge gradient image, and α represents setting ginseng
Number.
S403, using initial for described minimum Alternate thresholds as current detection threshold value T.
Exemplary, using Min (i) and H (i) > 0 as current detection threshold value T.
S404, determine the marginal point in described edge gradient image based on current detection threshold value T.
Exemplary, mellow lime for edge gradient image degree level is more than or equal to the pixel of current detection threshold value T as limit
Edge point.
S405, the left adjacent marginal point determining marginal point and right adjacent marginal point, generate at least one boundary chain.
In the present embodiment, the left adjacent marginal point that it is critical only that marginal point of boundary chain and right adjacent marginal point are generated really
Fixed, the left neighbour's marginal point determining marginal point in the first hunting zone that Fig. 4 b provides for the embodiment of the present invention four and right adjacent side edge
The exemplary plot of point.
Exemplary, as shown in Figure 4 b, when being in the first hunting zone (0 ° to 45 °) with the gradient direction of marginal point it is
Example, the determination process of left adjacent marginal point and right adjacent marginal point is specifically expressed as: it is marginal point that note is designated the grid of 0, remaining 8
Grid is the neighbor pixel of described marginal point, carries out left adjacent side edge according to the search order of+1 ,+2 ,+3 ,+the 4 of mark in figure
The determination of point, if it is determined that the pixel being designated+1 is marginal point, then described pixel is defined as a left side for described marginal point
Adjacent marginal point, otherwise, continues to determine whether the marginal point being designated+2 is left adjacent marginal point.If 4 grids all travel through terminate
Still it is not determined by left adjacent marginal point, then it is assumed that described marginal point does not exist left adjacent marginal point；In like manner, can be based on identical mode
Start to determine the right adjacent marginal point of described marginal point from the grid being designated-1.
Exemplary, after the left adjacent marginal point determining described marginal point and right adjacent marginal point, by each marginal point with
Corresponding left adjacent marginal point and right adjacent side edge point are attached；Based on consistency check criterion to each marginal point travel direction one
Cause checks, it is thus achieved that at least one meets the boundary chain of consistency check criterion.
S406, calculate and store the evaluation result of current detection threshold value T.
Exemplary, calculate the average edge chain length that current detection threshold value T is corresponding, and by described average edge chain length
It is recorded as the evaluation result of current detection threshold value T.
S407, judge whether described current detection threshold value T reaches upper threshold condition, if it is not, then perform step S408；If
It is then to perform step S409.
S408, carry out current detection threshold value T from increasing, forming new current detection threshold value T, and returning setting step-length
S404。
S409, the length value of all average edge chain lengths more stored, by average edge chain length maximum pair
The detection threshold value answered is as target detection threshold value.
S410, determine the optimal edge chain threshold value that described target detection threshold value is corresponding.
Exemplary, the longest edge edge point length value corresponding to target detection threshold value and the product value setting percentage ratio are made
For optimal edge chain threshold value.
Embodiment of the present invention example four provides the threshold value for Image Edge-Detection from determining the preferred embodiment of method, base
In the preferred embodiment, the strong noise interference in time being determined to reduce rim detection of threshold value that the embodiment of the present invention provides is described,
Can also preferably avoid the loss of important edges point during rim detection, it is ensured that the seriality at edge；Simultaneously, it is also possible to simplify
The selection course of detection threshold value and the process of calculating, the accuracy rate of testing result when improve rim detection.
Embodiment five
Fig. 5 determines the structural frames of device certainly for a kind of threshold value for Image Edge-Detection that the embodiment of the present invention four provides
Figure.This threshold value is the determination of detection threshold value in time determining device to be applicable to image to carry out rim detection, this device can by software and/
Or hardware realizes, and typically it is integrated in digital image processing system.As it is shown in figure 5, this threshold value is from determining that device includes: detection
Evaluation of result module 51 and targets threshold determine module 52.
Wherein, evaluation module 51, it is used for using at least two Alternate thresholds respectively pending image to be carried out
Image Edge-Detection, and respectively the testing result of at least two Alternate thresholds is evaluated according to the Evaluation Strategy set；
Targets threshold determines module 52, at least two evaluation result being compared according to the comparison criterion set,
To determine target detection threshold value.
In the present embodiment, this threshold value is from determining that device first passes through evaluation module 51 and uses at least two standby
Select threshold value that pending image carries out Image Edge-Detection respectively, and alternative at least two respectively according to the Evaluation Strategy set
The testing result of threshold value is evaluated；Then targets threshold determines that the module 52 comparison criterion according to setting is at least two evaluation
Result compares, to determine target detection threshold value.
The embodiment of the present invention five provides a kind of threshold value for Image Edge-Detection and certainly determines device, utilizes this threshold value certainly
Determine device, it is possible to reduce strong noise interference during rim detection, additionally it is possible to important edges point when preferably avoiding rim detection
Loss, it is ensured that the seriality at edge；Simultaneously, it is also possible to simplify selection course and the process of calculating of detection threshold value, improve
The accuracy rate of testing result during rim detection.
Further, this threshold value includes from determining that device also optimizes: gradient image determines that module and Alternate thresholds determine mould
Block.
Wherein, gradient image determines module, for determining the edge of pending image based on the edge detection operator set
Gradient image, wherein, the corresponding gradient information of each pixel in described edge gradient image, described gradient information includes
Gradient magnitude and gradient direction；
Alternate thresholds determines module, for determining at least two Alternate thresholds based on the threshold value alternative conditions set, wherein,
Described threshold value alternative conditions sets based on the gradient information of pixel in described edge gradient image.
Further, described evaluation module 51, specifically for: divide based at least two Alternate thresholds chosen
Do not determine the marginal point in described edge gradient image, and generate at least one boundary chain based on described marginal point；Statistics is described
The generation bar number of boundary chain, and determine the boundary chain length of each of the edges chain；Based on described generation bar number and each edge chain length
Degree calculates the average edge chain length that described Alternate thresholds is corresponding；Using described average edge chain length as described Alternate thresholds
Evaluation result.
On the basis of above-described embodiment, described generate at least one boundary chain based on described marginal point, including:
According to the gradient direction that each marginal point in the search order set and described edge gradient image is corresponding, determine respectively
The left adjacent marginal point of each marginal point and right adjacent marginal point；Each marginal point is clicked on corresponding left adjacent marginal point and right adjacent side edge
Row connects；Based on consistency check criterion to each marginal point travel direction consistency check, it is thus achieved that at least one meets unanimously
Property check criterion boundary chain.
Further, described consistency check criterion is: if the right adjacent marginal point that the left adjacent marginal point of marginal point is corresponding
It not described marginal point, then disconnect the connection of described marginal point and described left adjacent marginal point；If the right adjacent marginal point of marginal point
Corresponding left adjacent marginal point is not described marginal point, then disconnect the connection of described marginal point and described right adjacent marginal point.
Further, described targets threshold determines module 52, specifically for: compare at least two average edge chain lengths
Length value；The average edge chain length that described length value is maximum is defined as optimum evaluation result, described optimum evaluation result pair
The Alternate thresholds answered is designated as described target detection threshold value.
On the basis of above-described embodiment, this threshold value certainly determines that device also optimizes and includes: boundary chain threshold determination module, uses
In the optimal edge chain threshold value determining that described target detection threshold value is corresponding.
Further, described boundary chain threshold determination module, specifically for: determine described target detection threshold value corresponding edge
Maximal margin chain length angle value in chain；Calculate described maximal margin chain length angle value and the product value setting percentage ratio；Take advantage of described
Product value is as the optimal edge chain threshold value of described target detection threshold value.
Further, described at least two Alternate thresholds is from the beginning of the minimum initial Alternate thresholds set, to set step
Length is stepped up and obtains.
Note, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious change,
Readjust and substitute without departing from protection scope of the present invention.Therefore, although by above example, the present invention is carried out
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
Other Equivalent embodiments more can be included, and the scope of the present invention is determined by scope of the appended claims.
Claims (18)
1. the threshold value for Image Edge-Detection determines method certainly, it is characterised in that including:
Use at least two Alternate thresholds that pending image carries out Image Edge-Detection respectively, and according to the Evaluation Strategy set
Respectively the testing result of at least two Alternate thresholds is evaluated；
At least two evaluation result is compared, to determine target detection threshold value by the comparison criterion according to setting.
Method the most according to claim 1, it is characterised in that using at least two Alternate thresholds respectively to pending figure
Before carrying out Image Edge-Detection, also include:
The edge gradient image of pending image, wherein, described edge gradient image is determined based on the edge detection operator set
In the corresponding gradient information of each pixel, described gradient information includes gradient magnitude and gradient direction；
Determining at least two Alternate thresholds based on the threshold value alternative conditions set, wherein, described threshold value alternative conditions is based on described
In edge gradient image, the gradient information of pixel sets.
Method the most according to claim 2, it is characterised in that use at least two Alternate thresholds respectively to pending image
Carry out Image Edge-Detection, and respectively the testing result of at least two Alternate thresholds is commented according to the Evaluation Strategy set
Valency, specifically includes:
The marginal point in described edge gradient image is determined respectively based at least two Alternate thresholds chosen, and based on described limit
Edge point generates at least one boundary chain；
Add up the generation bar number of described boundary chain, and determine the boundary chain length of each of the edges chain；
Based on the average edge chain length that Alternate thresholds described in described generation bar number and each boundary chain length computation is corresponding；
Using described average edge chain length as the evaluation result of described Alternate thresholds.
Method the most according to claim 3, it is characterised in that described generate at least one edge based on described marginal point
Chain, including:
According to the gradient direction that each marginal point in the search order set and described edge gradient image is corresponding, determine each limit respectively
The left adjacent marginal point of edge point and right adjacent marginal point；
Each marginal point is attached with corresponding left adjacent marginal point and right adjacent side edge point；
Based on consistency check criterion to each marginal point travel direction consistency check, it is thus achieved that at least one meets concordance inspection
Look into the boundary chain of criterion.
Method the most according to claim 4, it is characterised in that described consistency check criterion is: if a left side for marginal point
The right adjacent marginal point that adjacent marginal point is corresponding is not described marginal point, then disconnect the company of described marginal point and described left adjacent marginal point
Connect；If the left adjacent marginal point that the right adjacent marginal point of marginal point is corresponding is not described marginal point, then disconnect described marginal point and institute
State the connection of right adjacent marginal point.
Method the most according to claim 3, it is characterised in that according to the comparison criterion set at least two evaluation result
Compare, to determine target detection threshold value, specifically include:
The relatively length value of at least two average edge chain lengths；
The average edge chain length that described length value is maximum is defined as optimum evaluation result, and described optimum evaluation result is corresponding
Alternate thresholds is designated as described target detection threshold value.
7. according to the arbitrary described method of claim 1-6, it is characterised in that after determining target detection threshold value, also include:
Determine the optimal edge chain threshold value that described target detection threshold value is corresponding.
Method the most according to claim 7, it is characterised in that the described optimum limit determining that described target detection threshold value is corresponding
Edge chain threshold value, specifically includes:
Determine the maximal margin chain length angle value in the chain of described target detection threshold value corresponding edge；
Calculate described maximal margin chain length angle value and the product value setting percentage ratio；
Using described product value as the optimal edge chain threshold value of described target detection threshold value.
Method the most according to claim 1, it is characterised in that described at least two Alternate thresholds is from the beginning of the minimum set
Beginning Alternate thresholds starts, and is stepped up and obtains setting step-length.
10. the threshold value for Image Edge-Detection determines device certainly, it is characterised in that including:
Evaluation module, is used for using at least two Alternate thresholds respectively pending image to be carried out image border inspection
Survey, and respectively the testing result of at least two Alternate thresholds is evaluated according to the Evaluation Strategy set；
Targets threshold determines module, for comparing at least two evaluation result according to the comparison criterion set, to determine
Target detection threshold value.
11. devices according to claim 10, it is characterised in that also include:
Gradient image determines module, for determining the edge gradient image of pending image based on the edge detection operator set,
Wherein, the corresponding gradient information of each pixel in described edge gradient image, described gradient information includes gradient magnitude
And gradient direction；
Alternate thresholds determines module, for determining at least two Alternate thresholds based on the threshold value alternative conditions set, wherein, described
Threshold value alternative conditions sets based on the gradient information of pixel in described edge gradient image.
12. methods according to claim 11, it is characterised in that described evaluation module, specifically for:
The marginal point in described edge gradient image is determined respectively based at least two Alternate thresholds chosen, and based on described limit
Edge point generates at least one boundary chain；
Add up the generation bar number of described boundary chain, and determine the boundary chain length of each of the edges chain；
Based on the average edge chain length that Alternate thresholds described in described generation bar number and each boundary chain length computation is corresponding；
Using described average edge chain length as the evaluation result of described Alternate thresholds.
13. devices according to claim 12, it is characterised in that described generate at least one edge based on described marginal point
Chain, including:
According to the gradient direction that each marginal point in the search order set and described edge gradient image is corresponding, determine each limit respectively
The left adjacent marginal point of edge point and right adjacent marginal point；
Each marginal point is attached with corresponding left adjacent marginal point and right adjacent side edge point；
Based on consistency check criterion to each marginal point travel direction consistency check, it is thus achieved that at least one meets concordance inspection
Look into the boundary chain of criterion.
14. devices according to claim 13, it is characterised in that described consistency check criterion is:
If the right adjacent marginal point that the left adjacent marginal point of marginal point is corresponding is not described marginal point, then disconnect described marginal point and institute
State the connection of left adjacent marginal point；If the left adjacent marginal point that the right adjacent marginal point of marginal point is corresponding is not described marginal point, then break
Open the connection of described marginal point and described right adjacent marginal point.
15. devices according to claim 12, it is characterised in that described targets threshold determines module, specifically for:
The relatively length value of at least two average edge chain lengths；
The average edge chain length that described length value is maximum is defined as optimum evaluation result, and described optimum evaluation result is corresponding
Alternate thresholds is designated as described target detection threshold value.
16. according to the arbitrary described device of claim 11-15, it is characterised in that after targets threshold determines module, also wrap
Include:
Boundary chain threshold determination module, for determining the optimal edge chain threshold value that described target detection threshold value is corresponding.
17. devices according to claim 16, it is characterised in that described boundary chain threshold determination module, specifically for:
Determine the maximal margin chain length angle value in the chain of described target detection threshold value corresponding edge；
Calculate described maximal margin chain length angle value and the product value setting percentage ratio；
Using described product value as the optimal edge chain threshold value of described target detection threshold value.
18. devices according to claim 10, it is characterised in that described at least two Alternate thresholds is from the minimum set
Initial Alternate thresholds starts, and is stepped up and obtains setting step-length.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610466010.3A CN106157301B (en) | 2016-06-23 | 2016-06-23 | A kind of certainly determining method and device of the threshold value for Image Edge-Detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610466010.3A CN106157301B (en) | 2016-06-23 | 2016-06-23 | A kind of certainly determining method and device of the threshold value for Image Edge-Detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106157301A true CN106157301A (en) | 2016-11-23 |
| CN106157301B CN106157301B (en) | 2019-02-05 |
Family
ID=57353647
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610466010.3A Active CN106157301B (en) | 2016-06-23 | 2016-06-23 | A kind of certainly determining method and device of the threshold value for Image Edge-Detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106157301B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111986221A (en) * | 2020-09-07 | 2020-11-24 | 北京凌云光技术集团有限责任公司 | Edge evaluation method and device based on gray scale or position information |
| CN112801032A (en) * | 2021-02-22 | 2021-05-14 | 西南科技大学 | Dynamic background matching method for moving target detection |
| CN113658153A (en) * | 2021-08-24 | 2021-11-16 | 凌云光技术股份有限公司 | Contour data feature point detection method and device |
| CN113705672A (en) * | 2021-08-27 | 2021-11-26 | 国网浙江省电力有限公司双创中心 | Threshold value selection method, system and device for image target detection and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101465001A (en) * | 2008-12-31 | 2009-06-24 | 昆山锐芯微电子有限公司 | Method for detecting image edge based on Bayer RGB |
| WO2010140159A2 (en) * | 2009-06-05 | 2010-12-09 | Hewlett-Packard Development Company, L.P. | Edge detection |
| CN202736125U (en) * | 2012-06-15 | 2013-02-13 | 黄淮学院 | Medical image edge detection system |
| CN104077773A (en) * | 2014-06-23 | 2014-10-01 | 北京京东方视讯科技有限公司 | Image edge detection method, and image target identification method and device |
| CN104700421A (en) * | 2015-03-27 | 2015-06-10 | 中国科学院光电技术研究所 | Edge detection algorithm based on canny self-adaptive threshold value |
-
2016
- 2016-06-23 CN CN201610466010.3A patent/CN106157301B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101465001A (en) * | 2008-12-31 | 2009-06-24 | 昆山锐芯微电子有限公司 | Method for detecting image edge based on Bayer RGB |
| WO2010140159A2 (en) * | 2009-06-05 | 2010-12-09 | Hewlett-Packard Development Company, L.P. | Edge detection |
| CN202736125U (en) * | 2012-06-15 | 2013-02-13 | 黄淮学院 | Medical image edge detection system |
| CN104077773A (en) * | 2014-06-23 | 2014-10-01 | 北京京东方视讯科技有限公司 | Image edge detection method, and image target identification method and device |
| CN104700421A (en) * | 2015-03-27 | 2015-06-10 | 中国科学院光电技术研究所 | Edge detection algorithm based on canny self-adaptive threshold value |
Non-Patent Citations (1)
| Title |
|---|
| 姚刚等: "一种自适应梯度幅值的边缘检测算法", 《计算机与数字工程》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111986221A (en) * | 2020-09-07 | 2020-11-24 | 北京凌云光技术集团有限责任公司 | Edge evaluation method and device based on gray scale or position information |
| CN111986221B (en) * | 2020-09-07 | 2024-05-24 | 凌云光技术股份有限公司 | Edge evaluation method and device based on gray level or position information |
| CN112801032A (en) * | 2021-02-22 | 2021-05-14 | 西南科技大学 | Dynamic background matching method for moving target detection |
| CN112801032B (en) * | 2021-02-22 | 2022-01-28 | 西南科技大学 | Dynamic background matching method for moving target detection |
| CN113658153A (en) * | 2021-08-24 | 2021-11-16 | 凌云光技术股份有限公司 | Contour data feature point detection method and device |
| CN113658153B (en) * | 2021-08-24 | 2024-01-26 | 凌云光技术股份有限公司 | Contour data feature point detection method and device |
| CN113705672A (en) * | 2021-08-27 | 2021-11-26 | 国网浙江省电力有限公司双创中心 | Threshold value selection method, system and device for image target detection and storage medium |
| CN113705672B (en) * | 2021-08-27 | 2024-03-26 | 国网浙江省电力有限公司双创中心 | Threshold selection method, system, device and storage medium for image target detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106157301B (en) | 2019-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109741356B (en) | Sub-pixel edge detection method and system | |
| CN106485275B (en) | A method of realizing that cover-plate glass is bonded with liquid crystal display positioning | |
| CN105279372B (en) | A kind of method and apparatus of determining depth of building | |
| CN106934790B (en) | A kind of evaluation method of image definition, the method focused automatically and related device | |
| CN103426164B (en) | Based on scallop size computing method and the scallop sorting system of Opencv graphical analysis | |
| CN106157301A (en) | A kind of threshold value for Image Edge-Detection is from determining method and device | |
| CN100571335C (en) | Image syncretizing effect real-time estimating method and device based on pixel space relativity | |
| CN102385753A (en) | Illumination-classification-based adaptive image segmentation method | |
| CN103177451B (en) | Based on the self-adapting window of image border and the Stereo Matching Algorithm of weight | |
| CN109064418A (en) | A kind of Images Corrupted by Non-uniform Noise denoising method based on non-local mean | |
| CN104683783B (en) | A kind of self adaptation depth map filtering method | |
| CN102819850A (en) | Method for detecting edge of color image on basis of local self-adaption color difference threshold | |
| CN104732197B (en) | Score detection device and method | |
| CN108596975A (en) | A kind of Stereo Matching Algorithm for weak texture region | |
| CN103914827A (en) | Method for visual inspection of shortages of automobile sealing strip profile | |
| CN102306307B (en) | Positioning method of fixed point noise in color microscopic image sequence | |
| CN103473776A (en) | Method and system for comparing image definition and automatic focusing control method | |
| CN105354570B (en) | A kind of pinpoint method and system of license plate right boundary | |
| CN103914849A (en) | Method for detecting red date image | |
| CN106651797A (en) | Determination method and apparatus for effective region of signal lamp | |
| CN105070094B (en) | A kind of parking stall measure system and method for detecting parking stalls based on machine vision | |
| CN109410235B (en) | Target tracking method fusing edge features | |
| CN104282013B (en) | A kind of image processing method and device for foreground target detection | |
| CN110188640B (en) | Face recognition method, face recognition device, server and computer readable medium | |
| CN104243970A (en) | 3D drawn image objective quality evaluation method based on stereoscopic vision attention mechanism and structural similarity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 100094 Beijing city Haidian District Cui Hunan loop 13 Hospital No. 7 Building 7 room 701 Patentee after: Lingyunguang Technology Co., Ltd Address before: 100094 Beijing city Haidian District xingshikou road Yiyuan cultural creative industry base building No. 2 Room 101 Patentee before: LUSTER LIGHTTECH GROUP Co.,Ltd. |
|
| CP03 | Change of name, title or address |