CN109410228A - Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features - Google Patents
Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features Download PDFInfo
- Publication number
- CN109410228A CN109410228A CN201810960749.9A CN201810960749A CN109410228A CN 109410228 A CN109410228 A CN 109410228A CN 201810960749 A CN201810960749 A CN 201810960749A CN 109410228 A CN109410228 A CN 109410228A
- Authority
- CN
- China
- Prior art keywords
- image
- scale
- edge
- fusion
- wavelet
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Abstract
The invention discloses a kind of internal wave of ocean detection algorithms based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, comprising the following steps: the multi-scale morphology based on wavelet modulus maxima obtains multi-scale edge information；Edge is detected using Multiscale Morphological to the low frequency subgraph picture obtained after wavelet decomposition；The method for recycling connected domain erases the small interference region of area；The result of modulus maximum multi-scale wavelet edge detection results and Multiscale Morphological edge detection is subjected to image co-registration, finally obtains Multi-structure elements multi-scale morphology image.Compared with existing texture filtering method, filter result of the invention can keep the structural information of image well, while can filter out some unnecessary grain details again；Structure detection proposed by the present invention and texture filtering algorithm compare oneself have algorithm the identification and holding of weak gradient-structure and the inhibition of multiple dimensioned, strong gradient texture and in terms of achieve preferable effect.
Description
Technical field
The present invention relates to field of Computer Graphics, especially a kind of sea based on Method Based on Multi-Scale Mathematical Morphology Fusion Features
Wave detection algorithm in ocean.
Background technique
Some local features that traditional edge detection method is mainly based upon image carry out the structural information in detection image.
For example, a series of gradient operators such as sobel operator, Roberts operator and Prewitt operator, mainly by grayscale image
First derivative is calculated to detect edge.Marr and Hildreth propose it is a kind of first image is carried out using Gaussian function it is smooth,
It reuses Laplace operator and carries out derivation, detect edge using zero crossing, this method is otherwise known as LOG (laplacian
Of gaussian) operator.Another more classical edge detection algorithm be canny operator it edge is regarded as bright
Discontinuous pixel in channel is spent, the edge of image is obtained by non-maxima suppression and hysteresis threshold processing step.But
It is canny operator before detecting edge, first with Gaussian smoothing filter to entire image smoothing processing, while inhibiting noise
Also part-structure edge has been obscured.And testing result is bigger to the dependence of scale, for different images, optimal scale ginseng
Number is often not quite similar.Above-mentioned edge detection method is designed for local edge in detection image, and there is no consider texture
Margin residual, testing result are unintelligible.
The frames of nearest some border detection algorithm combination machine learning introduces multiple features inputs, as gradient, brightness,
The border detection problem of image is regarded as the classification problem of a machine learning by the features such as texture, color and depth.Martin etc.
People proposes a kind of boundary response prediction algorithm Pb based on study using the color of image and texture as classifier input feature vector.
Since skirt response is related with scale, some multi-scale morphology devices based on Pb are also put forward one after another.Then,
Dollar et al. proposes BEL algorithm, mainly carries out edge using Boosting technology and object boundary learns, which can be real
The border detection of existing special object, but it is too dependent on training set.In addition, multiple dimensioned is also to promote edge inspection in recent research
Survey a strategy of effect.Lopez-Molina et al. proposes a kind of multiple dimensioned using Gaussian smoothing and Edge track realization
The method of edge detection.Ren has carried out multiple dimensioned extension to Pb method, calculates contrast on border on different scale, position
It sets as Analysis On Multi-scale Features vector, in this, as the input of classifier, training classifier obtains the testing result of Analysis On Multi-scale Features,
Effectively improve the performance of border detection.Ren and Bo proposes one kind by calculating the method for sparse coding gradient (SCG) to detect
Profile, algorithm achieve good performance.Ren et al. proposes to carry out closed algorithm to profile using probabilistic model.Jiang etc.
Propose it is a kind of using regression analysis and various features combined prediction super-pixel boundary whether be remarkable configuration method.Dollar
Et al. propose rapid edge and profile testing method.Sam and Charles also uses random forest training classifier, and prediction exists
Contrast on multiple contour directions.Et al. W propose that based on the border detection algorithm towards contrast, which is surround
The advantages of inhibiting model and border detection model based on machine learning, using the frame of machine learning, towards contrast sky
Between texture edge is inhibited.Perfect with deep learning theory and method, deep learning is also that contour detecting problem mentions
A good thinking is supplied.Kivinen et al., Maire et al., Bertasius et al. and Shen et al. are respectively with depth
The method of habit studies contour detecting problem, achieves good effect.Boundary based on study and profile inspection above
Survey method, be all centered on pixel image block or local neighborhood in extract feature, the method for reusing machine learning
It is trained classification, whether prediction pixel point is boundary or profile.Although detection effect is well, but effect relies on very much
In training set selection and training also than relatively time-consuming.
Mathematical morphology (Mathematical Morphology) is the Men Xueke in image procossing, it is established tight
On the basis of the mathematical theory of lattice, using the form of image as the subject of research object.Mathematical morphology is born in 1964, as one
The subject of the emerging image processing and analysis of door studies emphatically the collecting structure of image based on geometry.Initially it is
Analyze the mathematical method of geometries and structure.Now it solve inhibit noise, shape recognition, edge detection, image segmentation,
Texture analysis, the recovery and reconstruction of image are widely applied on the image processing problems such as compression of images.Its basic thought
It is to go to measure and extract the correspondingly-shaped in image with the structural element with certain form to reach to image analysis and identification
Purpose.
The method of morphology processing image has intuitive than the method for other airspaces or frequency domain image processing and analysis
On simplicity and preciseness mathematically, but in traditional edge detection method, usually using Laplace operator,
Roberts operator, Sobel operator, Prewitt operator, Gauss-Laplace operator, Canny operator etc..These algorithms are to making an uproar
The interference of sound is very sensitive, causes testing result unstable.Improved multiple scale detecting algorithm has good noise immunity, and
And the edge-smoothing detected, feature is clear, and real-time is good.
Edge extraction processing based on mathematical morphology better than based on the Boundary extracting algorithm differentiated, it unlike
Differential algorithm is sensitive like that noise, while the edge extracted is smoother, has in description signal aspect feature unique
Advantage.Not only noise can be effective filtered out, but also original detailed information in image can be retained, is one of edge detecting technology
Important breakthrough.
It needs to select reasonable method to carry out according to survey region situation and actual needs in internal wave of ocean and extraction of ocean eddies
The extraction of sharp side and vortex, or extracting method is improved.In addition, the selection of threshold value is the difficult point of Nei Bo and extraction of ocean eddies,
Influence of the selection of threshold value vulnerable to subjective factor at present, selected threshold value is excessively high to will lead to marine environment characteristic information there are defect,
Threshold value is too low, can reduce marine environment characteristic information positioning accuracy.At present still without a kind of ideal, objective, adaptive journey
Spend high Research on threshold selection.Therefore, the selection of reasonable threshold value still needs to carry out a large amount of work in marine environment feature information extraction
Make and studies.Existing research person is to Research on threshold selection expansion research in marine environment feature extraction at present, such as Bo et al. basis
Image accumulative histogram feature is extracted threshold value to sharp side and choose and detected using dual threshold to sharp side.In addition, at present
Most widely used gradient method and Canny algorithm are for strong edge (such as solid at the beginning of algorithm designs in interior wave extraction
Edge) extraction, the target edges of research are apparent.And ocean belongs to fluid, each hydrographic features are in satellite image in ocean
In show the feature at weak edge, i.e. edge is unobvious, therefore gradient method and Canny algorithm are not particularly suited for mentioning for fluid edge
It takes.
Summary of the invention
The purpose of the present invention is to provide a kind of, and the internal wave of ocean based on Method Based on Multi-Scale Mathematical Morphology Fusion Features detects calculation
Method.
Realize the technical solution of the object of the invention are as follows: a kind of internal wave of ocean based on Method Based on Multi-Scale Mathematical Morphology Fusion Features
Detection algorithm, comprising the following steps:
Step 1, using wavelet decomposition, test object is decomposed, obtains the high frequency subgraph and low frequency of source images
Image；
Step 2, edge is detected using Multiscale Morphological to the low frequency subgraph picture obtained after wavelet decomposition；
Step 3, the Multiscale Morphological edge detection results obtained to step 2 erase area using the method for connected domain
Less than the interference region of given threshold, Multi-structure elements Multiscale Fusion back edge detection image is obtained；
Step 4, modulus maximum multi-scale wavelet edge detection results and Multiscale Morphological edge detection results are carried out
Image co-registration finally obtains Multi-structure elements multi-scale morphology image.
The present invention compared with prior art, significant advantage are as follows: the present invention is by binaryzation original graph and binaryzation
Corrosion image removes interference region by the method for connected domain；Then the two-value of corrosion image is subtracted with original image binary map
Figure obtains novel more structural form edge detection results figures, and algorithm is more convenient and efficient, effectively erases image border result
Interference region, improves the accuracy of edge detection, remains the edge feature of image well.
Detailed description of the invention
Fig. 1 is the internal wave of ocean detection algorithm flow chart the present invention is based on Method Based on Multi-Scale Mathematical Morphology Fusion Features.
Fig. 2 is internal wave of ocean original image.
Fig. 3 is edge detection graph of the modulus maximum multi-scale wavelet to internal wave of ocean.
Fig. 4 is edge detection schematic diagram of the Method Based on Multi-Scale Mathematical Morphology to same figure.
Fig. 5 is after Multiscale Morphological detects and to carry out connected domain selection schematic diagram.
Fig. 6 is the internal wave of ocean result figure after image co-registration.
Fig. 7 is a kind of edge detection schematic diagram of canny operator to internal wave of ocean.
Specific implementation method
In conjunction with Fig. 1, a kind of internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, including following step
It is rapid:
Step 1, using wavelet decomposition, test object is decomposed, obtains the high frequency subgraph and low frequency of source images
Image；
Step 2, edge is detected using Multiscale Morphological to the low frequency subgraph picture obtained after wavelet decomposition；
Step 3, the Multiscale Morphological edge detection results obtained to step 2 erase area using the method for connected domain
Less than the interference region of given threshold；
Step 4, modulus maximum multi-scale wavelet edge detection results and Multiscale Morphological edge detection results are carried out
Image co-registration finally obtains Multi-structure elements multi-scale morphology image.
Further, step 1 specifically:
Using wavelet decomposition, test object is decomposed, finally obtains the high frequency subgraph and low frequency subgraph of source images
Picture.In terms of low frequency, only one low frequency subgraph picture.There are three high frequency subgraphs, be respectively it is horizontal, vertical, diagonal these three
The high frequency detail subgraph in direction.
The frequency of image: the index of gray-value variation severe degree is gradient of the gray scale on plane space.High frequency subgraph
Picture and low frequency subgraph picture are done convolution to original image by low-pass filter and high-pass filter and are obtained, and low-pass filter is decomposed:
[1 1]/sqrt (2) decomposes high-pass filter: [- 1 1]/sqrt (2).
Multi-scale morphology based on wavelet modulus maxima obtains multi-scale edge information: 1. selecting Db4 conduct
Wavelet function, internal wave image f (x, y) carry out dyadic wavelet transform；2. obtaining image level and vertical side by transformation coefficient
To wavelet conversion coefficient, find out the modulus value and argument of dyadic wavelet transform；3. finding out mould by the modulus value and argument of wavelet transformation
It is worth the Local modulus maxima along argument direction, the i.e. marginal point of image；4. edge point is reasonably connected and is accepted or rejected.It obtains
Low-frequency approximation subgraph and high frequency detail subgraph.
Further, edge is carried out using ash value mathematical morphology to the low-frequency approximation subgraph containing image bulk information
Detection obtains low-frequency image edge.
Edge detection is carried out to image using different structural elements, passes through side of the entity Weighted Fusion in conjunction with comentropy
Method carries out image co-registration to edge image, obtains the edge detection results G1 under single scale.
5 structural elements are expanded respectively, side is carried out to image in scale 2 with 5 structural elements after expansion
5 testing results of acquisition are carried out image co-registration according to the fusion method of step 1, obtain side when scale n=2 by edge detection
Edge testing result G2.
5 structural elements are expanded respectively, side is carried out to image in scale 3 with 5 structural elements after expansion
5 testing results of acquisition are carried out image co-registration according to the fusion method of step 1, obtain side when scale n=3 by edge detection
Edge testing result G3.
Similarly, it is examined according to the edge that above-mentioned steps carry out n scale (n takes 2~5 in the present embodiment) to 5 structural elements
It surveys and fusion, the blending image for obtaining n different scale is respectively as follows: G1 ..., Gn.By 5 kinds of structural element Image erosions as a result,
According to adaptive weighting, new corrosion grayscale image a G'(x, y are reconfigured out).By G'(x, y), two width grayscale image of f (x, y)
As binaryzation, image-region is carried out by the method for connected domain and is erased, the new image GL (x, y) of two width, fL (x, y) are obtained.It is logical
Formula is crossed, F (x, y)=fL (x, y)-GL (x, y) finally obtains the edge image F (x, y) of Multiscale Morphological processing.
Further, the small interference region of area is erased using the method for connected domain, obtains that Multi-structure elements are multiple dimensioned to be melted
Close back edge detection image.
Further, two obtained width edge image application image fusion functions are subjected to image co-registration, i.e. pixel space
It is added and is carried out the closed operation of appropriate threshold selection, obtains final fusion results.
Due to the interface that interior wave is two heterogeneity water bodies, it can be considered to be increased using region in image segmentation
Method by the ocean essential image segmentation of survey region at the similar region of several features, i.e., water body of different nature, so
The boundary between each water body is extracted afterwards, which is interior wave.When the ocean essential image to survey region is split, need
Suitable similarity criterion is selected, and passes through the similarity criterion for the similar region merging technique of feature.
Picture edge characteristic after over-segmentation is strengthened, to solve the problems, such as the weak edge of original image.Therefore
First ocean essential image is split, then edge detection is carried out to image after segmentation, better sharp side extraction effect can be obtained.
With reference to the accompanying drawing and specific embodiment is described in further details the present invention.
Embodiment 1
A kind of internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, comprising the following steps:
Step 1, wavelet decomposition: utilizing wavelet decomposition, decompose to test object, finally obtains high frequency of source images
Image and low frequency subgraph picture.In terms of low frequency, only one low-frequency approximation subgraph.There are three high frequency subgraphs, is water respectively
The high frequency detail subgraph in these three flat, vertical, diagonal directions.
A. selecting smooth function θ (x, y) is scaling function, finds out the first-order partial derivative ψ of function # (x, y)x(x, y) and ψy
(x, y), by ψx(x, y) and ψy(x, y) is considered as wavelet function.
B. wavelet transformation is carried out to image with following formula (1), (2), two for acquiring the horizontal and vertical direction of wavelet transformation
ComponentWith
Wherein f is source images；
C. modulus maximum and gradient direction of the image wavelet transform coefficient on gradient direction are found out.
The wavelet coefficient acquired calculates mould M according to the following formula (3) and (4)sF (x, y), argument AsF (x, y).
Due to the presence of noise, adaptive block threshold value is carried out to wavelet conversion coefficient modulus of local maximum image.
Export multi-scale morphology image.
Step 2, based on the edge detection of Multiscale Morphological: using more rulers to the low frequency subgraph picture obtained after wavelet decomposition
Spend morphologic detection edge.In view of the direction of image target area edge pixel point and the changeable feature of form, 5 kinds are constructed
3 × 3 structural elements, the linear structure element including 4 directions have respectively represented 0 °, 45 °, 90 ° and 135 ° direction and 1 circle
Dish structure element.This 5 structural elements can preferably extract all directions feature of target area, and can extract target side
Noise is filtered out while edge pixel.It is respectively indicated with matrix are as follows:
This 5 kinds of structural elements of b1, b2, b3, b4, b5 are selected, single scale operation is made, are carried out with morphological erosion type formula
Edge corrosion operation:
Gi (x, y)=f (x, y) Θ bi
Wherein, f (x, y) is original interior wave input picture, and i is structural element, and Gi (x, y) is that the image of i structural element is rotten
Lose result.
Step 3, by 5 kinds of structural element Image erosions as a result, reconfiguring out a new corrosion according to adaptive weighting
Grayscale image G'(x, y).If weight coefficient ai(x, y), i are structural element, and x, y represent Gi (x, y) Image erosion result xth row
Y column pixel.The thought of ai (x, y) weight coefficient: not changing this characteristic of the size of image according to Image erosion result, will
The Image erosion result of different structure element takes the pixel on same ranks to be compared.The present invention selects 5 structural elements,
Just there are 5 edge corrosion result figure G1 (x, y), G2 (x, y), G3 (x, y), G4 (x, y), G5 (x, y), with 5 pictures on ranks
Vegetarian refreshments gray value is set as Pi (x, y), and 5 pixels two are neither repeated to subtract each other to take absolute value.According to permutation and combination concept, m is a
Structural element just has m different pixel point values,It is a it is different subtract each other combination, select 5 structural elements, just there is 5 differences
Image erosion pixel point value Pi(x, y), (i=1,2,3,4,5) and It is a to subtract each other combination, i.e.,
E (x, y)={ ej,z(x, y)=| Pj(x, y)-Pz(x, y) | | 1≤j, z≤m；j<z}
Wherein, Pj(x, y), Pz(x, y) is that two different structure elements corrode pixel, and E (x, y) indicates Pj(x, y), Pz
The set that (x, y) pixel subtracts each other, j, z are structural element, eJ, z(x, y) is Pj(x, y), PzThe difference that (x, y) gray value subtracts each other, m
For the number of structural element.Threshold value λ, λ a value is set according to the complexity of image, more complicated value is smaller, more
Simple value is bigger.
When E1 (x, y)=ej, z (x, y)=| Pj (x, y)-Pz (x, y) |≤| 1≤j, z≤m；j<z}
And there are two the pixel point element Pj (x, y), Pz of not identical structural element in the E1 (x, y) each remained
(x, y).All pixels for remaining E1 (x, y) are subjected to statistic of classification according to the method that identical structural element is one kind,
Statistical result is set as Si, SiIndicate i-th of structural element edge corrosion pixel Pi(x, y) is in all E1Number (m in (x, y)
=5, that is, there is S1, S2, S3, S4, S5), the S that statistics is obtainediAccording to formula:
Wherein, the number of m representative structure element, ai(x, y) is weight coefficient.Pass through public affairs further according to this weight coefficient
Formula:
Wherein, Pnew(x, y) is indicated in G'(x, y) gray value of the pixel of xth row y column in image.According to the picture obtained
Vegetarian refreshments PnewThe new gray value of (x, y) traverses xth row y column by the 1st row 1 column, is combined into the more structural images edge images of single scale
Etch figures, i.e. G'(x, y).
By G'(x, y), two width Binary Sketch of Grey Scale Image of f (x, y) carries out image-region by the method for connected domain and erases.
The classification that connected domain is carried out by the concept of connected domain, is divided into several connected regions, calculates pixel in each connected region
The number of point is denoted as num.It can be obtained according to the ratio of the size of image and interference region size, when (wherein x, y are num≤xy/36
The ranks number of image) when, the connected domain is regarded as interference region, erases the interference region, and concrete operations are by the connection interference range
All pixels point in domain is all set as 0.Thus obtain the new bianry image of two width:
Pass through formula, F (x, y)=fL (x, y)-GL (x, y)
Final detection obtains the more structural form edge image F (x, y) of dynamic self-adapting weight.
Step 4, the small interference region of area is erased using the method for connected domain, finally obtains more regular internal wave of ocean
Image.
Step 5, the closed operation of appropriate threshold selection is finally added and carried out according to pixel space, obtains final fusion knot
Fruit.
The canny edge detection for carrying out original image manages filtering algorithm with three fringes under a kind of guidance of structure, to knot
Structure and texture take different filtering modes respectively, obtain the filter result that structure is kept and texture is smooth.By itself and previous step
Obtained fusion results image compares.
Embodiment 2
Illustrate technical effect of the invention below by emulation.
A kind of internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, comprising the following steps:
Step 1, using wavelet decomposition, test object is decomposed, finally obtains the high frequency subgraph of source images and low
Frequency subgraph.Fig. 2 is internal wave of ocean original image, and Fig. 3 is edge detection graph of the modulus maximum multi-scale wavelet to internal wave of ocean.
Step 2, edge inspection is carried out using ash value mathematical morphology to the low-frequency approximation subgraph containing image bulk information
Low-frequency image edge is measured, as shown in Figure 4.
Edge detection is carried out to image using different structural elements, passes through side of the entity Weighted Fusion in conjunction with comentropy
Method carries out image co-registration to edge image, obtains the edge detection results Gf1 under single scale.
5 structural elements are expanded respectively, side is carried out to image in scale 2 with 5 structural elements after expansion
5 testing results of acquisition are carried out image co-registration according to the fusion method of step 1, obtain side when scale n=2 by edge detection
Edge testing result Gf2.
Step 3,5 structural elements are expanded respectively, with 5 structural elements after expansion in scale 3 to image
Edge detection is carried out, 5 testing results of acquisition are subjected to image co-registration according to the fusion method of step 1, obtain scale n=3
When edge detection results Gf3.
Similarly, the edge detection and fusion that according to above-mentioned steps 5 structural elements are carried out with n scale (n takes 2~5), are obtained
The blending image for obtaining n different scale is respectively as follows: Gf1 ..., Gfn.By 5 kinds of structural element Image erosions as a result, according to adaptive
Weight reconfigures out new corrosion grayscale image a G'(x, y).By G'(x, y), two width Binary Sketch of Grey Scale Image of f (x, y),
Image-region is carried out by the method for connected domain to erase, and obtains the new image GL (x, y) of two width, fL (x, y).Pass through formula, F
(x, y)=fL (x, y)-GL (x, y) finally obtains the edge image F (x, y) of Multiscale Morphological processing.
Step 4, the small interference region of area is erased using the method for connected domain, after obtaining Multi-structure elements Multiscale Fusion
Edge-detected image, as shown in Figure 5.
Step 5, the closed operation of appropriate threshold selection is finally added and carried out according to pixel space, obtains final fusion knot
Fruit such as Fig. 6.
It is compared with existing conventional method: carrying out the canny edge detection of original image, under a kind of guidance of structure
Three fringes manage filtering algorithm, different filtering modes is taken respectively to structure and texture, obtain structure keep and texture it is smooth
Filter result, see Fig. 7.It is compared with fusion results image obtained in the previous step.From fig. 6, it can be seen that Fig. 7 is compared,
Novel more structural form edge detection fusion results figures, algorithm is more convenient and efficient, effectively erases the dry of image border result
Region is disturbed, the accuracy of edge detection is improved, remains the edge feature of image well.
Claims (5)
1. a kind of internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, which is characterized in that including following
Step:
Step 1, using wavelet decomposition, test object is decomposed, obtains the high frequency subgraph and low frequency subgraph picture of source images；
Step 2, edge is detected using Multiscale Morphological to the low frequency subgraph picture obtained after wavelet decomposition；
Step 3, the Multiscale Morphological edge detection results obtained to step 2, erase area using the method for connected domain and are less than
The interference region of given threshold obtains Multi-structure elements Multiscale Fusion back edge detection image；
Step 4, modulus maximum multi-scale wavelet edge detection results and Multiscale Morphological edge detection results are subjected to image
Fusion, finally obtains Multi-structure elements multi-scale morphology image.
2. the internal wave of ocean detection algorithm according to claim 1 based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, special
Sign is, the specific steps of step 1 wavelet decomposition are as follows:
Select Db4 as wavelet function, internal wave image f (x, y) carries out dyadic wavelet transform；
By transformation coefficient, the wavelet conversion coefficient of image level and vertical direction is obtained, finds out the modulus value of dyadic wavelet transform
And argument；
By the modulus value and argument of wavelet transformation, modulus value is found out along the Local modulus maxima in argument direction, the i.e. marginal point of image；
Edge point is connected and accepted or rejected, low frequency subgraph picture and high frequency subgraph are obtained.
3. the internal wave of ocean detection algorithm according to claim 1 based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, special
Sign is that step 2 carries out edge detection using grey value mathematical morphology to low frequency subgraph picture and obtains low-frequency image edge；Specifically
Are as follows:
Edge detection is carried out to image using different structural elements, passes through method pair of the entity Weighted Fusion in conjunction with comentropy
Edge image carries out image co-registration, obtains the edge detection results G1 under single scale；
5 structural elements are expanded respectively, edge inspection is carried out to image in scale 2 with 5 structural elements after expansion
It surveys, 5 testing results of acquisition is subjected to image co-registration according to the fusion method of step 1, obtain edge inspection when scale n=2
Survey result G2；
5 structural elements are expanded respectively, edge inspection is carried out to image in scale 3 with 5 structural elements after expansion
It surveys, 5 testing results of acquisition is subjected to image co-registration according to the fusion method of step 1, obtain edge inspection when scale n=3
Survey result G3；
The edge detection and fusion that according to above-mentioned steps 5 structural elements are carried out with n scale, obtain melting for n different scale
It closes image and is respectively as follows: G1 ..., Gn；By 5 kinds of structural element Image erosions as a result, according to adaptive weighting, one is reconfigured out
New corrosion grayscale image G'(x, y)；By G'(x, y), two width Binary Sketch of Grey Scale Image of f (x, y) is carried out by the method for connected domain
Image-region is erased, and the new image GL (x, y) of two width, fL (x, y) are obtained；Pass through formula F (x, y)=fL (x, y)-GL (x, y)
Finally obtain the edge image F (x, y) of Multiscale Morphological processing.
4. the internal wave of ocean detection algorithm according to claim 3 based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, special
Sign is, constructs 5 kind of 3 × 3 structural element, and the linear structure element including 4 directions respectively represents 0 °, 45 °, 90 ° and 135 °
Direction and 1 disc structure element；It is respectively indicated with matrix are as follows:
5. the internal wave of ocean detection algorithm according to claim 1 based on Method Based on Multi-Scale Mathematical Morphology Fusion Features, special
Sign is, step 4 specifically:
Two obtained width edge image application image fusion functions are subjected to image co-registration, i.e. pixel space is added and carries out properly
The closed operation that threshold value is chosen, obtains final fusion results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810960749.9A CN109410228A (en) | 2018-08-22 | 2018-08-22 | Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810960749.9A CN109410228A (en) | 2018-08-22 | 2018-08-22 | Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109410228A true CN109410228A (en) | 2019-03-01 |
Family
ID=65463672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810960749.9A Pending CN109410228A (en) | 2018-08-22 | 2018-08-22 | Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109410228A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109741287A (en) * | 2018-12-27 | 2019-05-10 | 湖南国科微电子股份有限公司 | Image orientation filtering method and device |
| CN110033471A (en) * | 2019-04-19 | 2019-07-19 | 福州大学 | A kind of wire detection method based on connected domain analysis and morphological operation |
| CN110118640A (en) * | 2019-05-14 | 2019-08-13 | 大连理工大学 | A kind of method that interior estimates feature in strong stratified fluid is extracted in laboratory |
| CN110147716A (en) * | 2019-04-02 | 2019-08-20 | 北京理工雷科电子信息技术有限公司 | Wave method for detecting area in a kind of SAR image combined based on frequency domain with airspace |
| CN110443821A (en) * | 2019-07-09 | 2019-11-12 | 成都理工大学 | Water body detection method and water body detection device based on image |
| CN110796086A (en) * | 2019-10-30 | 2020-02-14 | 中国科学院宁波工业技术研究院慈溪生物医学工程研究所 | Iris segmentation method of AS-OCT image based on local phase tensor algorithm |
| CN111241862A (en) * | 2020-01-21 | 2020-06-05 | 西安邮电大学 | Bar code positioning method based on edge characteristics |
| CN111563859A (en) * | 2020-05-19 | 2020-08-21 | 安徽大学 | Highlight removing method of double-edge-preserving filter |
| CN112862760A (en) * | 2021-01-19 | 2021-05-28 | 浙江大学 | Bearing outer ring surface defect area detection method |
| CN114994671A (en) * | 2022-05-31 | 2022-09-02 | 南京慧尔视智能科技有限公司 | Target detection method, device, equipment and medium based on radar image |
| CN116630225A (en) * | 2023-03-13 | 2023-08-22 | 中铁大桥局集团有限公司 | Method and device for identifying underwater foundation damage of railway bridge and processing equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916373A (en) * | 2010-08-11 | 2010-12-15 | 西安电子科技大学 | Road semiautomatic extraction method based on wavelet detection and ridge line tracking |
| CN106447686A (en) * | 2016-09-09 | 2017-02-22 | 西北工业大学 | Method for detecting image edges based on fast finite shearlet transformation |
| US20170337687A1 (en) * | 2015-11-30 | 2017-11-23 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for computer aided diagnosis |
-
2018
- 2018-08-22 CN CN201810960749.9A patent/CN109410228A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916373A (en) * | 2010-08-11 | 2010-12-15 | 西安电子科技大学 | Road semiautomatic extraction method based on wavelet detection and ridge line tracking |
| US20170337687A1 (en) * | 2015-11-30 | 2017-11-23 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for computer aided diagnosis |
| CN106447686A (en) * | 2016-09-09 | 2017-02-22 | 西北工业大学 | Method for detecting image edges based on fast finite shearlet transformation |
Non-Patent Citations (2)
| Title |
|---|
| 徐剑锋 等: "基于小波变换和多尺度形态学的图像边缘检测", 《设计与实现》 * |
| 闫敬文 等: "基于形态学小波变换的SAR图像数据海洋目标提取", 《工程图学学报》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109741287A (en) * | 2018-12-27 | 2019-05-10 | 湖南国科微电子股份有限公司 | Image orientation filtering method and device |
| CN110147716B (en) * | 2019-04-02 | 2021-02-19 | 北京理工雷科电子信息技术有限公司 | SAR image internal wave region detection method based on combination of frequency domain and airspace |
| CN110147716A (en) * | 2019-04-02 | 2019-08-20 | 北京理工雷科电子信息技术有限公司 | Wave method for detecting area in a kind of SAR image combined based on frequency domain with airspace |
| CN110033471A (en) * | 2019-04-19 | 2019-07-19 | 福州大学 | A kind of wire detection method based on connected domain analysis and morphological operation |
| CN110033471B (en) * | 2019-04-19 | 2022-09-13 | 福州大学 | Frame line detection method based on connected domain analysis and morphological operation |
| CN110118640A (en) * | 2019-05-14 | 2019-08-13 | 大连理工大学 | A kind of method that interior estimates feature in strong stratified fluid is extracted in laboratory |
| CN110118640B (en) * | 2019-05-14 | 2020-06-02 | 大连理工大学 | Method for extracting internal solitary wave characteristics in strong stratified fluid in laboratory |
| CN110443821A (en) * | 2019-07-09 | 2019-11-12 | 成都理工大学 | Water body detection method and water body detection device based on image |
| CN110443821B (en) * | 2019-07-09 | 2023-05-05 | 成都理工大学 | Image-based water body detection method and device |
| CN110796086A (en) * | 2019-10-30 | 2020-02-14 | 中国科学院宁波工业技术研究院慈溪生物医学工程研究所 | Iris segmentation method of AS-OCT image based on local phase tensor algorithm |
| CN111241862A (en) * | 2020-01-21 | 2020-06-05 | 西安邮电大学 | Bar code positioning method based on edge characteristics |
| CN111241862B (en) * | 2020-01-21 | 2023-06-02 | 西安邮电大学 | Bar code positioning method based on edge characteristics |
| CN111563859A (en) * | 2020-05-19 | 2020-08-21 | 安徽大学 | Highlight removing method of double-edge-preserving filter |
| CN112862760A (en) * | 2021-01-19 | 2021-05-28 | 浙江大学 | Bearing outer ring surface defect area detection method |
| CN112862760B (en) * | 2021-01-19 | 2023-11-10 | 浙江大学 | Bearing outer ring surface defect area detection method |
| CN114994671A (en) * | 2022-05-31 | 2022-09-02 | 南京慧尔视智能科技有限公司 | Target detection method, device, equipment and medium based on radar image |
| CN114994671B (en) * | 2022-05-31 | 2023-11-28 | 南京慧尔视智能科技有限公司 | Target detection method, device, equipment and medium based on radar image |
| CN116630225A (en) * | 2023-03-13 | 2023-08-22 | 中铁大桥局集团有限公司 | Method and device for identifying underwater foundation damage of railway bridge and processing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109410228A (en) | Internal wave of ocean detection algorithm based on Method Based on Multi-Scale Mathematical Morphology Fusion Features | |
| WO2018107939A1 (en) | Edge completeness-based optimal identification method for image segmentation | |
| CN104361582B (en) | Method of detecting flood disaster changes through object-level high-resolution SAR (synthetic aperture radar) images | |
| Yu et al. | A new edge detection approach based on image context analysis | |
| US8116522B1 (en) | Ship detection system and method from overhead images | |
| NagaRaju et al. | Morphological edge detection algorithm based on multi-structure elements of different directions | |
| Kaspers | Blob detection | |
| Celik et al. | Unsupervised colour image segmentation using dual-tree complex wavelet transform | |
| CN109544571A (en) | A kind of metallic phase image edge detection method based on mathematical morphology | |
| CN115861359B (en) | Self-adaptive segmentation and extraction method for water surface floating garbage image | |
| CN113837974A (en) | NSST (non-subsampled contourlet transform) domain power equipment infrared image enhancement method based on improved BEEPS (Bayesian particle swarm optimization) filtering algorithm | |
| Niu et al. | Geometrical-based generative adversarial network to enhance digital rock image quality | |
| CN114764801A (en) | Weak and small ship target fusion detection method and device based on multi-vision significant features | |
| Zhaohui et al. | Water body extraction from multi-source satellite images | |
| Li et al. | Automatic infrared ship target segmentation based on structure tensor and maximum histogram entropy | |
| Shen et al. | Segmenting multiple overlapping nuclei in H&E stained breast cancer histopathology images based on an improved watershed | |
| CN108242060A (en) | A kind of method for detecting image edge based on Sobel operators | |
| CN112163606B (en) | Infrared small target detection method based on block contrast weighting | |
| Deepa et al. | Improved watershed segmentation for apple fruit grading | |
| Manjula | Image edge detection and segmentation by using histogram thresholding method | |
| Yan-Li | Introduction to digital image pre-processing and segmentation | |
| Li et al. | Algorithm of Canny Operator Edge Pre-processing Based on Mathematical Morphology | |
| Chen et al. | A study of image segmentation algorithms combined with different image preprocessing methods for thyroid ultrasound images | |
| Wang et al. | Image segmentation method for an illumination highlight region of interior design effects based on the partial differential equation | |
| Mehrara et al. | Quad-pixel edge detection using neural network |
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: 20190301 |