CN107610188A - A kind of synchronous compression warp wavelet method of figure identification separation - Google Patents
A kind of synchronous compression warp wavelet method of figure identification separation Download PDFInfo
- Publication number
- CN107610188A CN107610188A CN201710681682.0A CN201710681682A CN107610188A CN 107610188 A CN107610188 A CN 107610188A CN 201710681682 A CN201710681682 A CN 201710681682A CN 107610188 A CN107610188 A CN 107610188A
- Authority
- CN
- China
- Prior art keywords
- image
- separation
- warp wavelet
- synchronous compression
- point
- 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
Abstract
The invention provides a kind of synchronous compression warp wavelet method of figure identification separation, methods described includes step：Image is pre-processed；Compression warp wavelet is synchronized to pretreated image and obtains corresponding bent wave component；Cluster separation is carried out to obtained bent wave component；To bent wave component march ripple inverse transformation also each composition of original image after separation.The invention provides a kind of synchronous compression warp wavelet method of figure identification separation, the wave vector composition of each position in image is analyzed by using synchronous compression algorithm, it is achieved thereby that the quantitative expression to image orientation information.Synchronous compression is combined with warp wavelet, has filled up the vacancy that directional information describes in warp wavelet.Realize has preferable analytical effect to the line Strange properties of image, and preferable rarefaction representation can be realized to the composition in image with anisotropic.Meanwhile this method possesses certain anti-noise ability in itself, be advantageous to carry out more Accurate Analysis to destination object.
Description
Technical field
The present invention relates to computer vision field, and in particular to a kind of synchronous compression warp wavelet side of figure identification separation
Method.
Background technology
Wavelet transformation has obtained the attention and favor of Research on Time Frequency person with the advantage of its multiscale analysis.But small echo
Conversion does not have direction factor, therefore occurs phenomena such as edge blurry in the analysis of 2-D data, and analytical effect is simultaneously paid no attention to
Think.
The content of the invention
In order to solve the above problems, the synchronous compression warp wavelet method separated is identified the invention provides a kind of figure,
Compression warp wavelet is firstly introduced into, then carries out wave vector cluster separation, last march ripple inverse transformation, so as to effectively solve again
Above mentioned problem.
Technical scheme provided by the invention is：A kind of synchronous compression warp wavelet method of figure identification separation, the side
Method includes step：Image is pre-processed；Compression warp wavelet is synchronized to pretreated image and obtains corresponding song
Wave component；Cluster separation is carried out to obtained bent wave component；To the bent wave component march ripple inverse transformation also original image after separation
Each composition.
The beneficial effects of the invention are as follows：The invention provides a kind of synchronous compression warp wavelet side of figure identification separation
Method, the wave vector composition of each position in image is analyzed by using synchronous compression algorithm, it is achieved thereby that to figure
As the quantitative expression of directional information.Synchronous compression is combined with warp wavelet, directional information in warp wavelet has been filled up and has described
Vacancy.Realize has preferable analytical effect to the line Strange properties of image, to having the composition of anisotropic in image
Preferable rarefaction representation can be realized.Meanwhile this method possesses certain anti-noise ability in itself, be advantageous to carry out destination object
More Accurate Analysis.
Brief description of the drawings
Fig. 1 is the overall flow figure of the synchronous compression warp wavelet of figure identification separation in first embodiment of the invention；
Fig. 2 is that the specific steps flow for synchronizing compression warp wavelet in first embodiment of the invention to image is illustrated
Figure；
Fig. 3 is that v is used in first embodiment of the inventionfThe real part of (a, θ, b)Bent wave energy is carried out
The specific steps flow chart redistributed；
Fig. 4 is the specific steps flow chart that first embodiment of the invention to obtained bent wave component cluster separation；
Fig. 5 is race and the step flow chart of separation definition of the point set of first embodiment of the invention；
Fig. 6 is the warp wavelet time-frequency schematic diagram of second embodiment of the invention；
Fig. 7 is the sinusoidal curve ripple schematic diagram of second embodiment of the invention；
Fig. 8 is the sinusoidal curve ripple schematic diagram under second embodiment of the invention different parameters；
Fig. 9 is the sinusoidal curve reconstructed wave schematic diagram under second embodiment of the invention different parameters；
Figure 10 is noise suppressed schematic diagram in second embodiment of the invention；
Figure 11 is to choose proper angle parameter in second embodiment of the invention to carry out image separation schematic diagram.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing to embodiment party of the present invention
Formula is further described, the particular technique details hereinafter mentioned, such as：Method, equipment etc., are only better understood from reader
Technical scheme, does not represent that present invention is limited only by following ins and outs.
The first embodiment of the present invention provides a kind of synchronous compression warp wavelet method of figure identification separation, passes through profit
The wave vector composition of each position in image is analyzed with synchronous compression algorithm, realizes the amount to image orientation information
Change expression, there is preferable analytical effect to the line Strange properties of image, can be real to the composition in image with anisotropic
Now preferable rarefaction representation.Referring to Fig. 1, Fig. 1 is the synchronous compression Qu Bo of figure identification separation in first embodiment of the invention
The overall flow figure of conversion, specific steps include：
S101：Image is pre-processed, specifically included：Intercept valid data；By the irregular data in valid data
Normalization；Edge-smoothing processing is carried out to the data after normalization.
S102：Compression warp wavelet is synchronized to pretreated image and obtains corresponding bent wave component.
S103：Cluster separation is carried out to obtained bent wave component.
S104：To bent wave component march ripple inverse transformation also each composition of original image after separation, the concrete form after reduction
For
The synchronous compression warp wavelet method storage of the figure identification separation and calculating cost are higher.Assuming that given number
Size according to f (x) is L × L, wherein there is the component that K wave number is O (L).The energy of each synchronous compressionSurround
It is distributed in around its 2D wave vector, distance isIt therefore meets TF (v, b)>=δ non-zero mesh point is in 4D phase spaces
Classification sum be KL4∈, it is seen that cluster is larger and impracticable.In order to reduce cost, we can be first in each position
On similar clustering is applied to wave vector so that each position obtains O (K) individual cluster.Then clustering method is applied to again
Size is O (KL after simplifying in 4D phase spaces2) point concentrate.
Referring to Fig. 2, Fig. 2 is the specific steps stream for synchronizing compression warp wavelet in first embodiment of the invention to image
Journey schematic diagram, is specifically included：
S201：Carry out warp wavelet and obtain object function f (x) bent ripple expression formula and gradient.Wherein, the Qu Bobiao reaches
Formula is Wf(a, θ, b), gradient are
S202：The f (x) is the K superpositions with obvious curve singularity image.
S203：Calculate local wave vector estimation and the energy concentrated on around it.Wherein, the local wave vector is estimated asEnergy is Tf(a, θ, b).In above formula, and a ∈ [1, ∞) represent amplitude yardstick；The θ
∈ [0,2 π) represent deflection；The b represents the coordinate of two-dimensional spatial location；(a, θ) ∈ p, b ∈ B and Wf(a, θ, b) ≠ 0；Institute
The point that P is one group of Fourier domain is stated, Fourier domain is：[- L/2, L/2)2, L is domain scale size；The B is discrete uniform
Mesh space：LB×LB, LBFor space scale size, B={ (n1/LB, n2/LB)：0≤n1, n2＜ LB, n1, n2∈Z}。
S204：Bent wave energy is redistributed using the real part that local wave vector is estimated.Bent wave energy after distribution
ForWhereinL is two-dimensional grid chi
Size is spent, L is domain scale size；The value of the s and t are 1/2<s<t<1.
With continued reference to Fig. 3, Fig. 3 is that v is used in first embodiment of the inventionfThe real part of (a, θ, b)It is right
The specific steps flow chart that bent wave energy is redistributed, comprises the following steps：
S301：Build two-dimentional Descartes's network that step-length is Δ.
S302：Discrete fourier domain variable v meets V={ (n1Δ,n2Δ):n1,n2∈
Z}。
S303：By each v=(n1Δ, n2Δ) ∈ V and cell using v as grid element center
It is associated.
Referring to Fig. 4, Fig. 4 is the specific steps stream that first embodiment of the invention to obtained bent wave component cluster separation
Cheng Tu, specifically include：
S401：Select point set S to be sorted.
S402：Set the threshold value d in 2D spatial domains0, vector radius threshold R0And angle threshold θ0。
S403：Obtain K cluster result S1,S2,...,SK。
S404：K class S1,S2,...,SKMeet each SiAll it is to meet (d0, θ0, R0) race.
S405：Different class SiAnd SjIn condition (d0, θ0, R0) under separate.
S406：It is describedNear the phase space related positioned at K image.
S407：According to K classification results Wf(a, θ, b) will be separated into
Referring to Fig. 5, Fig. 5 is race and the step flow chart of separation definition of the point set of first embodiment of the invention, specific to wrap
Include：
S501：Any point p in 4D spaces is described as (xp, ap, θp)。
S502：The xpIt is 2D spatial domains midpoint p coordinate, apFor point p wave vector radius, θpPressed from both sides for point p wave vector
Angle.
S503：(ap cosθp, ap sinθp) for point p wave vector projection.
S504：If point p and point q meet condition | xp-xq|≤d0, | ap-aq|≤R0And
min{|θp-θq|, 2 π-| θp-θq|}≤θ0It is to meet condition (d a pair then to think (p, q)0, θ0, R0) consecutive points.
S505：IfSo that (p1, q1), (qn, p2), (qi, qi+ 1) for i=1,
2 ..., n-1 is to meet condition (d0, θ0, R0) consecutive points, then it is to meet condition (d to say point set S0, θ0, R0) cluster.
S506：IfWithWithout satisfaction (d0, θ0, R0) consecutive points pair, then claim point set S1And S2In bar
Part (d0, θ0, R0) under separate.
By performing the first embodiment of the present invention, all technical characteristics in the claims in the present invention are obtained for detailed
Illustrate.
Prior art is different from, the embodiment provides a kind of synchronous compression warp wavelet of figure identification separation
Method, the wave vector composition of each position in image is analyzed by using synchronous compression algorithm, it is achieved thereby that right
The quantitative expression of image orientation information.Synchronous compression is combined with warp wavelet, directional information in warp wavelet has been filled up and has retouched
The vacancy stated.Realize has preferable analytical effect to the line Strange properties of image, in image have anisotropic into
Preferable rarefaction representation can be realized by dividing.
The second embodiment of the present invention and first embodiment are identical in method, but second embodiment is with specific actual
The data verification validity and practicality of the present invention program.Referring to Fig. 6, Fig. 6 is the warp wavelet of second embodiment of the invention
Time-frequency schematic diagram, is specifically included：Frequency axis 601, time shaft 602, amplitude axis 603, time domain song ripple 604 and frequency domain superposition ripple 605.
Wherein, yardstick a=2, direction θ=9, position b=(135,50).It is worth noting that, in yardstick a=1, i.e., under thick yardstick,
Bent wave function does not possess directionality, is now equal to wavelet transformation.
Referring to Fig. 7, Fig. 7 is the sinusoidal curve ripple schematic diagram of second embodiment of the invention, including high frequency waves 701 and low frequency wave
702。
Referring to Fig. 8, Fig. 8 is the sinusoidal curve ripple schematic diagram under second embodiment of the invention different parameters.Wherein, sine wave
The yardstick of bent ripple 801 is 1, angle 1；The yardstick of sinusoidal curve ripple 802 is 2, angle 3；The yardstick of sinusoidal curve ripple 803 is
3, angle 7；The yardstick of sinusoidal curve ripple 804 is 4, angle 10；The yardstick of sinusoidal curve ripple 805 is 5, angle 28.
Referring to Fig. 9, Fig. 9 is the sinusoidal curve reconstructed wave schematic diagram under second embodiment of the invention different parameters.Wherein, just
The yardstick of string curve ripple 901 is 1；The yardstick of sinusoidal curve ripple 902 is 2；The yardstick of sinusoidal curve ripple 903 is 3；Sinusoidal curve ripple
904 yardstick is 4；The yardstick of sinusoidal curve ripple 905 is 5.As can be seen that when yardstick is relatively low, what it showed is image
Profile information, i.e., thick dimensional information.With the increase of yardstick, its information shown then increasingly details, and thin dimensional information.
And the part that can be seen that two different frequencies for two images corresponding to 2 and 3 from yardstick is successfully separated.
In bent wave analysis, usual noise signal all possesses bent wave component in all directions, and wave vector is larger.Cause
This, in order to improve the noise robustness of algorithm, the present invention adds the mechanism of threshold denoising during image separates.
Threshold value δ >=0 is defined, is met for energy size
Tf(v, b) >=δ, v ∈ V, b ∈ B
Bent wave component, we are ignored.Peel off the composition larger with wave vector for feature Distribution value in addition
|vf(a, θ, b) | ＞ R
We are regarded as noise contribution and ignored.Wherein R is image radius.Referring to Figure 10, Figure 10 is the present invention
Noise suppressed schematic diagram in second embodiment, including：Noise suppression effect 1001, threshold value when threshold value is 0.001 are
Making an uproar when noise suppression effect 1003 and threshold value when noise suppression effect 1002, threshold value when 0.002 are 0.003 are 0.004
Sound inhibition 1004.Choose suitable angle parameter, it becomes possible to be successfully separated image.With continued reference to Figure 11, Figure 11 is this
Proper angle parameter is chosen in invention second embodiment and carries out image separation schematic diagram, including high frequency waves 1101 and low frequency wave
1102.It can be seen that separating effect is preferable.
It is different from the first embodiment of the present invention, the second embodiment of the present invention is by introducing the example of specific sine wave
Son, the validity and practicality of technical solution of the present invention are demonstrated from the angle of reality.Meanwhile the second embodiment of the present invention
The method for demonstrating the present invention possesses certain anti-noise ability, is advantageous to carry out more Accurate Analysis to destination object.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.
Claims (10)
- A kind of 1. synchronous compression warp wavelet method of figure identification separation, it is characterised in that：Comprise the following steps：Image is entered Row pretreatment；Compression warp wavelet is synchronized to pretreated image and obtains corresponding bent wave component；To obtained Qu Bo Composition carries out cluster separation；To bent wave component march ripple inverse transformation also each composition of original image after separation.
- A kind of 2. synchronous compression warp wavelet method of figure identification separation as claimed in claim 1, it is characterised in that：It is described Carrying out pretreatment to image includes：Intercept valid data；Irregular data in valid data is normalized；After normalization Data carry out edge-smoothing processing.
- A kind of 3. synchronous compression warp wavelet method of figure identification separation as claimed in claim 1, it is characterised in that：It is described Compression warp wavelet is synchronized to image to be included：Carry out warp wavelet and obtain object function f (x) bent ripple expression formula Wf(a, θ, b) and gradientThe f (x) is the K superpositions with obvious curve singularity image；Calculate local ripple Vector estimates vf(a, θ, b) and the energy T concentrated on around itf(a, θ, b)；In above formula, a ∈ [1, ∞] represent amplitude chi Degree；The θ ∈ [0,2 π) represent deflection；The b represents the coordinate of two-dimensional spatial location.
- A kind of 4. synchronous compression warp wavelet method of figure identification separation as claimed in claim 3, it is characterised in that：It is described Local wave vector is estimated as(a, θ) ∈ P, b the ∈ B and Wf(a, θ, b) ≠ 0；The P For the point of one group of Fourier domain, Fourier domain is：[- L/2, L/2)2, L is domain scale size；The B is discrete uniform grid Space：LB×LB, LBFor space scale size, B={ (n1/LB, n2/LB)：0≤n1, n2＜ LB, n1, n2∈Z}。
- A kind of 5. synchronous compression warp wavelet method of figure identification separation as claimed in claim 4, it is characterised in that：Use vfThe real part of (a, θ, b)Bent wave energy is redistributed to obtainIt is describedL is two-dimensional grid Scale size, L are domain scale size；The value of the s and t are 1/2<s<t<1.
- A kind of 6. synchronous compression warp wavelet method of figure identification separation as claimed in claim 5, it is characterised in that：Described ArriveSpecifically include：Build two-dimentional Descartes's network that step-length is Δ；From Dissipate Fourier variable v and meet V={ (n1Δ,n2Δ):n1,n2∈Z}；By each v=(n1Δ, n2Δ) ∈ V with using v as grid The cell at centerIt is associated.
- A kind of 7. synchronous compression warp wavelet method of figure identification separation as claimed in claim 1, it is characterised in that：It is described Cluster separation is carried out to obtained bent wave component to comprise the following steps：Select point set S to be sorted；Set the threshold value in 2D spatial domains d0, vector radius threshold R0And angle threshold θ0；Obtain K cluster result S1,S2,...,SK；K class S1,S2,...,SKMeet Each SiAll it is to meet (d0, θ0, R0) race；Different class SiAnd SjIn condition (d0, θ0, R0) under separate；It is describedNear the phase space related positioned at K image；According to K classification results Wf(a, θ, b) will divide From for
- A kind of 8. synchronous compression warp wavelet method of figure identification separation as claimed in claim 7, it is characterised in that：It is described Each SiAll it is to meet (d0, θ0, R0) race and different class SiAnd SjIn condition (d0, θ0, R0) under the definition that separates include it is following Step：Any point p in 4D spaces is described as (xp, ap, θp)；The xpIt is 2D spatial domains midpoint p coordinate, apFor point p Wave vector radius, θpFor point p wave vector angle；(ap cosθp, ap sinθp) for point p wave vector projection；If point p and Point q meets condition | xp-xq|≤d0, | ap-aq|≤R0And min | θp-θq|, 2 π-| θp-θq|}≤θ0It is a pair then to think (p, q) Meet condition (d0, θ0, R0) consecutive points；IfSo that (p1, q1), (qn, p2), (qi, qi+ 1) for i=1,2 ..., n-1 is to meet condition (d0, θ0, R0) consecutive points, then it is to meet condition (d to say point set S0, θ0, R0) Cluster；IfWithWithout satisfaction (d0, θ0, R0) consecutive points pair, then claim point set S1And S2In condition (d0, θ0, R0) under separate.
- A kind of 9. synchronous compression warp wavelet method of figure identification separation as claimed in claim 1, it is characterised in that：It is described right The concrete form of bent wave component march ripple inverse transformation also each composition of original image after separation is：
- A kind of 10. synchronous compression warp wavelet method of figure identification separation as claimed in claim 1, it is characterised in that： The cluster of each position is obtained to wave vector application clustering procedure on each position, cluster separation method is then applied to 4D phases again Point after simplifying in space is concentrated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710681682.0A CN107610188A (en) | 2017-08-10 | 2017-08-10 | A kind of synchronous compression warp wavelet method of figure identification separation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710681682.0A CN107610188A (en) | 2017-08-10 | 2017-08-10 | A kind of synchronous compression warp wavelet method of figure identification separation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107610188A true CN107610188A (en) | 2018-01-19 |
Family
ID=61064988
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710681682.0A Pending CN107610188A (en) | 2017-08-10 | 2017-08-10 | A kind of synchronous compression warp wavelet method of figure identification separation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107610188A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108519619A (en) * | 2018-04-11 | 2018-09-11 | 中国石油大学(华东) | Based on the time-frequency analysis technology for becoming privileged warp wavelet |
| CN109410178A (en) * | 2018-09-28 | 2019-03-01 | 武汉科技大学 | A kind of workpiece crack detection method and system |
| CN109917459A (en) * | 2017-12-12 | 2019-06-21 | 中国石油天然气股份有限公司 | A kind of method, apparatus and system for suppressing seismic noise |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103905831A (en) * | 2014-04-18 | 2014-07-02 | 西安电子科技大学 | Strip wave transform image compression method based on graphics processor |
| CN105302773A (en) * | 2015-09-18 | 2016-02-03 | 河海大学 | Wavelet transformation method based on synchronization compression fractional order |
-
2017
- 2017-08-10 CN CN201710681682.0A patent/CN107610188A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103905831A (en) * | 2014-04-18 | 2014-07-02 | 西安电子科技大学 | Strip wave transform image compression method based on graphics processor |
| CN105302773A (en) * | 2015-09-18 | 2016-02-03 | 河海大学 | Wavelet transformation method based on synchronization compression fractional order |
Non-Patent Citations (1)
| Title |
|---|
| HAIZHAO YANG等: "Synchrosqueezed Curvelet Transform for 2D Mode Decomposition", 《SIAM JOURNAL ON MATHEMATICAL ANALYSIS》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109917459A (en) * | 2017-12-12 | 2019-06-21 | 中国石油天然气股份有限公司 | A kind of method, apparatus and system for suppressing seismic noise |
| CN108519619A (en) * | 2018-04-11 | 2018-09-11 | 中国石油大学(华东) | Based on the time-frequency analysis technology for becoming privileged warp wavelet |
| CN109410178A (en) * | 2018-09-28 | 2019-03-01 | 武汉科技大学 | A kind of workpiece crack detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Looney et al. | Multiscale image fusion using complex extensions of EMD | |
| US9706298B2 (en) | Method and apparatus for localization of an acoustic source and acoustic beamforming | |
| CN107610188A (en) | A kind of synchronous compression warp wavelet method of figure identification separation | |
| CN109859126A (en) | A kind of vedio noise reduction method, apparatus, electronic equipment and storage medium | |
| CN102663695A (en) | DR image denoising method based on wavelet transformation and system thereof | |
| CN110082135A (en) | Equipment fault recognition methods, device and terminal device | |
| Patel et al. | Separated component-based restoration of speckled SAR images | |
| WO2021057309A1 (en) | Tracked target determination method and related device | |
| CN102945548A (en) | Directional pyramid filtering-based image processing method and device | |
| CN104637494A (en) | Double-microphone mobile equipment voice signal enhancing method based on blind source separation | |
| CN105580074B (en) | Signal processing system and method | |
| Rao et al. | Satellite image fusion using fast discrete curvelet transforms | |
| US20150312663A1 (en) | Source separation using a circular model | |
| Pradeep et al. | VLSI Implemetation of Euler Number Computation and Stereo Vision Concept for CORDIC based Image Registration | |
| CN104182944A (en) | Optical image denoising method based on serial connection of curvelet transform and wavelet transform | |
| CN104978720A (en) | Video image raindrop removal method and apparatus | |
| CN110458746A (en) | Add method and device, the electronic equipment of watermark | |
| CN108445452A (en) | The sound bearings improved frequency domain SRP method of estimation | |
| CN103747268B (en) | A kind of layered self-adapting threshold video denoising method | |
| CN103513288A (en) | Compensation direction filtering method of two-dimensional grid data | |
| Tuyet et al. | Reducing impurities in medical images based on curvelet domain | |
| Wang et al. | Blind source separation of electromagnetic detection signals | |
| Li et al. | Weak signal blind source separation in passive radar system with strong interference | |
| Mazur et al. | A new clustering approach for solving the permutation problem in convolutive blind source separation | |
| Heimann et al. | Jointly Resampling and Reconstructing Corrupted Images for Image Classification |
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: 20180119 |