CN106548183B - A method of edge extracting quickly being carried out to fuzzy object - Google Patents
A method of edge extracting quickly being carried out to fuzzy object Download PDFInfo
- Publication number
- CN106548183B CN106548183B CN201610954078.6A CN201610954078A CN106548183B CN 106548183 B CN106548183 B CN 106548183B CN 201610954078 A CN201610954078 A CN 201610954078A CN 106548183 B CN106548183 B CN 106548183B
- Authority
- CN
- China
- Prior art keywords
- edge
- gradient
- information
- image
- partial derivative
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of methods for quickly carrying out edge extracting to fuzzy object, by light source after spatial light modulator SLM modulation, it is incident upon in target, after target reflection/transmission, detector receives a secondary fuzzy image information under the action of scattering medium, the displacement on four direction is carried out according to Roberts and gradient operator to the random speckle of projection, detector records the partial derivative that four data calculate image all directions respectively, after multiple repairing weld, carry out double velocity correlation calculating, obtain the strong related information of four partial derivatives, find out gradient information, finally two gradient informations are added, reach and the inclined marginal information of any direction is quickly rebuild.The present invention reduces reconstruction information quantity, reduces the difficulty to target identification, improves the arithmetic speed of system；Blurred picture can be accurately identified；Sampling number needed for rebuilding target image is low, reduces system processes data quantity, it is easier to hardware realization.
Description
Technical field
The invention belongs to image processing edges to extract field, especially a kind of quickly to carry out edge extracting to fuzzy object
Method.
Background technique
As environmental pollution is got worse, haze weather is gone on a journey to people to live and be impacted in city.At present tradition at
As although technology is mature, to image deblurring processing without quickly and effectively algorithm.Ghost imaging is since it is with anti-atmosphere
The characteristic of disturbance has obtained the extensive concern of domestic and foreign scholars.Ghost imaging calculating speed is fast, and anti-scattering ability is strong, but due to
Its high sampling rate and low signal-to-noise ratio become the key factor for restricting ghost image development.
In recent years, research team both domestic and external reduces picture noise, improves the figure of ghost image mainly to atmospheric turbulance is resisted
Image quality amount expansion research, such as difference ghost imaging (DGI), normalization ghost imaging (NGI), iterated denoising ghost imaging (IDGI) and time
(TCDGI) etc. is imaged in related differential ghost.These algorithms are although a degree of to improve picture quality, but increases algorithm
Computation complexity, and nyquist sampling rate is also far longer than 100%, and image reconstruction speed is not improved.2015,
Xue-feng Liu et al. proposes that (28Dec 2015, Vol.23, No.26, DOI:10.1364/ is imaged in gradient operator ghost
OE.23.033802, OPTICSEXPRESS 33803) edge of target is reconfigured quickly, it reduces needed for reconstruction image
Sample rate, but the edge contour of the target of this method selection all possesses special angle, it can not be to the inclined side of any direction
Edge extracts.2016, the imaging of Tianyi Mao proposition displacement speckle ghost (Volume 8, Number 4, August2016,
DOI:10.1109/JPHOT.2016.2578934), profile information at any angle can be extracted, but algorithm is multiple
Miscellaneous, sample rate ratio GGI is greatly improved, and increases system data storage burden.
Summary of the invention
The invention reside in a kind of method for quickly carrying out edge extracting to fuzzy object is provided, allow computer to fuzzy
The identification of targeted cache high quality.
The technical solution for realizing the aim of the invention is as follows: a method of edge extracting quickly being carried out to fuzzy object,
The displacement on four direction is carried out according to Roberts and gradient operator to the random speckle of projection first, is successively incident upon target
On, detector records the partial derivative that four data calculate image all directions respectively；Then after multiple repairing weld, second order pass is carried out
Online is calculated, and is obtained the strong related information of four partial derivatives, is found out gradient information, is finally added two gradient informations, is reached pair
The inclined marginal information of any direction is rebuild.
The present invention compared with prior art, remarkable advantage are as follows: (1) reduce reconstruction information quantity, reduce to target know
Other difficulty improves the arithmetic speed of system.(2) compared with tradition imaging, there is anti-scattering ability, mould can be accurately identified
Paste image.(3) sampling number needed for rebuilding target image is low, reduces system processes data quantity, it is easier to hardware realization.
The present invention is described in further detail below in conjunction with the accompanying drawings.
Detailed description of the invention
Fig. 1 is the flow chart for the method that the present invention quickly carries out edge extracting to fuzzy object.
Fig. 2 is the random speckle direction of displacement schematic diagram for the method that the present invention quickly carries out edge extracting to fuzzy object.
The effect that Fig. 3 base of the present invention quickly carries out the method for edge extracting to fuzzy object shows figure, wherein figure (a) is original
Beginning target image (rice image), figure (b) are the blurred picture (blurred picture) under scattering process, and figure (c) is to extract image
(edge extracting effect).
Fig. 4 is the present invention and gradient operator ghost imaging algorithm Contrast on effect, and figure (a) is double slit target, schemes (b) for gradient calculation
Sub- ghost imaging algorithm extracts object edge effect, and figure (c) is that the present invention extracts object edge effect.
Fig. 5 is the signal-to-noise ratio comparison diagram that the present invention extracts image SNR with displacement speckle ghost imaging under same case.
Specific embodiment
The method that the present invention quickly carries out edge extracting in the approach application of edge extracting image procossing to fuzzy object,
By rebuilding object edge information, in conjunction with the principle for calculating ghost imaging, light source, will after spatial light modulator (SLM) modulation
Speckle after four direction displacement is incident upon in target, and detector separately detects four groups of data and storing data is used as and once adopts
Sample.After target reflects (transmission), detector receives a secondary fuzzy image information under the action of scattering medium.To throwing
The random speckle penetrated carries out the displacement on four direction according to Roberts and gradient operator, and detector records four data respectively
The partial derivative of image all directions is calculated, image border is extracted.After multiple repairing weld, double velocity correlation calculating is carried out, obtains four
The strong related information of partial derivative, finds out gradient information, is finally added two gradient informations, reaches to the inclined side of any direction
Edge information reconstruction, to rebuild the edge image of target.For double slit target ((a) of Fig. 4), traditional gradient operator ghost imaging
Algorithm is shown in (b) of Fig. 4 for the effect of double slit target, and (c) for effect part Fig. 4 that the present invention obtains.
In conjunction with Fig. 1, the specific implementation step that the present invention quickly carries out the method for edge extracting to fuzzy object is as follows:
Step 1, n times sampling is carried out for the blurred picture of m × n, will generate every time and projected and is in target random scattered
Spot relative to projection speckle respectively to left and right, upper left, upper right be displaced a pixel (only by one picture of the speckle displacement of projection
Element.Calculating process is exactly the matrix of projection speckle to be moved a position in computer, and the second row becomes the first row, and the third line becomes second
Row ... the first row change last line), i.e., hypothesis picture size is m row n column, and sampling number n times sample k for arbitrarily primary,
Speckle displacement can be indicated by following equation:
In above-mentioned formula, x and y are pixel coordinate, SkRepresent the random speckle that kth time sampling generates, S1 kRepresent SkAlong the side x
To movement, S2 kRepresent SkIt moves in the y-direction, S3 kRepresent SkIt tilts 45 ° of directions along x to move, S4 kRepresent Sk45 ° of directions are tilted along y
Mobile, this is the description being specifically displaced, and is exactly SkTo the left, to the right, it is left tiltedly it is upper right tiltedly on movement, specific direction of displacement is as schemed
Shown in 2.
Step 2, displacement speckle projects target, is recorded by target by bucket detector, it may be assumed that
Di kBucket detector is represented in kth time sampling, the numerical value detected under i-th kind of random speckle displacement.Si kIt is step
S in 1kFour kinds displacement after speckle, i=1,2,3,4.Represent matrix dot product.O (x, y) represents target information matrix.
Step 3, the data obtained using step 2 calculate the partial derivative in image all directions, specific formula is as follows:
Formula (6) is the marginal information that all directions are calculated in kth time sampling, is calculated and is schemed using data obtained in step 2
As the partial derivative of all directions.Di edge(k)For the partial derivative of all directions kth time sampling.It is specific as shown in Figure 2.
Step 4, after n times sample, two secondary association operations is carried out to all directions partial derivative, extract related information.Tool
Body formula is as follows:
Formula (7) is secondary incidence formula, wherein SiTo project speckle on the image, SiDisplacement mode such as step 1
Middle Si kIdentical, formula (7) calculating detector records the strong associated data information of target all directions partial derivative in data, to rebuild side
Edge information image.EiFor the strong related information of calculated all directions partial derivative.
Step 5, two secondary association calculated results of all directions partial derivative according to obtained in step 4 calculate image two
The gradient of different directions.Specific formula is as follows:
Edge in formula1And Edge2For the gradient value for calculating image different directions.
Step 6, according to step 5 obtain as a result, by Edge1And Edge2Addition obtains final reconstruction image.It is specific public
Formula is as follows:
GIedge=Edge1+Edge2 (10)
GI in formula (10)edgeThe clearly object edge image as reconstructed.One secondary 256*256 size is obscured
Rice figure rebuild (in Fig. 3, (a) be target original image, (b) be blurred picture), rebuild effect such as Fig. 3 (c) shown in.Phase
For being displaced speckle ghost imaging algorithm, signal noise ratio (snr) of image is greatly improved, signal-to-noise ratio comparison such as Fig. 5.
Claims (5)
1. it is a kind of quickly to fuzzy object carry out edge extracting method, it is characterised in that first to the random speckle of projection according to
Roberts and gradient operator carry out the displacement on four direction, are successively incident upon in target, detector records four data respectively
Calculate the partial derivative of image all directions；Then after multiple repairing weld, double velocity correlation calculating is carried out, the strong of four partial derivatives is obtained
Related information finds out gradient information, is finally added two gradient informations, reaches to the inclined marginal information weight of any direction
It builds；
Specific step is as follows:
Step 1, n times sampling is carried out for the blurred picture of m × n, will generate every time and projects the random speckle in target point
Not to left and right, upper left, upper right are displaced a pixel；
Step 2, using four kinds of field informations obtained in bucket detector successively recording step 1；
Step 3, the field information according to obtained in step 2 calculates the partial derivative on the four direction of image；
Step 4, after n times sample, double velocity correlation calculating is carried out to the partial derivative on four direction, extracts related information；
Step 5, the result according to obtained in step 4 calculates the gradient information of different directions；
Step 6, the gradient information in step 5 is superimposed, the sharp edge image for the fuzzy object as a result rebuild passes through two
Kind gradient is added, and achievees the purpose that the inclined edge of opposite any direction extracts；
In step 1, the displacement on four direction is carried out to random speckle, it may be assumed that
In above-mentioned formula, x and y are pixel coordinate, SkRepresent the random speckle that kth time sampling generates, S1 kRepresent SkIt moves in the x-direction
It is dynamic, S2 kRepresent SkIt moves in the y-direction, S3 kRepresent SkIt tilts 45 ° of directions along x to move, S4 kRepresent Sk45 ° of directions are tilted along y to move.
2. the method according to claim 1 for quickly carrying out edge extracting to fuzzy object, it is characterised in that in step 3
In, the partial derivative on four direction is calculated:
Di edge(k)For the partial derivative of all directions kth time sampling, i represents i-th kind of random speckle displacement, Si kIt is SkFour kinds displacement after
Speckle, i=1,2,3,4,Matrix dot product is represented, O (x, y) represents target information matrix.
3. the method according to claim 2 for quickly carrying out edge extracting to fuzzy object, it is characterised in that in step 4
In, repetition steps 1 and 2,3 pairs of targets sample, and after n times sample, carry out double velocity correlation to the partial derivative on four direction
It calculates, i.e.,
Wherein SiTo project speckle on the image, i is direction number, EiBelieve for the strong association of calculated all directions partial derivative
Breath.
4. the method according to claim 3 for quickly carrying out edge extracting to fuzzy object, it is characterised in that in step 5
In, two gradient informations are calculated according to the two secondary association calculated results that step 4 obtains；
Edge1And Edge2For the gradient value for calculating image different directions.
5. the method according to claim 4 for quickly carrying out edge extracting to fuzzy object, it is characterised in that in step 6
In, two gradient informations of step 5 are superimposed, obtain final object edge information:
GIedge=Edge1+Edge2 (10)
GIedgeThe clearly object edge image as reconstructed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610954078.6A CN106548183B (en) | 2016-10-27 | 2016-10-27 | A method of edge extracting quickly being carried out to fuzzy object |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610954078.6A CN106548183B (en) | 2016-10-27 | 2016-10-27 | A method of edge extracting quickly being carried out to fuzzy object |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106548183A CN106548183A (en) | 2017-03-29 |
| CN106548183B true CN106548183B (en) | 2019-04-12 |
Family
ID=58392860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610954078.6A Active CN106548183B (en) | 2016-10-27 | 2016-10-27 | A method of edge extracting quickly being carried out to fuzzy object |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106548183B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909591B (en) * | 2017-09-29 | 2021-06-01 | 华北水利水电大学 | Boundary extraction method and system based on structured light illumination and single-pixel imaging |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7471341B2 (en) * | 2004-02-18 | 2008-12-30 | Analog Devices, Inc. | Apparatus, methods, and articles for a blind ghost cancellation on an SD video |
| CN101599174A (en) * | 2009-08-13 | 2009-12-09 | 哈尔滨工业大学 | Method for outline extraction of level set medical ultrasonic image area based on edge and statistical nature |
| CN105469398A (en) * | 2015-11-23 | 2016-04-06 | 南京航空航天大学 | Deformation speckle generation method based on reverse mapping method |
-
2016
- 2016-10-27 CN CN201610954078.6A patent/CN106548183B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7471341B2 (en) * | 2004-02-18 | 2008-12-30 | Analog Devices, Inc. | Apparatus, methods, and articles for a blind ghost cancellation on an SD video |
| CN101599174A (en) * | 2009-08-13 | 2009-12-09 | 哈尔滨工业大学 | Method for outline extraction of level set medical ultrasonic image area based on edge and statistical nature |
| CN105469398A (en) * | 2015-11-23 | 2016-04-06 | 南京航空航天大学 | Deformation speckle generation method based on reverse mapping method |
Non-Patent Citations (1)
| Title |
|---|
| 基于局部Hadamard调制的迭代去噪鬼成像;张伟良 等;《光学学报》;20160430;第36卷(第4期);全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106548183A (en) | 2017-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109791696B (en) | Method, device and method for locating event cameras for 3D reconstruction of a scene | |
| Wang et al. | Fusing meter-resolution 4-D InSAR point clouds and optical images for semantic urban infrastructure monitoring | |
| CN103729854B (en) | A kind of method for detecting infrared puniness target based on tensor model | |
| CN102360498B (en) | Reconstruction method for image super-resolution | |
| CN109544677A (en) | Indoor scene main structure method for reconstructing and system based on depth image key frame | |
| CN105898278B (en) | A kind of three-dimensional video-frequency conspicuousness detection method based on binocular Multidimensional Awareness characteristic | |
| Chen et al. | Dynamic occlusion detection and inpainting of in situ captured terrestrial laser scanning point clouds sequence | |
| CN103871039B (en) | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection | |
| CN110555908A (en) | three-dimensional reconstruction method based on indoor moving target background restoration | |
| CN105869146A (en) | Saliency fusion-based SAR image change detection method | |
| CN104268880A (en) | Depth information obtaining method based on combination of features and region matching | |
| CN103826032A (en) | Depth map post-processing method | |
| CN105654511A (en) | Quick detecting and tracking method for weak moving object | |
| CN106097317A (en) | A kind of many spot detection based on discrete cosine phase information and localization method | |
| CN117274515A (en) | Visual SLAM method and system based on ORB and NeRF mapping | |
| CN105469428B (en) | A kind of detection method of small target based on morphologic filtering and SVD | |
| CN108961385A (en) | A kind of SLAM patterning process and device | |
| CN105678814B (en) | A kind of building facade image of combination phase correlation analysis repeats texture detection | |
| CN106548183B (en) | A method of edge extracting quickly being carried out to fuzzy object | |
| Chen et al. | Multidirectional edge detection based on gradient ghost imaging | |
| Rao et al. | Extreme feature regions detection and accurate quality assessment for point-cloud 3D reconstruction | |
| CN108876807A (en) | A kind of real-time piggyback satellite image motion object detection tracking | |
| Hu et al. | A new method and implementation of blind restoration algorithm for moving fuzzy license plate image based on frequency-domain characteristics | |
| Ouyang et al. | Research on DENOISINg of cryo-em images based on deep learning | |
| CN114972332A (en) | Bamboo laminated wood crack detection method based on image super-resolution reconstruction 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |