CN108961176A - Range gating three-dimensional imaging is adaptive bilateral with reference to restorative procedure - Google Patents
Range gating three-dimensional imaging is adaptive bilateral with reference to restorative procedure Download PDFInfo
- Publication number
- CN108961176A CN108961176A CN201810611102.5A CN201810611102A CN108961176A CN 108961176 A CN108961176 A CN 108961176A CN 201810611102 A CN201810611102 A CN 201810611102A CN 108961176 A CN108961176 A CN 108961176A
- Authority
- CN
- China
- Prior art keywords
- image
- depth
- value
- depth image
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000002146 bilateral effect Effects 0.000 title claims abstract description 47
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 40
- 238000003384 imaging method Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 230000000694 effects Effects 0.000 claims description 9
- 241001269238 Data Species 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 238000012217 deletion Methods 0.000 claims description 4
- 230000037430 deletion Effects 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 2
- 239000004576 sand Substances 0.000 claims description 2
- 210000001367 artery Anatomy 0.000 claims 1
- 238000004080 punching Methods 0.000 claims 1
- 210000003462 vein Anatomy 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 10
- 238000010586 diagram Methods 0.000 description 5
- 230000008439 repair process Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000003111 delayed effect Effects 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 244000141353 Prunus domestica Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
Abstract
A kind of range gating three-dimensional imaging is adaptive bilateral with reference to the method repaired, include the following steps: step 1: two-dimensional space sectioning image adjacent according to range gating first, it is indicated respectively with A, B frame, sets 3D threshold value, and obtain original depth image to be repaired using 3-d inversion algorithm；Step 2: original depth image and A frame image being secondly partially filled with missing data using the adaptive bilateral algorithm that refers to, the depth image being partially filled with；Step 3: the depth image being partially filled with and B frame image completely being filled into missing data using the adaptive bilateral algorithm that refers to again, obtain filling complete depth image；Step 4: will finally fill the image of complete depth image and (A+B)/2 using adaptive bilateral with reference to algorithm smothing filtering, the depth image after obtaining complete, smooth reparation.The present invention can achieve the cavity for the depth image that effective completion range gating 3 dimension imaging technology obtains and the purpose of removal noise.
Description
Technical field
The invention belongs to technical field of image processing more particularly to the reparation skills of range gating three-dimensional imaging depth image
Art.
Background technique
Range-gated Imager is a kind of Active Imaging technology, using pulse laser as lighting source, using gating at
For picture device as detector, passing through the delay between control laser pulse and gate pulse can be achieved space slice imaging, filtering back
Scape, while the back scattering noise such as inhibit atmosphere, realize target acquisition.
It in fact, every width gating image not only includes luminance information, while also including range information.Several identical gate-widths,
Exist between the image of different delayed time apart from upper information association.The three of target can be finally inversed by by algorithm using the information association
Tie up information.The algorithm of laser-gated three-dimensional imaging at present mainly have stepping delay three-dimensional imaging, gain modulation three-dimensional imaging and away from
Three kinds from energy correlation three-dimensional imaging (also known as super-resolution three-dimensional imaging).
The stepping three-dimensional imaging that is delayed causes greatly real-time very poor due to original data volume, it is difficult to for online real-time three-dimensional at
Picture.Gain modulation three-dimensional imaging leads to ambient noise and device since two width slice maps are obtained under different system gains
Part noise is difficult to effectively filter out.Apart from energy correlation three-dimensional imaging algorithm, it is only necessary to modify to delay not co-extensive to obtain
At present gating image, so that it may which, using the distance of the relationship Inversion Calculation target between distance and energy, complexity is low, data
Few (minimum two frame) is measured, and under identical parameters, there is smaller three-dimensional reconstruction compared to gain modulation three-dimensional imaging algorithm
Section and there is higher range accuracy, therefore become in recent years in practical applications, online real time laser gates three-dimensional imaging
Technical way.
Although range gating 3 dimension imaging technology has, spatial resolution is high, operating distance is remote and effectively inhibit atmosphere or
The characteristics of back scattering of water, can solve the problems, such as that the existing operating distance of conventional three-dimensional imaging is close, spatial resolution is low, and
The target three-dimensional information of the big depth of field, High Range Resolution can be obtained quickly, in real time, but this method acquisition is depth image
Or three-dimensional point cloud image still has many places to be repaired, is mainly reflected in objective optics reflection characteristic and illumination light
Being unevenly distributed etc. by force causes 3-D image there are hole region (missing depth data) and with certain noise.Depth image
Cavity and noise can make the distance resolution for obtaining target three-dimensional information reduce, while can also reduce range gating three-dimensional imaging
The operating distance of technology, so that three-dimensional information of the technology when obtaining fine structure target and distant object will appear mould
Paste, the serious data that will lead to acquisition are full of prunes.Therefore adjust the distance gate three-dimensional imaging acquisition depth image into
Row filling cavity region, removal noise are a urgent problems.
The algorithm for repairing depth image at present is primarily directed to binocular stereo vision, method of structured light, time flight calculation
The depth image of the acquisitions such as method, there are no the calculations of the depth image reparation obtained directed entirely to range gating 3 dimension imaging technology
Method.
Summary of the invention
Place in view of the deficiency of the prior art, it is a kind of adaptive bilateral it is a primary object of the present invention to propose
With reference to the method for reparation, made an uproar with reaching cavity and the removal of the depth image that effective completion range gating 3 dimension imaging technology obtains
The purpose of sound.
In order to achieve the above objectives, technical solution provided by the invention is as follows:
A kind of range gating three-dimensional imaging is adaptive bilateral with reference to the method repaired, and includes the following steps:
Step 1: two-dimensional space sectioning image adjacent according to range gating first is indicated with A, B frame respectively, sets 3D threshold
Value, and original depth image to be repaired is obtained using 3-d inversion algorithm；
Step 2: original depth image and A frame image being secondly partially filled with missing number using the adaptive bilateral algorithm that refers to
According to the depth image being partially filled with；
Step 3: again completely filling the depth image being partially filled with using the adaptive bilateral algorithm that refers to B frame image
Missing data obtains filling complete depth image；
Step 4: will finally fill the image of complete depth image and (A+B)/2 using adaptive bilateral flat with reference to algorithm
Sliding filtering, the depth image after obtaining complete, smooth reparation.
It can be seen from the above technical proposal that the invention has the following advantages:
Using the present invention, due to being by depth each in neighborhood of pixel points to be repaired in the adaptive bilateral reference algorithm of proposition
Value after degree data weighted average is assigned to pixel to be repaired, if pixel to be repaired has depth data, the value being assigned to
Play the role of smothing filtering；If pixel depth data to be repaired lacks, the value being assigned to plays completion depth data
Effect.So the adaptive bilateral algorithm that refers to proposed by the present invention can be used for filling cavity and remove noise.
Using the present invention, when due to calculating the weighted average of each depth data in neighborhood, weight coefficient is by depth image sky
Between domain weight and the gray scale domain weight of two-dimensional slice image composition, two-dimensional slice image is depth image and basis, and includes rich
Rich target texture information.So compared to only being operated to depth image, it is proposed by the present invention adaptive bilateral with reference to algorithm
Fill up or smooth data closer to target truthful data.
Using the present invention, due to the 3D inverting threshold value in completion depth data with reference to two-dimensional slice image, when needing to mend
Ability completion depth when the gray value of the corresponding two-dimensional slice image same position of depth data pixel is greater than 3D inverting threshold value entirely
Data.It is really needed so can accurately be filled up when adaptive bilateral reference algorithm filling cavity proposed by the present invention
The position of filling reduces excessively filling or generation the phenomenon that lack of fill to the greatest extent.
Using the present invention, due to only lacking in depth data and the position of completion needed to refer to algorithm using adaptively bilateral
Filling cavity is only used in the position for having depth data adaptive bilateral with reference to algorithm depth of smoothness data.So the present invention mentions
The adaptive bilateral pixel that operation can be reduced with reference to algorithm out has faster reparation speed.
Using the present invention, due to for each depth data to be repaired, neighborhood window value, airspace weight and value
Domain weight can be adaptive, so, adaptive bilateral reference algorithm proposed by the present invention is compared to fixed window value, fixed airspace
For the algorithm of weight and fixed codomain weight, filling or smooth depth data are closer to target true three-dimension information.
Detailed description of the invention
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference
Attached drawing, the present invention is described in more detail as after, in which:
Fig. 1 is the adaptive bilateral flow diagram with reference to the method repaired of range gating three-dimensional imaging of the present invention；
Fig. 2 is the flow diagram of step 2-3 in Fig. 1；
Fig. 3 is the flow diagram of step 4 in Fig. 1；
Fig. 4 is the two-dimensional slice image according to the embodiment of the present invention, in which: (a) two dimension slicing A frame image, (b) two dimension is cut
Piece B frame image, (c) two dimension slicing (A+B)/2 frame image；
Fig. 5 is the depth for utilizing triangle to generate apart from energy related algorithm according to two-dimensional slice image of the embodiment of the present invention
Image and its repair after as a result, (a) original depth image, (b) depth image being partially filled with according to A frame, (c) according to B frame
The depth image completely filled, (d) depth image that the reparation after smoothing denoising is completed.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference
Attached drawing, the present invention is described in more detail.It should be noted that in attached drawing or specification description, similar or identical portion
Divide and all uses identical figure number.The implementation for not being painted or describing in attached drawing is those of ordinary skill in technical field
Known form.In addition, though can provide the demonstration of the parameter comprising particular value herein, it is to be understood that parameter is without definite etc.
In corresponding value, but can be similar to be worth accordingly in acceptable error margin or design constraint.
The invention discloses a kind of adaptive bilateral methods with reference to reparation depth image, to reach effective completion distance choosing
The cavity for the depth image that logical 3 dimension imaging technology obtains and the purpose of removal noise.
The adaptive bilateral process signal with reference to the method repaired of Fig. 1 range gating three-dimensional imaging of the present invention in attached drawing
Figure, the present invention can be divided mainly into 4 parts as we know from the figure: 3-d inversion generates original depth image to be repaired, original depth
Depth image that image reference A frame is partially filled with, turns finally to the depth image completely filled referring again to B frame
(A+B)/2 the depth image of complete correction is obtained after frame smoothing denoising.Since the 3-d inversion of first part is intended merely to obtain
The content that three dimensional depth image, the not instead of present invention are mainly explained, the object of main function of the present invention, therefore depth is not done to it
The explanation entered.Subsequent three parts are main implementation steps of the invention, in order to illustrate side proposed by the present invention in more detail
It is as shown in Figure 2 and Figure 3 to be drawn specific detailed flow diagram by method again for three parts behind in Fig. 1.In the following, we are referring to figure
1, Fig. 2, flow diagram shown in Fig. 3, in detail explanation the present embodiment.
Specific step is as follows for the present embodiment:
Step 1: two-dimensional space sectioning image adjacent according to range gating first is indicated with A, B frame respectively, sets 3D threshold
Value, and original depth image to be repaired is obtained using 3-d inversion algorithm.Choose author laboratory independent research laser
The night plant two-dimensional slice image data that range gating 3-D imaging system obtains, adjacent two-dimensional space slice map
Shown in the enhancing image of picture such as attached drawing Fig. 4 (a), Fig. 4 (b).A frame indicates the sectioning image at range data acquisition system 201m, B
Frame indicates the sectioning image after opposite A frame delay 50ns.According to the gray value of background in A, B frame, 3D threshold value is set as 5 (i.e. A frames
Or pixel of the gray value less than 5 is not involved in 3-d inversion in B frame), and utilize range gating super-resolution 3-d inversion algorithm
(actually triangle is apart from energy correlation 3-d inversion algorithm), obtains original depth-map to be repaired by adjacent A, B frame
Picture.Shown in original depth image such as attached drawing Fig. 5 (a) to be repaired, it can be seen from the figure that there is the sky of many depth data missings
Hole, value NaN are expressed as white cavity in image kind；In addition, there are also many spikes in image, when depth image noise
It is caused.Therefore illustrate that original depth picture quality to be repaired is very poor, differed farther out with true target three-dimensional information.Below
How specific explanation repairs the original depth image step by step.
It is that triangle is calculated apart from energy correlation that 3-d inversion algorithm used in original depth image is obtained in the step 1
Method, in the algorithm, laser pulse and gate pulse are also rectangular pulse, gate gate-width and laser pulse width is equal in magnitude, because
This, under the effect of echo broadening effect, target range is triangle to energy envelope, passes through the energy ash established between contiguous slices
Degree can get the range information of target than relationship, so as to obtain depth image according to adjacent sectioning image.
Step 2: original depth image and A frame image being secondly partially filled with missing number using the adaptive bilateral algorithm that refers to
According to the depth image being partially filled with；
Step 3: again completely filling the depth image being partially filled with using the adaptive bilateral algorithm that refers to B frame image
Missing data obtains filling complete depth image；
Wherein in step 2- step 3:
It is adaptively bilateral that with reference to completion missing data, specific step is as follows:
Step 1a: the position that depth data is lacked in original depth image is found.The upper left corner is opened from original depth image
Begin, find the cavity position for lacking depth data in original depth image, white is expressed as in image, the pixel that value is NaN
The as position of depth data missing；
Step 2a: for the position of each depth data missing, according to the neighbour centered on depth data deletion sites
The size of the variance adjustment neighborhood window of all depth datas, makes window value constantly in permitted minimum value in domain
It is changed between Window_min and maximum value Window_max.It is first minimum value by the initial value design of neighborhood window value
Window_min, the variance initial value design minimum value Var_min in neighborhood.Calculate the neighbour centered on first missing data position
Variance in domain, and compared with the initial value of setting, if variance yields is greater than initial value, the neighborhood window of second missing data position
Value reduces 1, if being less than initial value, increases by 1, constant if being equal to initial value.And so on, by the variance in the neighborhood of current location
The size of next neighborhood window value is determined compared with the variance in prior location neighborhood.In the present embodiment, Window_min
=3, Window_max=15, Var_min=0.
Step 3a: depth data is assigned to the value after depth data each in neighborhood at depth data deletion sites weighted average
The pixel of missing, to the depth data of completion missing, the point of completion is the centre of neighbourhood；The value being assigned to is with following expression
Formula (1) indicates:
I ' (x, y) is the value for being assigned to centre of neighbourhood missing data pixel, and Ω is the territory of pixel (x, y), one
As in the case of be rectangular area centered on (x, y), w (i, j) is the weight coefficient at pixel (i, j), wpIt is normalization
Parameter, I (i, j) are each depth data values in contiguous range.Wherein:
Bilateral for this method proposition refers to algorithm, and weight coefficient w (i, j) is by depth image spatial domain weight wsWith two
Tie up the gray scale domain weight w of sectioning imagerComposition, it may be assumed that
W (i, j)=ws(i, j) wr(i, j) (8)
Wherein wsAnd wrIt is expressed as follows with Gaussian function:
σs、σrFor the standard deviation based on Gaussian function, G (i, j) is gray scale of the two-dimensional slice image at pixel (i, j)
Value.
σs、σrDetermine the bilateral performance with reference to algorithm.In σsIn the case where fixation, σrExcessive, different gray scale differences are corresponding
Weight w is larger, then loses the effect for retaining marginal information using grey scale change, bilateral degenerate with reference to algorithm is gaussian filtering
Device；σrToo small, weight w is too sensitive to different gray scale differences, then loses the effect of filtering.It is bilateral to consider pixel with reference to algorithm
Between spatial coherence while, it is also considered that the similarity degree of pixel value, thus denoising when be also able to maintain image details letter
Breath is not lost.
Step 4a: the standard deviation sigma of depth image airspace weight Gaussian functionsCorresponding adjustment is done according to neighborhood window value；
The standard deviation of all pixels point position in rectangular neighborhood window is regarded to the standard deviation of the neighborhood window airspace weight Gaussian function
σs, since window value adaptively constantly adjusts, the also adaptive constantly adjustment of airspace weight；Two-dimensional slice image codomain weight
The standard deviation sigma of Gaussian functionrCorresponding adjustment is done according to gray value all in two-dimensional slice image neighborhood window；By rectangle neighbour
The standard deviation of all gray values regards the standard deviation sigma of the neighborhood window codomain weight Gaussian function in the window of domainr, due in window
Gray value constantly changes, therefore the also adaptive constantly adjustment of codomain weight.
In the step 4a, why σs、σrIt determines the bilateral performance with reference to algorithm, therefore it cannot be arbitrarily set as
Fixed value.By the standard deviation sigma of depth image airspace weight Gaussian functionsCorresponding adjustment is done according to neighborhood window value, rectangle is adjacent
The standard deviation of all pixels point position regards the standard deviation sigma of the neighborhood window airspace weight Gaussian function in the window of domains, due to window
The adaptive constantly adjustment of mouth value, therefore the also adaptively constantly adjustment of airspace weight.By two-dimensional slice image codomain weight Gaussian function
Several standard deviation sigmasrCorresponding adjustment is done according to gray value all in two-dimensional slice image neighborhood window, in rectangular neighborhood window
The standard deviation of all gray values regards the standard deviation sigma of the neighborhood window codomain weight Gaussian functionr, since gray value is not in window
Disconnected variation, therefore the also adaptive constantly adjustment of codomain weight.
According to step 3a- step 4a it is found that Gauss in the neighborhood window value and field of each depth data to be repaired
The standard deviation of function is constantly changed according to the depth data of depth data and current field window in previous field window
, therefore the value for being assigned to field center more can really reflect the actual conditions in region instantly, so adaptively making
Filling or smooth depth data are more acurrate.
Step 5a: in completion depth data with reference to the 3D inverting threshold value being arranged in step 1；When needing completion depth data
Ability completion depth data when the gray value of the corresponding two-dimensional slice image same position of pixel is greater than 3D inverting threshold value, otherwise
Skip the pixel；
Step 6a: traversing entire original depth image, repeats step 1a- to step 5a, until completion meets the institute of condition
Until the depth data for having position.
Wherein in step 5a- step 6a:
When deciding whether to fill up the depth data of missing, with reference to the 3D inverting threshold value 5 being arranged in step 1, work as needs
Ability completion depth data when the gray value of the corresponding A frame same position of completion depth data pixel is greater than 5, otherwise just skips this
Pixel.It can be avoided in this way and the position of all missing depth datas all filled out, cause excessively to fill.The above are for a certain
The operation that the position of depth data missing is carried out, will traverse entire original depth image, repeat step 1a- to step 5a, directly
Until completion meets the depth data of all positions of condition (corresponding A frame gray value is greater than 5), can just it obtain with reference to A frame
The depth image being partially filled with.Shown in the depth image such as Fig. 5 (b) being partially filled with according to A frame, it is recognised that comparing from figure
Original depth image in Fig. 5 (a), has filled up partial depth data, and the branch such as image upper left is gradually complete.With
Upper filling is filled and imperfect just for A frame, therefore also does the depth image after filling above with B frame again
Similar operation is with complete depth of cracking closure image.So repeating step 1a- to step 5a again, changed when by codomain weight computing
At the gray value of B frame, 5 pixel is greater than with reference to gray value in B frame, the depth image being partially filled with and B frame image are used
Until the depth data of the adaptive bilateral all positions for meeting condition (corresponding B frame gray value is greater than 5) with reference to algorithm completion, obtain
To filling, depth image such as Fig. 5 (c) that complete depth image is completely filled according to B frame is shown, it is recognised that phase from figure
Than the original depth image in Fig. 5 (a) and the depth image being partially filled in Fig. 5 (b), most depth datas of missing
By completion, the branch in image is almost complete.Illustrate that adaptive bilateral reference algorithm proposed by the present invention can be effective
The filling distance gating 3 dimension imaging technology obtain depth image.
Step 4: will finally fill the image of complete depth image and (A+B)/2 using adaptive bilateral flat with reference to algorithm
Sliding filtering, the depth image after obtaining complete, smooth reparation.
Wherein in step 4:
It is adaptively bilateral that with reference to algorithm smoothing denoising, specific step is as follows:
Step 1b: the position filled and have depth data in complete depth image is found；
Step 2b: the method for depth of smoothness data is consistent with step 2a- step 4a, it is only necessary to by two-dimensional slice image A frame
Or B frame replaces with (A+B)/2 frame, and the position for lacking depth data is replaced with to the position of depth data；
Step 3b: traversal entirely fills complete depth image, repeats step 1b- step 2b, until smooth all positions
Depth data until.
After filling complete depth image, due to there is the pixel of depth data not pass through any operation, depth number also originally
According to there is many spikes, the depth data newly filled may also introduce spike, therefore also carry out to all depth datas smooth
Filtering, obtains smooth depth image.It is calculated to reduce, carries out smothing filtering only in the position for having depth data, therefore
The position filled and have depth data in complete depth image is found first.For each depth data, need its neighborhood
Value after interior all depth datas weighted average replaces the value of the centre of neighbourhood to achieve the effect that depth of smoothness data, side
Method is consistent with mono- step 4a of step 2a, it is only necessary to which two-dimensional slice image A frame or B frame are replaced with (A+B)/2 frame (such as attached drawing Fig. 4
(c) shown in), the position for lacking depth data is replaced with to the position of depth data.Not with completion missing depth data
With depth of smoothness data are without necessarily referring to 3D threshold value.The above depth of smoothness data, will also be all over for a certain pixel
It goes through and entirely fills complete depth image, repeat step 1b- step 2b, until the depth data of smooth all positions,
It can obtain repairing the depth image completed.Repair shown in the depth image such as Fig. 5 (d) completed, it can be seen that compared to it is smooth it
Before, the spike of image is much less, and image is also smoothened very much.In order to measure smooth effect, with signal-to-noise ratio come objective
Quantization.The signal-to-noise ratio of original depth image is 25.7889dB, and the signal-to-noise ratio for repairing the depth image of completion is 31.9161dB,
Probably improve 6dB.Illustrate it is proposed by the present invention it is adaptive it is bilateral with reference to algorithm can effectively smooth range gating it is three-dimensional
The depth image that imaging technique obtains.
So far, attached drawing is had been combined the present embodiment is described in detail.According to above description, those skilled in the art
Adaptively bilateral to range gating three-dimensional imaging depth map of the invention there should be clear understanding with reference to restorative procedure.
In addition, the definition of the above method is not limited in various specific structures, shape or the mode mentioned in embodiment, this
Field those of ordinary skill simply can be changed or be replaced to it.
In conclusion it is proposed by the present invention adaptive bilateral with reference to the method repaired, reach effective completion range gating
The cavity for the depth image that 3 dimension imaging technology obtains and the purpose of removal noise, and authentic and valid with completion depth data,
Depth of smoothness effect data is obvious and is able to maintain image detail information, while having both the advantages that repair speed fast, adaptable.
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
It describes in detail bright, it should be understood that the above is only a specific embodiment of the present invention, is not intended to restrict the invention, it is all
Within the spirit and principles in the present invention, any modification, equivalent substitution, improvement and etc. done should be included in guarantor of the invention
Within the scope of shield.
Claims (4)
1. a kind of range gating three-dimensional imaging is adaptive bilateral with reference to the method repaired, include the following steps:
Step 1: two-dimensional space sectioning image adjacent according to range gating first is indicated with A, B frame respectively, sets 3D threshold value,
And original depth image to be repaired is obtained using 3-d inversion algorithm；
Step 2: original depth image and A frame image are secondly partially filled with missing data using the adaptive bilateral algorithm that refers to,
The depth image being partially filled with；
Step 3: the depth image being partially filled with and B frame image completely being filled into missing using the adaptive bilateral algorithm that refers to again
Data obtain filling complete depth image；
Step 4: finally smoothly filtering the image for filling complete depth image and (A+B)/2 using the adaptive bilateral algorithm that refers to
Wave, the depth image after obtaining complete, smooth reparation.
2. range gating three-dimensional imaging according to claim 1 is adaptive bilateral with reference to the method repaired, wherein in step 1
Obtaining 3-d inversion algorithm used in original depth image is triangle apart from energy related algorithm；In the algorithm, laser arteries and veins
Punching and gate pulse are also rectangular pulse, gate gate-width and laser pulse width is equal in magnitude, therefore, are acted in echo broadening effect
Under, target range is triangle to energy envelope, can get target than relationship by the energy gray scale established between contiguous slices
Range information, so as to obtain depth image according to adjacent sectioning image.
3. range gating three-dimensional imaging according to claim 1 is adaptively bilateral to refer to the method repaired, wherein step 2-
In step 3:
It is adaptively bilateral that with reference to completion missing data, specific step is as follows:
Step 1a: finding the position that depth data is lacked in original depth image, white is expressed as in image, the picture that value is NaN
Vegetarian refreshments is the position of depth data missing；
Step 2a: neighborhood window is adjusted according to the variance of all depth datas in the neighborhood centered on depth data deletion sites
Size, change window value constantly between permitted minimum value Window_min and maximum value Window_max；It is first
It is first minimum value by the initial value design of neighborhood window value, the variance initial value design minimum value in neighborhood；Calculate first missing number
According to the variance in the neighborhood centered on position, and compared with the initial value of setting, if variance yields is greater than initial value, second missing number
Reduce 1 according to the neighborhood window value of position, if being less than initial value, increases by 1, it is constant if being equal to initial value；And so on, it will be current
Variance in the neighborhood of position determines the size of next neighborhood window value compared with the variance in prior location neighborhood；
Step 3a: depth data missing is assigned to the value after depth data each in neighborhood at depth data deletion sites weighted average
Pixel, to the depth data of completion missing, the point of completion is the centre of neighbourhood；The value being assigned to is with following expression formula
(1) it indicates:
I ' (x, y) is the value for being assigned to centre of neighbourhood missing data pixel, and Ω is the territory of pixel (x, y), general feelings
It is the rectangular area centered on (x, y) under condition, w (i, j) is the weight coefficient at pixel (i, j), wpIt is normalized parameter,
I (i, j) is each depth data value in contiguous range；Wherein:
Bilateral for this method proposition refers to algorithm, and weight coefficient w (i, j) is by depth image spatial domain weight wsIt is cut with two dimension
The gray scale domain weight w of picturerComposition, it may be assumed that
W (i, j)=ws(i, j) wr(i, j)
(3)
Wherein wsAnd wrIt is expressed as follows with Gaussian function:
Wherein σs、σrFor the standard deviation based on Gaussian function, G (i, j) is gray scale of the two-dimensional slice image at pixel (i, j)
Value；
Step 4a: the standard deviation sigma of depth image airspace weight Gaussian functionsCorresponding adjustment is done according to neighborhood window value；By rectangle
The standard deviation of all pixels point position regards the standard deviation sigma of the neighborhood window airspace weight Gaussian function in neighborhood windows, due to
Window value adaptively constantly adjusts, therefore the also adaptive constantly adjustment of airspace weight；Two-dimensional slice image codomain weight Gaussian function
Several standard deviation sigmasrCorresponding adjustment is done according to gray value all in two-dimensional slice image neighborhood window；By rectangular neighborhood window
The standard deviation of interior all gray values regards the standard deviation sigma of the neighborhood window codomain weight Gaussian functionr, due to gray value in window
Constantly variation, therefore the also adaptive constantly adjustment of codomain weight；
Step 5a: in completion depth data with reference to the 3D inverting threshold value being arranged in 1 step 1 of claims；When needing completion
Ability completion depth number when the gray value of the corresponding two-dimensional slice image same position of depth data pixel is greater than 3D inverting threshold value
According to otherwise just skipping the pixel；
Step 6a: traversing entire original depth image, repeats step 1a- to step 5a, until completion meets all positions of condition
Until the depth data set.
4. range gating three-dimensional imaging according to claim 1 is adaptively bilateral to refer to the method repaired, wherein step 4
In:
It is adaptively bilateral that with reference to algorithm smoothing denoising, specific step is as follows:
Step 1b: the position filled and have depth data in complete depth image is found；
Step 2b: the method for depth of smoothness data is consistent with step 2a- step 4a, it is only necessary to by two-dimensional slice image A frame or B frame
(A+B)/2 frame is replaced with, the position for lacking depth data is replaced with to the position of depth data；
Step 3b: traversal entirely fills complete depth image, repeats step 1b- step 2b, until the depth of smooth all positions
Until degree evidence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810611102.5A CN108961176B (en) | 2018-06-14 | 2018-06-14 | Self-adaptive bilateral reference restoration method for range-gated three-dimensional imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810611102.5A CN108961176B (en) | 2018-06-14 | 2018-06-14 | Self-adaptive bilateral reference restoration method for range-gated three-dimensional imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108961176A true CN108961176A (en) | 2018-12-07 |
| CN108961176B CN108961176B (en) | 2021-08-03 |
Family
ID=64488685
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810611102.5A Active CN108961176B (en) | 2018-06-14 | 2018-06-14 | Self-adaptive bilateral reference restoration method for range-gated three-dimensional imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108961176B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110189265A (en) * | 2019-05-06 | 2019-08-30 | 中国科学院半导体研究所 | Apart from energy correlation three-dimensional imaging super pixel restorative procedure |
| CN116228601A (en) * | 2023-05-08 | 2023-06-06 | 山东中都机器有限公司 | Visual monitoring method for coal leveling effect of bidirectional coal leveling of train |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101860667A (en) * | 2010-05-06 | 2010-10-13 | 中国科学院西安光学精密机械研究所 | Method of quickly eliminating composite noise in images |
| CN103438863A (en) * | 2013-06-13 | 2013-12-11 | 中国科学院半导体研究所 | Range gating code super-resolution three-dimensional imaging method |
| CN105701778A (en) * | 2016-01-11 | 2016-06-22 | 赛诺威盛科技（北京）有限公司 | Method of removing metal artifact from CT image |
| CN106910242A (en) * | 2017-01-23 | 2017-06-30 | 中国科学院自动化研究所 | The method and system of indoor full scene three-dimensional reconstruction are carried out based on depth camera |
-
2018
- 2018-06-14 CN CN201810611102.5A patent/CN108961176B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101860667A (en) * | 2010-05-06 | 2010-10-13 | 中国科学院西安光学精密机械研究所 | Method of quickly eliminating composite noise in images |
| CN103438863A (en) * | 2013-06-13 | 2013-12-11 | 中国科学院半导体研究所 | Range gating code super-resolution three-dimensional imaging method |
| CN105701778A (en) * | 2016-01-11 | 2016-06-22 | 赛诺威盛科技（北京）有限公司 | Method of removing metal artifact from CT image |
| CN106910242A (en) * | 2017-01-23 | 2017-06-30 | 中国科学院自动化研究所 | The method and system of indoor full scene three-dimensional reconstruction are carried out based on depth camera |
Non-Patent Citations (1)
| Title |
|---|
| SUN LIANG 等: "Lower-upper-threshold correlation for underwater range-gated imaging self-adaptive enhancement", 《APPLIED OPTICS》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110189265A (en) * | 2019-05-06 | 2019-08-30 | 中国科学院半导体研究所 | Apart from energy correlation three-dimensional imaging super pixel restorative procedure |
| CN116228601A (en) * | 2023-05-08 | 2023-06-06 | 山东中都机器有限公司 | Visual monitoring method for coal leveling effect of bidirectional coal leveling of train |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108961176B (en) | 2021-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107862293B (en) | Radar color semantic image generation system and method based on countermeasure generation network | |
| CN106157267B (en) | Image defogging transmissivity optimization method based on dark channel prior | |
| CN102789635B (en) | Image enhancement method and image enhancement device | |
| CN104240194A (en) | Low-light-level image enhancement algorithm based on parabolic function | |
| DE102019118838A1 (en) | VIRTUAL PHOTOGRAMMETRY | |
| CN110189265B (en) | Distance energy correlation three-dimensional imaging super pixel repairing method | |
| Zheng et al. | Underwater image enhancement algorithm based on CLAHE and USM | |
| TWI394096B (en) | Method for tracking and processing image | |
| CN109903379A (en) | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling | |
| EP1465116A1 (en) | Computer graphics | |
| CN110288539A (en) | A kind of mobile clear method of underwater picture with dark channel prior in color combining space | |
| CN108961176A (en) | Range gating three-dimensional imaging is adaptive bilateral with reference to restorative procedure | |
| CN110866882A (en) | Layered joint bilateral filtering depth map restoration algorithm based on depth confidence | |
| CN102096913B (en) | Multi-strategy image fusion method under compressed sensing framework | |
| Shivakumar et al. | Real time dense depth estimation by fusing stereo with sparse depth measurements | |
| CN109087370A (en) | A kind of spongy defect image generation method of casting | |
| CN103136753B (en) | A kind of depth image segmentation method based on mean shift algorithm and mathematical morphology | |
| Fu et al. | Scene-awareness based single image dehazing technique via automatic estimation of sky area | |
| CN110223240A (en) | Image defogging method, system and storage medium based on color decaying priori | |
| Wei et al. | A new fast single-image defog algorithm | |
| CN113077504B (en) | Large scene depth map generation method based on multi-granularity feature matching | |
| CN110349113B (en) | Adaptive image defogging method based on dark primary color priori improvement | |
| CN103400381B (en) | A kind of Method for Underwater Target Tracking based on optical imagery | |
| CN104835121A (en) | Infinite norm constraint and maximum entropy principle-based hue mapping method | |
| WO2022214821A3 (en) | Monocular depth estimation |
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 |