CN108648221A - A kind of depth map cavity restorative procedure based on mixed filtering - Google Patents
A kind of depth map cavity restorative procedure based on mixed filtering Download PDFInfo
- Publication number
- CN108648221A CN108648221A CN201810443580.XA CN201810443580A CN108648221A CN 108648221 A CN108648221 A CN 108648221A CN 201810443580 A CN201810443580 A CN 201810443580A CN 108648221 A CN108648221 A CN 108648221A
- Authority
- CN
- China
- Prior art keywords
- pixel
- priority
- edge
- information
- neighborhood
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000001914 filtration Methods 0.000 title claims abstract description 42
- 230000008439 repair process Effects 0.000 claims abstract description 18
- 230000002146 bilateral effect Effects 0.000 claims abstract description 16
- 230000004927 fusion Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 4
- 238000000205 computational method Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 18
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 7
- 238000012797 qualification Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000004438 eyesight Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000010429 evolutionary process Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of depth map cavity restorative procedure based on mixed filtering, steps are as follows：Obtain depth image to be repaired；Identify its hole region；Calculate the priority of pixel in cavity；Priority is put into higher than the pixel of threshold value in priority query and is filled up；To the end of hole region is repaired and then repair non-hole region；For hole region and non-hole region, non-edge is first repaired respectively and repairs fringe region again.The present invention proposes is arranged priority for all pixels in cavity, and priority depends on support and confidence level of the neighborhood territory pixel to central cavity pixel.As unknown pixel is constantly padded, empty edge constantly reduces, while updating the priority of the pixel at empty edge, and the update threshold value of adaptivity, ensures that the pixel of highest priority is preferentially filled up.Texture and structural information is added in joint bilateral filtering principle in the present invention, has preferable treatment effect for texture and complicated image.
Description
Technical field
The invention belongs to depth map recovery technique fields in three-dimensional reconstruction, and in particular to a kind of depth based on mixed filtering
The method that figure cavity is repaired.
Background technology
Daily living scene is 3 D stereo scene, i.e., is all to have numerous three-dimensional informations to be interwoven, and the mankind
It is also on the basis of human visual perception and processing to receive these information.Three-dimensional reconstruction mainly passes through stereopsis at present
Feel technology converts two-dimensional signal to three-dimensional information, and the three-dimensional reconstruction that data realize two-dimensional bodies is extracted from image.Currently, having
Two kinds of mainstream technologys can rebuild three-dimensional scenic：One is the method using multiple view is rebuild, pass through video camera
Either camera is inferred to three-dimensional information and the display of the object in scene or scene using binocular or multi-view stereo vision；
Another is to rebuild threedimensional model on the basis of cromogram and depth image by way of " depth+texture ".Texture is color
Color image is the texture information for describing object, different from texture coloured image, and the gray value of depth image indicates objects in images
It is remoter to be worth bigger expression distance for the distance between video camera.
In depth image, there are many acquisition modes in acquisition process, the limitation due to equipment itself and outside environmental elements
Interference, such as：The calibrated error of sensor hardware and the error of offset, the measurement accuracy of offset can illumination condition by
It influences, the influence etc. of body surface texture material so that collected depth image can have the region that depth value is 0, i.e., empty
Hole, this can lead to the poor effect of three-dimensional reconstruction.
Mainly pass through filtering algorithm, Tomas i in 1998 and Maduchi first for image repair in three-dimensional reconstruction at present
The secondary theory for proposing bilateral filtering algorithm, passes through improvement over the years later, and Anh Vu Le in 2014 et al. propose to be based on direction
Joint bilateral filtering and part the joint bilateral filtering algorithm based on direction, in its formula in the calculating of space proximity
Further that edge direction factor is added, i.e. kernel is the gaussian filtering based on direction, and in depth image cavity is repaired, figure
As in pixel whether in cavity, whether be divided into four kinds in object boundary and carry out type, and a kind of algorithm is found by experiment
Combination, make to use different algorithms for different in the case of, significantly more efficient and accurate progresss cavity reparation.But
It is, based on the algorithm and its innovatory algorithm of joint bilateral filtering, to focus on more and improved in the calculating of central cavity pixel depth value
With it is perfect, strategy is filled up without excessive perfect for whole empty pixel.It may result in some unnecessary errors, example in this way
It is known like this to close on although surrounding adjacent pixels and the similarity of this center pixel are higher such as certain unknown center pixel
Number of pixels is less, inevitably can cause to miss because of the loss of learning of surrounding neighbors pixel if first carrying out the calculating of the pixel
Difference.In addition, the main structure information and texture information of image are not added in repair process for current algorithm, this can influence side
The accuracy that edge pixel calculates and the repair ability to image entirety.
Invention content
The present invention is directed at least solve the technical problems existing in the prior art, especially innovatively propose a kind of based on mixed
Close the depth map cavity restorative procedure of filtering.
In order to realize the above-mentioned purpose of the present invention, the present invention provides a kind of, and the depth map cavity based on mixed filtering is repaired
Method comprising following steps：
S1 obtains depth image to be repaired；
S2 identifies hole region, and the pixel that depth value is 0 is empty pixel；
S3 is arranged priority initial threshold, calculates the priority of all hole-filling area pixels, priority passes through cavity
The support r1 and confidence level r2 of pixel are determined；
S4, the pixel that priority is more than threshold value are put into priority query, and by the pixel in queue according to from big to small
Sequence carry out depth value calculating；
S5 extracts its main structure information and texture information by Total Variation from coloured image；
S6 first determines whether empty pixel is in fringe region for hole region, if being in non-edge,
Using joint bilateral filtering algorithm of the part based on direction of fusion structure and texture information, main calculation methods are depth value neighbours
What the depth value weighted average of domain pixel obtained, the weights of each neighborhood territory pixel by the neighborhood territory pixel that is got from image with
Spatial domain proximity, grey value similarity and structural similarity, the texture similarity of its center pixel obtain, if empty non-edge
Still there is the pixel that do not fill up in region, then returns to step S4；If it is not, starting to fill up fringe region, tied using fusion
The joint bilateral filtering algorithm based on direction of structure and texture information, the reparation neighborhood window that non-edge repairs algorithm are certainly
Adaptability；
S7 to the end of all hole region repairings and then repairs all non-hole regions, the non-non- side of hole region
The reparation of edge region is the three side filtering algorithm of joint by fusion structure and texture information, the side of filling up of non-cavity non-edge
Method and the non-edge of hole region to fill up mode identical.
The present invention solves the problems, such as that depth image occurs empty in gatherer process, is proposed first against whole filling-up hole strategy
Priority is set for all pixels in cavity, priority depends on neighborhood territory pixel to the support of central cavity pixel and can
Reliability.During hole-filling, the priority of empty edge pixel and the update threshold value of adaptivity can be updated, is ensured excellent
The first highest pixel of grade can be filled up preferentially.The hole-filling priority policy of the present invention considers not only pixel in field
Location information, gray value, depth value information, and consider neighborhood territory pixel number, priority value cycle calculations avoid because neglecting
Slightly empty pixel calculates error caused by filling up priority orders.
The present invention adds the main structure information and texture information of image in current existing filtering algorithm, according to image
In pixel whether in boundary, whether in cavity be divided into four classes, be non-empty non-edge pixels, the non-cavity picture in edge respectively
Element, empty non-edge pixels and empty edge pixel carry out empty reparation respectively, obtain the reparation result figure of better quality.
Two kinds of letters of the entirely depth map cavity restorative procedure combination texture based on mixed filtering proposed by the present invention and structure
Breath also has preferable handling result, this method not only to have validity, also general applicability the image of texture complexity.
Since objects in images marginal information amount is big, and for the accuracy requirement of pixel value height, therefore the present invention is first right
The pixel of non-edge is filled up, and when completion, the neighborhood territory pixel of fringe region is substantially repaired.Fringe region is repaiied again
Multiple, at this moment, acquisition is put forward more information by fringe region reparation, and the effect of reparation is more preferable.
The present invention a kind of preferred embodiment in, calculate pixel to be filled up neighborhood territory pixel point support r1 and can
Reliability r2, neighborhood territory pixel point support r1 be by pixel to be filled around known neighborhood territory pixel number divided by filter window in neighbour
Domain pixel number, the filter window are dimensioned to k*k, and the k is the positive integer more than 1；Confidence level r2 be neighborhood territory pixel with
Similarity between pixel to be filled up, including space proximity and grey value similarity, formula are as follows：
Wherein,For spatial neighbor degree,For grey value similarity,
Wherein, subscript d indicates information in depth map, and subscript c indicates that information comes from cromogram, subscript s representation spaces
Domain, subscript r indicate pixel coverage domain,Indicate the spatial neighbor degree between pixel, qxFor the abscissa of pixel q, qyFor picture
The ordinate of vegetarian refreshments q, the neighborhood territory pixel point of pixel, p centered on qxFor the abscissa of pixel p, pixel centered on p, pyFor picture
The ordinate of vegetarian refreshments p,Indicate the grey value similarity between pixel q points and center pixel p points, I in neighborhoodpFor p
The gray value of point, IqFor the gray value of q points, σsFor the spatial domain parameter in Euclidean distance formula, σrFor the standard deviation of Gauss formula.
By above method, the neighborhood territory pixel point support r1 and confidence level of pixel to be filled up accurately quickly is calculated
R2 ensures the smooth calculating of priority.
In the another kind preferably embodiment of the present invention, in the step S3, the computational methods of priority m are：M=r1
+ λ × r2 enters ＞ 0,
Wherein, enter the parameter of weight shared by support r1 and confidence level r2 when calculating priority to determine.
The present invention is provided with preferential before the depth for calculating each empty pixel to be measured for all unknown empty pixels
Grade, that is, fill up the degree of priority of sequence.All empty pixels can be ranked up according to its priority in hole region, fill up empty
When hole, it can improve according to the descending progress depth value calculating of priority orders of each pixel to be measured and fill up efficiency, avoid
Because ignoring caused by empty pixel fills up priority orders and calculating error.
In the another kind preferably embodiment of the present invention, the method for extracting main structure information and texture information is：Image I
The two-part linear combination of main structure information S, texture information T is resolved into, is indicated as follows：
I=S+T,
Define Total Variation：Wherein, Ω is image-region,
Establish energy theorem：
It seeksAllow main structure information S to be output close to artwork I, 1 × TV of λ (S) are TV
Regularization term, λ 1 are regularization parameters, the weight of first item and Section 2 in balanced type,
Texture information is just acquired by formula T=I-S in image,
The pixel is obtained for each pixel of hole region based on main structure information S, the texture information T sought
The structural similarity and texture similarity of point and neighborhood territory pixel point, specific method are：
Wherein, centered on q pixel neighborhood territory pixel point, pixel centered on p,Indicate pixel q in neighborhood
The similarity of main structure information, S between point and center pixel p pointspIndicate the main structure information of central pixel point p, SqIndicate neighborhood
The main structure information of pixel q,Indicate the texture information in neighborhood between pixel q points and center pixel p points
Similarity, TpIndicate the main structure information of central pixel point p, TqIndicate the main structure information of neighborhood territory pixel point q.
Both information are added to depth value weight by the present invention by calculating structural similarity and texture similarity
In calculating so that the depth value of central cavity pixel calculates more accurate and effective.
In the another kind preferably embodiment of the present invention, the cavity pixel complementing method is：
1) double using the joint based on direction of fusion structure and texture information for the pixel of hollow sectors fringe region
Side filtering algorithm is filled up：
Wherein, DpRefer to the depth value of center pixel to be measured, neighborhood φpFor：
φp={ q=(qx, qy)|px-w≤qx≤px+ w, py-w≤qy≤py+ w }, using (2w+1) × (2w+1) sizes
Filter window, the w are positive integer；
In the method, due to using Directional Gaussian Filter algorithms in the proximity of spatial domain so that
Can be the neighborhood territory pixel tax that same or similar direction is in empty pixel gradient during calculating neighborhood territory pixel weights
Give higher weight so that weights are accurately calculated.
2) for the pixel of hollow sectors non-edge, using the part based on direction of fusion structure and texture information
Joint bilateral filtering algorithm is filled up：
Neighborhood ΩpIt is adaptivity, size is decided by that empty pixel is in conplane nearest edge to it
Distance；
3) it for the pixel of non-hollow sectors non-edge, is filtered using three side of joint of fusion structure and texture information
Algorithm repair and is filled up：
Wherein,For p points and q point depth value similarities；
4) for the pixel of non-hollow sectors fringe region, using it is identical with hollow sectors fringe region fill up algorithm into
Row fills up reparation.
It, will be all to be repaired due to cavity and the difference of noise and the particularity of edge pixel during image repair
Double image element according to whether in cavity, whether positioned at edge be divided into four classes, use the mixed filtering calculation of algorithms of different combination respectively
Method is repaired, and is improved the accuracy of edge pixel with this, is reduced the error of hole-filling, improves the repairing effect of image.
In the another kind preferably embodiment of the present invention, adjustment non-edge repairs the reparation neighborhood window of algorithm
The method of size is：Window size is decided by empty pixel to being at a distance from conplane nearest edge with it.
The additional aspect and advantage of the present invention will be set forth in part in the description, and will partly become from the following description
Obviously, or practice through the invention is recognized.
Description of the drawings
The above-mentioned and/or additional aspect and advantage of the present invention will become in the description from combination following accompanying drawings to embodiment
Obviously and it is readily appreciated that, wherein：
Fig. 1 is the flow of the depth map cavity restorative procedure based on mixed filtering in a kind of preferred embodiment of the present invention
Figure；
Fig. 2 is experimental group contrast effect figure in a kind of preferred embodiment of the present invention, wherein Fig. 2 (a) is to be repaired original
Image, Fig. 2 (b) are the depth map with cavity；Fig. 2 (c) is the texture information figure of extraction；Fig. 2 (d) is the structural information of extraction
Figure；Fig. 2 is the repairing effect figure of the pixel of hollow sectors fringe region；Fig. 2 (f) is the pixel of hollow sectors non-edge
Repairing effect figure；Fig. 2 (g) is the repairing effect figure of the pixel of non-hollow sectors non-edge；Fig. 2 (h) is non-hollow sectors
The repairing effect figure of the pixel of fringe region；Fig. 2 (i) is repairing effect enlarged drawing；Fig. 2 (j) is original image；
Fig. 3 is that adaptively adjustment calculates the schematic diagram of the size of pixel window in a kind of preferred embodiment of the present invention.
Specific implementation mode
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, and is only used for explaining the present invention, and is not considered as limiting the invention.
In the description of the present invention, unless otherwise specified and limited, it should be noted that term " installation ", " connected ",
" connection " shall be understood in a broad sense, for example, it may be mechanical connection or electrical connection, can also be the connection inside two elements, it can
, can also indirectly connected through an intermediary, for the ordinary skill in the art to be to be connected directly, it can basis
Concrete condition understands the concrete meaning of above-mentioned term.
The present invention provides a kind of depth map cavity restorative procedure based on mixed filtering, the present invention exist according to depth image
The problem of acquisition process, it is proposed that corresponding recovery technique to obtain more accurate depth image, while also promoting
The development of the research field of depth information is used into stereoscopic vision, three-dimensional reconstruction etc., as shown in Figure 1 comprising following step
Suddenly：
S1 obtains complex pattern to be repaired, has cavity on the complex pattern to be repaired.In a kind of preferred embodiment of the present invention
In, cavity can be taken above using GroundTruth as perfect depth map, result figure and GroundTruth will be repaired
It is compared, it can be determined that experiment effect.So the present invention imitates depth image collecting device using artificial,
Take on GroundTruth cavity (cavity that general depth collecting device occurs is easy edge region, thus it is a kind of more
In preferred embodiment, it is also edge region to take cavity), it is as the depth map with cavity in this experiment, i.e., to be repaired
Complex pattern.
S2 identifies that hole region, the region that wherein depth value is 0 are cavity.
S3 calculates the priority of all hole-filling area pixels.The computational methods of priority m are：M=r1+ λ × r2, λ
>0,
Wherein, λ is the parameter of weight shared by support r1 and confidence level r2 when determining to calculate priority.
Neighborhood territory pixel point support r1 is the ratio that total neighborhood territory pixel is accounted for by the known neighborhood territory pixel for the empty pixel filled up
It determines, the filter window is dimensioned to k*k, and the k is the positive integer more than 1；
Confidence level r2 spatial domain proximity and grey value similarity between neighborhood territory pixel and pixel to be filled up determine that formula is such as
Under：
Wherein,For spatial neighbor degree,For grey value similarity,
Wherein, subscript d indicates information in depth map, and subscript c indicates that information comes from cromogram, subscript s representation spaces
Domain, subscript r indicate pixel coverage domain,Indicate the spatial neighbor degree between pixel, qxFor the abscissa of pixel q, qyFor picture
The ordinate of vegetarian refreshments q, the neighborhood territory pixel point of pixel, p centered on qxFor the abscissa of pixel p, pixel centered on p, pyFor picture
The ordinate of vegetarian refreshments p,Indicate the grey value similarity between pixel q points and center pixel p points, I in neighborhoodpFor
The gray value of p points, IqFor the gray value of q points, σsFor the spatial domain parameter in Euclidean distance formula, σrFor the standard deviation of Gauss formula,
Determine the smoothness of Gaussian filter function, σrBigger, the frequency band of Gaussian function is wider, and what is showed on the image is more flat
It is sliding.Pass through σrIt can determine the performance of this algorithm, while also define the variation range of each similarity factor between pixel.
The present invention is provided with preferential before the depth for calculating each empty pixel to be measured for all unknown empty pixels
Grade, that is, fill up the degree of priority of sequence.All empty pixels can be according to its priority in hole region, and will be greater than threshold value
Pixel is put into priority query, when filling cavity, can preferentially be filled up to the pixel in priority query, be improved and fill up
Efficiency avoids calculating error because ignoring caused by empty pixel fills up priority orders.
When calculating, an initial threshold is set for priority m first, which is to calculate priority according to first
The data obtained and set the highest values of m be initial threshold, the present invention in initial threshold be 1.5.Later, pass through m=r1+ λ × r2
Calculate the priority m of all pixels in hole region.All empty pixels more than the threshold value are inserted into priority query
In TSet, and calculate the depth value of each pixel in the queue.Due to often calculating a pixel value can all cavity be closed on to it
The support and confidence level of pixel have an impact, thus the priority of all unknown pixels is not unalterable, be need with
The depth for more and more empty pixels is calculated and the newer process of dynamic.Therefore whenever having filled up institute in current queue
After having empty pixel, the priority of empty edge pixel and the update threshold value of adaptivity can be updated, and will according to new threshold value
In empty pixel arrangement to queue.Here threshold value is that adaptivity changes, if pixel not mended in queue, that is, waits filling up
Pixel quantity is 0, then threshold value is reduced by 0.1, otherwise increases 0.1, by adjusting threshold value so that the pixel of highest priority is always
It can preferentially be calculated.
S4, the sequence descending according to priority orders carry out depth value calculating to empty pixel.
S5 extracts its main structure information and texture information by coloured image, for each pixel of hole region,
Obtain the structural similarity and texture similarity of the pixel and neighborhood territory pixel point.
In another preferred embodiment of the present invention, the method for extracting main structure information and texture information is：
Image I resolves into the two-part linear combination of main structure information S, texture information T, indicates as follows：
I=S+T,
Define Total Variation：Wherein, Ω is image-region,
Establish energy theorem：
Allow main structure information S to be output close to artwork I, 1 × TV of λ (S) are TV canonicals
Changing item, λ 1 is regularization parameter, the weight of first item and Section 2 in balanced type,
Texture information is just acquired by formula T=I-S in image,
The pixel is obtained for each pixel of hole region based on main structure information S, the texture information T sought
The structural similarity and texture similarity of point and neighborhood territory pixel point, specific method are：
Wherein, centered on q pixel neighborhood territory pixel point, pixel centered on p,Indicate pixel q in neighborhood
The similarity of main structure information, S between point and center pixel p pointspIndicate the main structure information of central pixel point p, SqIndicate neighborhood
The main structure information of pixel q,Indicate the texture information in neighborhood between pixel q points and center pixel p points
Similarity, TpIndicate the main structure information of central pixel point p, TqIndicate the main structure information of neighborhood territory pixel point q.
Structural similarity and texture similarity is added in the present invention in existing algorithm, and both information are added to neighbour
In the calculating of domain pixel depth value weight so that the depth value of central cavity pixel calculates more accurate and effective.
Specifically calculating process is：Energy function based on the full variation of image is in the full statistical property two for becoming subitem and noise
It realizes and minimizes under term restriction, therefore, it is original image, u to enable u0For by the image of noise pollution, then
u0=u+n
In formula, n is random Gaussian noise, and mean value is 0, i.e. E (n)=0, and variance E (n2)=σ2, then image u's is total
Variation Model can be expressed as：
Wherein, Ω indicates image-region.Total variation considers the minimization problem of the energy function with qualifications, limit
Fixed condition is the statistical information of noise.
The statistical information of noise includes the mean value qualifications ∫ of noiseΩ(u-u0) dxdy=μ and variance qualifications ∫Ω
(u-u0)2Dxdy=σ2.The minimization problem of energy function under qualifications can be coupled by coefficient non-limiting condition come
Research is equivalent to and is converted into the non-limiting minimization problem of solution：
Wherein, γ, λ are two regularization parameters and Lagrange multiplier.Since the mean value of system noise is 0, then exist
In above formula
∫Ω(u-u0)2Dxdy=∫ΩNdxdy=0
Due toA constant term, or more above formula can be expressed as：
Therefore, the energy function model of image total variation is：
By the minimum value for seeking energy function so that noise is eliminated, to obtain clear image u.
In order to acquire the minimum value of energy function E (u), it is necessary first to obtain the Euler's square of E (u) by Variation TheoremSo
Wherein,For gradient operator,It indicates to seek divergence to the content in bracket, therefore full variation
The formula of model is evolved into：
In order to make the solution of full variational formulation meet qualifications, then the solution u (x, y, t) of above formula at any time is required for full
Sufficient variance qualifications, i.e.,
∫Ω(u (x, y, t)-u0)2Dxdy=σ2
In equation evolutionary process, need to ensure that the variance of u (x, y, t) is moderate constant, therefore
Expansion can obtain
It willSubstitute the u in above formulat, can obtain
So, it is only necessary to the value of regularization parameter λ can be so that above formula establishment, can then ensure that full variation is original
Non trivial solution meets variance qualifications, and limited minimization problem is just addressed.By above formula can derived parameter λ calculating it is public
Formula is：
In actual calculating process, positive number ε one small is added in the denominator, it in this way can be to avoid appearance 0 in denominator
Situation, so can obtain
Write above formula as discrete form, then its m+1 times discrete iteration general formula is, wherein Δ t is iteration step length.
Above-mentioned iteration general formula explanation, when image u is unknown, the initial value of iteration is u(0)=u0；By successive iteration
Afterwards so thatIllustrate u(m+1)=u(m), iteration reaches stable state, then u=u(m)As stable solution, that is, final restore approach value by the image of noise pollution.
Total Variation is that the angle of TV regularizations can also be used to the structure in constraint image and texture.Work as texture scale
When smaller, noise is normally behaved as, Total Variation is for eliminating noise；It is Total Variation when texture scale is larger
The main structure and texture in image can be efficiently extracted.
S6 first determines whether empty pixel is in fringe region for hole region, if being in non-edge,
Using joint bilateral filtering algorithm of the part based on direction of fusion structure and texture information, main calculation methods are depth value neighbours
What the depth value weighted average of domain pixel obtained, the weights of each neighborhood territory pixel by the neighborhood territory pixel that is got from image with
Spatial domain proximity, grey value similarity and structural similarity, the texture similarity of its center pixel obtain, if empty non-edge
Still there is the pixel that do not fill up in region, then returns to step S4；If it is not, starting to fill up fringe region, tied using fusion
The joint bilateral filtering algorithm based on direction of structure and texture information, the reparation neighborhood window that non-edge repairs algorithm are certainly
Adaptability, and the range of this algorithm is fixed；
S7 to the end of all hole region repairings and then repairs all non-hole regions, the non-non- side of hole region
The reparation of edge region is the three side filtering algorithm of joint by fusion structure and texture information, the side of filling up of non-cavity non-edge
Method and the non-edge of hole region to fill up mode identical
In the another kind preferably embodiment of the present invention, the cavity pixel complementing method is：
1) double using the joint based on direction of fusion structure and texture information for the pixel of hollow sectors fringe region
Side filtering algorithm is filled up：
Wherein, DpRefer to the depth value of center pixel to be measured, neighborhood φpFor：
φp={ q=(qx, qy)|px-w≤qx≤px+ w, py-w≤qy≤py+ w }, using (2w+1) × (2w+1) sizes
Filter window, the w are positive integer；
In the method, due to using Directional Gaussian Filter algorithms in the proximity of spatial domain so that
Can be the neighborhood territory pixel tax that same or similar direction is in empty pixel gradient during calculating neighborhood territory pixel weights
Give higher weight so that weights are accurately calculated.
2) for the pixel of hollow sectors non-edge, using the part based on direction of fusion structure and texture information
Joint bilateral filtering algorithm is filled up：
Neighborhood ΩpIt is adaptivity, size is decided by that empty pixel is in conplane nearest edge to it
Distance；
3) it for the pixel of non-hollow sectors non-edge, is filtered using three side of joint of fusion structure and texture information
Algorithm repair and is filled up：
Wherein,For p points and q point depth value similarities；
4) for the pixel of non-hollow sectors fringe region, using it is identical with hollow sectors fringe region fill up algorithm into
Row fills up reparation.
It, will be all to be repaired due to cavity and the difference of noise and the particularity of edge pixel during image repair
Double image element according to whether in cavity, whether positioned at edge be divided into four classes, use the mixed filtering calculation of algorithms of different combination respectively
Method is repaired, and is improved the accuracy of edge pixel with this, is reduced the error of hole-filling, improves the repairing effect of image.
The present invention solves the problems, such as that depth image occurs empty in gatherer process, is proposed first against whole filling-up hole strategy
Priority is set for all pixels in cavity, priority depends on neighborhood territory pixel to the support of central cavity pixel and can
Reliability.During hole-filling, as unknown pixel is constantly padded, empty edge constantly reduces, this algorithm can not
The priority of the pixel at the disconnected empty edge of update, while the update threshold value of meeting adaptivity, ensure that the pixel of highest priority is total
It can preferentially be filled up.The hole-filling priority policy of the present invention considers not only the location information of pixel, gray scale in field
Value, depth value information, and consider neighborhood territory pixel number, priority value cycle calculations avoid filling up because ignoring empty pixel
Error is calculated caused by priority orders.
The entirely depth map cavity restorative procedure based on mixed filtering proposed by the present invention is in joint bilateral filtering principle
On the basis of combine two kinds of information of texture and structure, also have preferable handling result for the image of texture complexity, this method is not only
With validity, also general applicability.
Since objects in images marginal information amount is big, and for the accuracy requirement of pixel value height, therefore the present invention is first right
The pixel of non-edge is filled up, and when completion, the neighborhood territory pixel of fringe region is substantially repaired.Fringe region is repaiied again
Multiple, at this moment, acquisition is put forward more information by fringe region reparation, and the effect of reparation is more preferable.
Fig. 2 is the reparation example of a complex pattern to be repaired, wherein Fig. 2 (a) is original image to be repaired, and Fig. 2 (b) is band
The depth map in cavity；Fig. 2 (c) is the texture information figure of extraction；Fig. 2 (d) is the structural information figure of extraction；Fig. 2 is hollow sectors
The repairing effect figure of the pixel of fringe region；Fig. 2 (f) is the repairing effect figure of the pixel of hollow sectors non-edge；Fig. 2
(g) it is the repairing effect figure of the pixel of non-hollow sectors non-edge；Fig. 2 (h) is the pixel of non-hollow sectors fringe region
Repairing effect figure；Fig. 2 (i) is repairing effect enlarged drawing；Fig. 2 (j) is original image.
In the present embodiment, as shown in figure 3, in empty non-edge, the side of the size of neighborhood window is repaired in adjustment
Method is：Window size is decided by empty pixel to being at a distance from conplane nearest edge with it.
Whether the present invention adds the main structure information and texture information of image in the algorithm, locate according to the pixel in image
In boundary, whether in cavity it is divided into four classes, is non-empty non-edge pixels, the non-empty pixel in edge, empty non-edge respectively
Pixel and empty edge pixel carry out empty reparation respectively, obtain the reparation result figure of better quality.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiments or example in can be combined in any suitable manner.
Although an embodiment of the present invention has been shown and described, it will be understood by those skilled in the art that：Not
In the case of being detached from the principle of the present invention and objective a variety of change, modification, replacement and modification can be carried out to these embodiments, this
The range of invention is limited by claim and its equivalent.
Claims (6)
1. a kind of depth map cavity restorative procedure based on mixed filtering, which is characterized in that include the following steps：
S1 obtains depth image to be repaired；
S2 identifies hole region, and the pixel that depth value is 0 is empty pixel；
S3 is arranged priority initial threshold, calculates the priority of all hole-filling area pixels, priority passes through empty pixel
Support r1 and confidence level r2 determined；
S4, the pixel that priority is more than threshold value are put into priority query, and the pixel in queue is suitable according to from big to small
Sequence carries out the calculating of depth value；
S5 extracts its main structure information and texture information by Total Variation from coloured image；
S6 first determines whether empty pixel is in fringe region for hole region, if being in non-edge, uses
Joint bilateral filtering algorithm of the part based on direction of fusion structure and texture information, main calculation methods are depth value neighborhood pictures
The depth value weighted average of element obtains, the weights of each neighborhood territory pixel by the neighborhood territory pixel that is got from image with wherein
Spatial domain proximity, grey value similarity and structural similarity, the texture similarity of imago element obtain, if empty non-edge
Still there is the pixel that do not fill up, then returns to step S4；If it is not, start to fill up fringe region, using fusion structure and
The joint bilateral filtering algorithm based on direction of texture information, the reparation neighborhood window that non-edge repairs algorithm is adaptive
Property；
S7 to the end of all hole region repairings and then repairs all non-hole regions, non-hole region non-edge area
It is three side filtering algorithm of joint by fusion structure and texture information that domain, which is repaired, the complementing method of non-cavity non-edge with
The non-edge of hole region to fill up mode identical.
2. the depth map cavity restorative procedure according to claim 1 based on mixed filtering, which is characterized in that the step
In S2, neighborhood territory pixel the point support r1 and confidence level r2 of pixel to be filled up are calculated,
Neighborhood territory pixel point support r1 is to account for the ratio of total neighborhood territory pixel by the known neighborhood territory pixel for the empty pixel filled up to determine,
The filter window is dimensioned to k*k, and the k is the positive integer more than 1；
Confidence level r2 spatial domain proximity and grey value similarity between neighborhood territory pixel and pixel to be filled up determine that formula is as follows：
Wherein,For spatial neighbor degree,For grey value similarity,
Wherein, subscript d indicates information in depth map, and subscript c indicates that information comes from cromogram, subscript s representation spaces domain, mark
R indicates pixel coverage domain,Indicate the spatial neighbor degree between pixel, qxFor the abscissa of pixel q, qyFor pixel q's
Ordinate, the neighborhood territory pixel point of pixel, p centered on qxFor the abscissa of pixel p, pixel centered on p, pyFor pixel p's
Ordinate,Indicate the grey value similarity between pixel q points and center pixel p points, I in neighborhoodpFor the ash of p points
Angle value, IqFor the gray value of q points, σsFor the spatial domain parameter in Euclidean distance formula, σrFor the standard deviation of Gauss formula.
3. the depth map cavity restorative procedure according to claim 1 based on mixed filtering, which is characterized in that the step
In S3, the computational methods of priority m are：
M=r1+ λ × r2, λ>0,
Wherein, λ is the parameter of weight shared by support r1 and confidence level r2 when determining to calculate priority.
4. the depth map cavity restorative procedure according to claim 1 based on mixed filtering, which is characterized in that extract main knot
Structure information and the method for texture information are：
Image I resolves into the two-part linear combination of main structure information S, texture information T, indicates as follows：
I=S+T,
The definition Total Variation of image structure information：
TV (S)=∫Ω| ▽ S | dxdy,
Energy theorem is established for image：
WhereinAllow main structure information S to be output close to artwork I, 1 × TV of λ (S) are TV canonicals
Changing item, λ 1 is regularization parameter, the weight of first item and Section 2 in balanced type,
Texture information is just acquired by formula T=I-S in image,
Based on main structure information S, the texture information T sought, for each pixel of hole region, obtain the pixel with
The structural similarity and texture similarity of neighborhood territory pixel point, specific method are：
Wherein, centered on q pixel neighborhood territory pixel point, pixel centered on p,Indicate neighborhood in pixel q points with
The similarity of main structure information, S between center pixel p pointspIndicate the main structure information of central pixel point p, SqIndicate neighborhood territory pixel
The main structure information of point q,Indicate the similar of the texture information in neighborhood between pixel q points and center pixel p points
Degree, TpIndicate the main structure information of central pixel point p, TqIndicate the main structure information of neighborhood territory pixel point q.
5. the depth map cavity restorative procedure according to claim 1 based on mixed filtering, which is characterized in that the cavity
Pixel complementing method is：
1) for the pixel of hollow sectors fringe region, using the bilateral filter of the joint based on direction of fusion structure and texture information
Wave algorithm is filled up：
Wherein, DpRefer to the depth value of center pixel to be measured, neighborhood φpFor：φp={ q=(qx,qy)|px-w≤qx≤px+w,
py-w≤qy≤py+ w }, using the filter window of (2w+1) × (2w+1) sizes, the w is positive integer；
2) for the pixel of hollow sectors non-edge, the part based on direction using fusion structure and texture information is combined
Bilateral filtering algorithm is filled up：
Neighborhood ΩpAdaptivity, size be decided by empty pixel to its be in conplane nearest edge away from
From；
3) for the pixel of non-hollow sectors non-edge, using the three side filtering algorithm of joint of fusion structure and texture information
Repair and fills up：
Wherein,For p points and q point depth value similarities；
4) it for the pixel of non-hollow sectors fringe region, fills up algorithm using identical with hollow sectors fringe region and is filled out
It studies for a second time courses one has flunked multiple.
6. the depth map cavity restorative procedure according to claim 5 based on mixed filtering, which is characterized in that adjustment cavity
The method of size that non-edge repairs the reparation neighborhood window of algorithm is：Window size be decided by empty pixel to at it
Distance in conplane nearest edge.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810443580.XA CN108648221B (en) | 2018-05-10 | 2018-05-10 | Depth map hole repairing method based on hybrid filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810443580.XA CN108648221B (en) | 2018-05-10 | 2018-05-10 | Depth map hole repairing method based on hybrid filtering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108648221A true CN108648221A (en) | 2018-10-12 |
| CN108648221B CN108648221B (en) | 2021-09-17 |
Family
ID=63754436
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810443580.XA Active CN108648221B (en) | 2018-05-10 | 2018-05-10 | Depth map hole repairing method based on hybrid filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108648221B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636732A (en) * | 2018-10-24 | 2019-04-16 | 深圳先进技术研究院 | A kind of empty restorative procedure and image processing apparatus of depth image |
| CN109685732A (en) * | 2018-12-18 | 2019-04-26 | 重庆邮电大学 | A kind of depth image high-precision restorative procedure captured based on boundary |
| CN109712067A (en) * | 2018-12-03 | 2019-05-03 | 北京航空航天大学 | A kind of virtual viewpoint rendering method based on depth image |
| CN110599513A (en) * | 2019-09-04 | 2019-12-20 | 南京邮电大学 | Binocular vision image edge detection and target tracking method |
| CN110866882A (en) * | 2019-11-21 | 2020-03-06 | 湖南工程学院 | Layered joint bilateral filtering depth map restoration algorithm based on depth confidence |
| CN111044988A (en) * | 2020-01-06 | 2020-04-21 | 青梧桐有限责任公司 | Fine processing method for laser radar scanning house type graph |
| CN111179195A (en) * | 2019-12-27 | 2020-05-19 | 西北大学 | Depth image hole filling method and device, electronic equipment and storage medium thereof |
| CN111311521A (en) * | 2020-03-12 | 2020-06-19 | 京东方科技集团股份有限公司 | Image restoration method and device and electronic equipment |
| CN111369611A (en) * | 2018-12-25 | 2020-07-03 | 曜科智能科技（上海）有限公司 | Image pixel depth value optimization method and device, equipment and storage medium thereof |
| CN111383185A (en) * | 2018-12-29 | 2020-07-07 | 海信集团有限公司 | Hole filling method based on dense disparity map and vehicle-mounted equipment |
| CN111986124A (en) * | 2020-09-07 | 2020-11-24 | 北京凌云光技术集团有限责任公司 | Filling method and device for missing pixels of depth image |
| CN112651897A (en) * | 2020-12-30 | 2021-04-13 | 成都星时代宇航科技有限公司 | Pixel repairing method and device, electronic equipment and computer readable storage medium |
| US11074708B1 (en) | 2020-01-06 | 2021-07-27 | Hand Held Products, Inc. | Dark parcel dimensioning |
| CN113242419A (en) * | 2021-04-30 | 2021-08-10 | 电子科技大学成都学院 | 2D-to-3D method and system based on static building |
| CN113763273A (en) * | 2021-09-07 | 2021-12-07 | 北京的卢深视科技有限公司 | Face complementing method, electronic device and computer readable storage medium |
| CN113935914A (en) * | 2021-10-08 | 2022-01-14 | 北京的卢深视科技有限公司 | Depth image restoration method, electronic device and storage medium |
| CN114418902A (en) * | 2022-03-31 | 2022-04-29 | 深圳市华汉伟业科技有限公司 | Image restoration method and computer-readable storage medium |
| CN116721143A (en) * | 2023-08-04 | 2023-09-08 | 南京诺源医疗器械有限公司 | Depth information processing device and method for 3D medical image |
| CN117314792A (en) * | 2023-11-27 | 2023-12-29 | 广州市大湾区虚拟现实研究院 | Depth image restoration method based on RGB image edge capture |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310420A (en) * | 2013-06-19 | 2013-09-18 | 武汉大学 | Method and system for repairing color image holes on basis of texture and geometrical similarities |
| CN103955891A (en) * | 2014-03-31 | 2014-07-30 | 中科创达软件股份有限公司 | Image restoration method based on block matching |
| CN104601972A (en) * | 2014-12-17 | 2015-05-06 | 清华大学深圳研究生院 | Method for synthesizing free viewpoint by image inpainting |
| CN106485672A (en) * | 2016-09-12 | 2017-03-08 | 西安电子科技大学 | Improved Block- matching reparation and three side Steerable filter image enchancing methods of joint |
| CN106851248A (en) * | 2017-02-13 | 2017-06-13 | 浙江工商大学 | Based on openness image repair priority computational methods |
| CN106920263A (en) * | 2017-03-10 | 2017-07-04 | 大连理工大学 | Undistorted integration imaging 3 D displaying method based on Kinect |
-
2018
- 2018-05-10 CN CN201810443580.XA patent/CN108648221B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310420A (en) * | 2013-06-19 | 2013-09-18 | 武汉大学 | Method and system for repairing color image holes on basis of texture and geometrical similarities |
| CN103955891A (en) * | 2014-03-31 | 2014-07-30 | 中科创达软件股份有限公司 | Image restoration method based on block matching |
| CN104601972A (en) * | 2014-12-17 | 2015-05-06 | 清华大学深圳研究生院 | Method for synthesizing free viewpoint by image inpainting |
| CN106485672A (en) * | 2016-09-12 | 2017-03-08 | 西安电子科技大学 | Improved Block- matching reparation and three side Steerable filter image enchancing methods of joint |
| CN106851248A (en) * | 2017-02-13 | 2017-06-13 | 浙江工商大学 | Based on openness image repair priority computational methods |
| CN106920263A (en) * | 2017-03-10 | 2017-07-04 | 大连理工大学 | Undistorted integration imaging 3 D displaying method based on Kinect |
Non-Patent Citations (6)
| Title |
|---|
| ANH VU LE 等: ""Directional Joint Bilateral Filter for Depth Images"", 《SENSORS (BASEL)》 * |
| FEI QI 等: ""Structure guided fusion for depth map inpainting"", 《PATTERN RECOGNITION LETTERS》 * |
| LI XU 等: ""Structure Extraction from Texture via Relative Total Variation"", 《ACM TRANSACTION ON GRAPHICS》 * |
| YINGHUA SHEN 等: ""Depth map enhancement method based on joint bilateral filter"", 《2014 7TH INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING》 * |
| 胡天佑 等: ""基于超像素分割的深度图像修复算法"", 《光电子·激光》 * |
| 陈丽: ""纹理图像的结构提取方法研究"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636732A (en) * | 2018-10-24 | 2019-04-16 | 深圳先进技术研究院 | A kind of empty restorative procedure and image processing apparatus of depth image |
| CN109712067A (en) * | 2018-12-03 | 2019-05-03 | 北京航空航天大学 | A kind of virtual viewpoint rendering method based on depth image |
| CN109685732A (en) * | 2018-12-18 | 2019-04-26 | 重庆邮电大学 | A kind of depth image high-precision restorative procedure captured based on boundary |
| CN109685732B (en) * | 2018-12-18 | 2023-02-17 | 重庆邮电大学 | High-precision depth image restoration method based on boundary capture |
| CN111369611B (en) * | 2018-12-25 | 2023-06-02 | 曜科智能科技（上海）有限公司 | Image pixel depth value optimization method, device, equipment and storage medium thereof |
| CN111369611A (en) * | 2018-12-25 | 2020-07-03 | 曜科智能科技（上海）有限公司 | Image pixel depth value optimization method and device, equipment and storage medium thereof |
| CN111383185A (en) * | 2018-12-29 | 2020-07-07 | 海信集团有限公司 | Hole filling method based on dense disparity map and vehicle-mounted equipment |
| CN111383185B (en) * | 2018-12-29 | 2023-09-22 | 海信集团有限公司 | Hole filling method based on dense parallax map and vehicle-mounted equipment |
| CN110599513A (en) * | 2019-09-04 | 2019-12-20 | 南京邮电大学 | Binocular vision image edge detection and target tracking method |
| CN110866882B (en) * | 2019-11-21 | 2021-09-07 | 湖南工程学院 | Layered joint bilateral filtering depth map repairing method based on depth confidence |
| CN110866882A (en) * | 2019-11-21 | 2020-03-06 | 湖南工程学院 | Layered joint bilateral filtering depth map restoration algorithm based on depth confidence |
| CN111179195B (en) * | 2019-12-27 | 2023-05-30 | 西北大学 | Depth image cavity filling method and device, electronic equipment and storage medium thereof |
| CN111179195A (en) * | 2019-12-27 | 2020-05-19 | 西北大学 | Depth image hole filling method and device, electronic equipment and storage medium thereof |
| US11074708B1 (en) | 2020-01-06 | 2021-07-27 | Hand Held Products, Inc. | Dark parcel dimensioning |
| CN111044988A (en) * | 2020-01-06 | 2020-04-21 | 青梧桐有限责任公司 | Fine processing method for laser radar scanning house type graph |
| CN111311521A (en) * | 2020-03-12 | 2020-06-19 | 京东方科技集团股份有限公司 | Image restoration method and device and electronic equipment |
| CN111986124B (en) * | 2020-09-07 | 2024-05-28 | 凌云光技术股份有限公司 | Filling method and device for missing pixels of depth image |
| CN111986124A (en) * | 2020-09-07 | 2020-11-24 | 北京凌云光技术集团有限责任公司 | Filling method and device for missing pixels of depth image |
| CN112651897A (en) * | 2020-12-30 | 2021-04-13 | 成都星时代宇航科技有限公司 | Pixel repairing method and device, electronic equipment and computer readable storage medium |
| CN112651897B (en) * | 2020-12-30 | 2024-05-03 | 成都星时代宇航科技有限公司 | Pixel repairing method, device, electronic equipment and computer readable storage medium |
| CN113242419A (en) * | 2021-04-30 | 2021-08-10 | 电子科技大学成都学院 | 2D-to-3D method and system based on static building |
| CN113763273A (en) * | 2021-09-07 | 2021-12-07 | 北京的卢深视科技有限公司 | Face complementing method, electronic device and computer readable storage medium |
| CN113935914A (en) * | 2021-10-08 | 2022-01-14 | 北京的卢深视科技有限公司 | Depth image restoration method, electronic device and storage medium |
| CN114418902B (en) * | 2022-03-31 | 2022-07-08 | 深圳市华汉伟业科技有限公司 | Image restoration method and computer-readable storage medium |
| CN114418902A (en) * | 2022-03-31 | 2022-04-29 | 深圳市华汉伟业科技有限公司 | Image restoration method and computer-readable storage medium |
| CN116721143B (en) * | 2023-08-04 | 2023-10-20 | 南京诺源医疗器械有限公司 | Depth information processing device and method for 3D medical image |
| CN116721143A (en) * | 2023-08-04 | 2023-09-08 | 南京诺源医疗器械有限公司 | Depth information processing device and method for 3D medical image |
| CN117314792A (en) * | 2023-11-27 | 2023-12-29 | 广州市大湾区虚拟现实研究院 | Depth image restoration method based on RGB image edge capture |
| CN117314792B (en) * | 2023-11-27 | 2024-04-19 | 广州市大湾区虚拟现实研究院 | Depth image restoration method based on RGB image edge capture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108648221B (en) | 2021-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108648221A (en) | A kind of depth map cavity restorative procedure based on mixed filtering | |
| CN104408701B (en) | A kind of large scene video image joining method | |
| CN105374019B (en) | A kind of more depth map fusion methods and device | |
| CN102592124B (en) | Geometrical correction method, device and binocular stereoscopic vision system of text image | |
| CN104574393B (en) | A kind of three-dimensional pavement crack pattern picture generates system and method | |
| CN106023303B (en) | A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud | |
| CN108876749A (en) | A kind of lens distortion calibration method of robust | |
| CN105184857B (en) | Monocular vision based on structure light ranging rebuilds mesoscale factor determination method | |
| CN112367514A (en) | Three-dimensional scene construction method, device and system and storage medium | |
| CN109000557B (en) | A kind of nuclear fuel rod pose automatic identifying method | |
| CN101409791B (en) | Video processing system and method | |
| CN105069804B (en) | Threedimensional model scan rebuilding method based on smart mobile phone | |
| CN106709948A (en) | Quick binocular stereo matching method based on superpixel segmentation | |
| CN105608671A (en) | Image connection method based on SURF algorithm | |
| CN108596975A (en) | A kind of Stereo Matching Algorithm for weak texture region | |
| CN110189339A (en) | The active profile of depth map auxiliary scratches drawing method and system | |
| CN107392947A (en) | 2D 3D rendering method for registering based on coplanar four point set of profile | |
| CN106709878B (en) | A kind of rapid image fusion method | |
| CN109523595A (en) | A kind of architectural engineering straight line corner angle spacing vision measuring method | |
| CN101900536A (en) | Method for measuring object surface appearance based on digital picture method | |
| CN109544599A (en) | A kind of three-dimensional point cloud method for registering based on the estimation of camera pose | |
| CN105029691B (en) | A kind of cigarette void-end detection method based on three-dimensional reconstruction | |
| CN103384343B (en) | A kind of method and device thereof filling up image cavity | |
| CN107945221A (en) | A kind of three-dimensional scenic feature representation based on RGB D images and high-precision matching process | |
| CN108663026A (en) | A kind of vibration measurement method |
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 |