CN106447782B - Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image - Google Patents
Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image Download PDFInfo
- Publication number
- CN106447782B CN106447782B CN201610628087.6A CN201610628087A CN106447782B CN 106447782 B CN106447782 B CN 106447782B CN 201610628087 A CN201610628087 A CN 201610628087A CN 106447782 B CN106447782 B CN 106447782B
- Authority
- CN
- China
- Prior art keywords
- point
- dimensional
- skin
- face
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/30—Polynomial surface description
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2210/00—Indexing scheme for image generation or computer graphics
- G06T2210/41—Medical
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Algebra (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
A kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image, the described method comprises the following steps: 1) facial MRI image pretreatment: using 3 d-dem Gaussian smoothing operator, be filtered pretreatment to facial MRI image;2) based on the skin of face tissue reconstruction amendment of connection grid area criterion, steps are as follows: (2.1) correct reconstructed results using trellis connectivity criterion, first have to a reconstructed results according to connectivity and are divided into several connected components;(2.2) result is separated into and needs to solve its area after several connected regions, data are triangle gridding, so being calculated as summing after calculating each triangle gridding area to area.The skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image that smaller, reconstruction that the present invention provides a kind of errors works well.
Description
Technical field
The invention belongs to the three-dimensional reconstruction fields of medical image, are related to a kind of skin of face three-dimensional rebuilding method.
Background technique
Facial soft tissue layering is rebuild mainly to skin, subcutaneous fat, muscle, the tissues such as cartilage carry out rebuilding respectively and
Overlapping display, doctor can not only be intuitive to see the intrinsic geometry of each tissue whereby, can also obtain the phase between tissue
To information such as positional relationships.Therefore, the development of the research not only may be implemented it is preoperative planning with improve plastic and aesthetic surgery at
Power also has important reference significance in terms of clinical medical aided education.
Many scholars both domestic and external expand the three-dimensional reconstruction research work based on medical image, but for soft group of face
The three-dimensional reconstruction research knitted is relatively fewer, and reconstructed object is mostly single organization, the researches that multilayer and layering are rebuild.Together
When, since there is computed tomography (Computed Tomography, CT) image at low cost, pixel value can directly reflect
Tissue density, the features such as being readily appreciated that and handling, most of researchs are using CT image as data source.But CT figure is not so good as magnetic resonance
Image (magnetic resonance imaging, MRI) can more clearly reflect soft tissue structure, therefore, carry out and be based on MRI
The research work that the facial soft tissue layering of image is rebuild is very significant.
Summary of the invention
Error in order to overcome the shortcomings of existing skin of face three-dimensional reconstruction mode is larger, poor, the present invention that rebuilds effect
Provide that a kind of error is smaller, rebuilds the skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image that works well.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image, comprising the following steps:
1) facial MRI image pretreatment
Using 3 d-dem Gaussian smoothing operator, pretreatment is filtered to facial MRI image, reduces noise in image pair
The influence of reconstructed results, the Gaussian kernel of the 3*3*3 used in treatment process.
The process for using three-dimensional Gaussian core to filter is Gaussian kernel, which is placed in three-dimensional voxel, and is successively translated makes in core
The excessively each pixel of the heart channel of Hang-Shaoyin, and the gray value of each pixel of Gauss kernel covering is multiplied with Gaussian kernel value, it is replaced after product addition
Fall the gray value of core central point.Analyzing three-dimensional Gaussian kernel filtering is it is seen that in addition to boundary point, and each point is after filtering
Value the influence of 26 points (being free of itself) can be all amounted to by each 9 points of adjacent 8 points and upper and lower level of its same layer, so
It can achieve denoising effect well.
2) based on the skin of face tissue reconstruction amendment of connection grid area criterion
Although very complete by the skin histology reconstructed results surface that three-dimensional Gaussian smoothing processing and MC algorithm obtain
It is whole and true to nature, but there is also reconstruction errors inside it.There are many chaotic configurations in the inside of reconstructed results, these structures are obvious
It is not belonging to skin of face tissue.On the one hand the appearance of these structures is because there are its hetero-organization and skin histologies in MRI image
Brightness it is suitable, be on the other hand because human body head invagination anatomical structure (ear canal, nostril) caused by, for former reason
The error of formation, the present invention, which devises, a kind of corrects reconstructed results based on the criterion of trellis connectivity.This method it is specific
Operating procedure is as follows:
(2.1) reconstructed results are corrected using trellis connectivity criterion, first has to a reconstructed results and is divided according to connectivity
For several connected components.
(2.2) result is separated into and needs to solve its area after several connected regions, because data are the triangulation network
Lattice, so be calculated as summing after calculating each triangle gridding area to area, triangle area can by formula (2-1) and
Formula (2-2) calculates, wherein a, and b, c are triangle side length, and s is semi-perimeter, and S is triangle area.
SiFor the area of i-th of triangle, then grid gross area StBeing represented by formula (2-3), wherein n is triangular plate number
Final result SfTake StMaximum grid, i.e. Sf=Max (St)。
Further, the skin of face three-dimensional rebuilding method is further comprising the steps of: 3) being layered ray method and correct Three-dimensional Gravity
Build result;
There are many chaotic configurations in the inside for the reconstructed results that MC method obtains, these structures are obviously not belonging to facial skin
Skin tissue.The appearance of these structures be on the one hand it is suitable with the brightness of skin histology because there are its hetero-organizations in MRI image, separately
It on the one hand is because caused by human body head invagination anatomical structure (ear canal, nostril).Previous action, by being connected to grid area
Criterion, effectively eliminated because other tissue intensities it is suitable caused by influence, but invaginate for head and fall into anatomical structure
It does not eliminate also.
Observation by the result generated to MC algorithm for reconstructing, it is seen that extra structure exists only in reconstructed results
It is internal.And if observer is simultaneously from from top (crown is in downward direction) and bottom, (from shoulder to crown direction) is not carried out
It cannot find internal extra structure, this explanation can be obtained by very if observer only observes from horizontal direction
The reconstructed results to tally with the actual situation.So the thinking of a kind of pair of MC algorithm improvement is to carry out at surface extraction to MC reconstructed results
Reason, final only to retain the structure that can see from horizontal direction, the structure remained in this way is just needed skin of face group
The reconstructed results knitted.Based on such thinking, this paper presents a kind of MC innovatory algorithms, MC reconstructed results are hierarchically handled, to every
Layer retention surface structure gets rid of the internal redundant structure that cannot be seen from horizontal direction, cleverly solves MC weight in this way
Algorithm is built in the problems in skin of face reconstruction, this method is named as layering ray method herein.
(3.1) MC three-dimensional reconstruction result is converted into a series of two dimension slicings, concrete operations are using plane between certain
Every going cutting MC reconstructed results, the cross surface cut each time is one layer of two dimension slicing, when plane completes all cut
After cutting operation, a series of two dimensional slice data corresponding with three-dimensional reconstruction result is just generated.Layering obtains two dimension slicing
After data, the data in each layer are several lines data, and next operation is to get rid of in these lines significantly not
Belong to the lines of surface texture.
The operation of this part is completed using using lines boundary includes information criterion.This Method And Principle is
If the boundary of a lines includes by another lines, which is not belonging to surface portion certainly, can be from current plane
Middle removal.
The inside lines of upper part have obtained effective removal after treatment, but lower half portion is located at the interior of ear canal position
Portion's structure is then kept down as former state, because the internal structure that ear canal structure remove these should is connected directly with outer profile,
So boundary includes that criterion can be removed effectively apart from ear hole, the internal structure of nostril remotely.But to being close to ear hole, nose
Structure redundant structure at hole does not work then.
(3.2) pass through last point of processing, the lines for being significantly not belonging to surface have been had been removed in slice, next
Processing target be get rid of ear canal etc. invagination hole configurations tissue around non-surface texture.In processing in this section, this
Invention simulates the process that human eye is observed from horizontal direction, only retains the structure that can be observed from horizontal direction, and of the present invention
This method is known as the boundary extraction algorithm based on ray.
Its base step of boundary extraction algorithm based on ray is first using level, and vertical and deflection is 45 ° and 135 °
Four cluster straight cuts two dimensional slice data, every straight line can intersect with contour line in section and be formed intersection point, and ray is cut
The case where face, sees the b in Fig. 3-10) figure.It is most marginal in the point that cut-off line intersects with lines after straight line intersects with contour line
Two points, marginal point take 135 ° of direction straight line referring to formula (3-2), and other three kinds of straight line situations are referring to formula (3-3), X in formula
Indicate the set of x coordinate value in all crosspoints, Y indicates the set of y-coordinate in all crosspoints, and P1, P2 are two obtained
Marginal point.The straight line in each direction is according to a certain distance uniformly by section, available a series of marginal point, difference
Extract and merge available four point sets of marginal point of each cut direction.
P1x=min (X), P1y=max (Y);P2x=max (X), P2y=min (Y) (3-2)
P1x=min (X), P1y=min (Y);P2x=max (X), P2y=max (Y) (3-3)
It obtains needing to carry out point set mutual analysis comparison after four point sets and merge, be pertaining only to boundary to finally obtain
The point of structure, the consolidation strategy used herein are as follows: if a point can be used as the profile point in final result, and if only if in addition
Three points, which are concentrated, at least has the point for being less than some given threshold at a distance from the point.By screening merging treatment, finally
As a result the point in is the point that the direction at least differing 45o from two can be observed, is the point for belonging to surface texture, threshold value
Size determines maximum allowable invagination constructional depth, according to the feature of facial hole institutional framework, the threshold value that the present invention chooses
It is 5 millimeters.
By the extraction operation of this step, the point of obtained final result is all the point on three-dimensional reconstruction result surface, is had
The internal structure for eliminating recess inside in lines to effect and being formed.
Further, the skin of face three-dimensional rebuilding method is further comprising the steps of: 4) facial chin hole repairing
The skin of face tissue reconstruction error as caused by facial hole invagination structure has largely been improved,
But because herein in processes in only the horizontal direction on cut, and cut the distance between larger (1 millimeter), and
Cut surface at facial chin almost with this paper is parallel, causes to sample less, finally makes hole occur in revised result
Hole, so next having carried out the processing of holes filling to correction result herein.
Holes filling method is to find boundary edge, then boundary edge head and the tail is connected, the boundary edge of these connections just forms
New tri patch is added completing the filling of hole to carry out triangle division to hole in one hole, holes filling operation,
Because the hole at chin missing is smaller and is parallel to the horizontal plane substantially, holes filling not will cause too big error.
Technical concept of the invention are as follows: the present invention is having studied MRI imaging technical principle and DICOM (Digital
Imaging and Communications in Medicine) on the basis of format medical image standard, complete based on face
The three-dimensional reconstruction of the skin histology of portion's MRI image.In addition, the Algorithms Integration used in experimentation is write and is completed
Medical Image Processing and three-dimensional reconstruction tool software can easily carry out medical image checks analysis and progress facial tissue three
Dimension is rebuild.
The three-dimensional reconstruction of skin of face is based on MC (Marching cubes) method for reconstructing, and uses three-dimensional Gaussian and filter
Reconstructed results are optimized with connected surface area criterion, propose a kind of surface extraction method based on layering ray method, it can
Effectively to remove MC algorithm for reconstructing because of face ear canal, mouth etc. invaginates reconstruction error caused by structure.
Beneficial effects of the present invention are mainly manifested in: being completed the three-dimensional reconstruction of skin of face using MC method, and used
Three-dimensional Gaussian filtering and connected surface area criterion optimize reconstructed results, propose a kind of table based on layering ray method
Face extracting method, can effectively remove MC algorithm for reconstructing because of face ear canal, and mouth etc. invaginates reconstruction error caused by structure.
Detailed description of the invention
Fig. 1 is reconstruction effect before and after three-dimensional Gaussian filter effect, wherein (a) is without the reconstruction knot by filtering processing
Fruit;It (b) is to use the reconstructed results after gaussian filtering.
Fig. 2 is the inside redundant structure in reconstructed results.
Fig. 3 is connection grid area criterion process flow diagram.
Fig. 4 is effect before and after the surface extraction based on connection grid area criterion, wherein a) for without surface extraction
Reconstructed results, b) it is result after connectivity criterion carries out the tiny chaotic configuration that removes in result of surface extraction.
Fig. 5 is that lines boundary includes method Boundary Extraction schematic diagram, wherein left-half is right half part before extraction operation
After extraction operation.
Fig. 6 is ray method Boundary Extraction, wherein a) is b before processing operation) be four kinds of rays cutting signals, c) it is to handle
After operation.
Fig. 7 is layering ray method flow diagram.
Fig. 8 is to extract the point after all boundary points merge to converge.
Fig. 9 is boundary point cloud amendment reconstructed results effect, wherein be a) that internal structure is effectively removed b) is chin
Place appears cavity.
Figure 10 is the tissue that skin of face rebuilds amendment removal, wherein a) observes final reconstructed results, b for different directions)
It is the redundant structure removed during correcting reconstructed results for the reconstructed results comprising internal reconstruction error, c).
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Figure 10, a kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image, including following step
It is rapid:
1) facial MRI image is read in, one of sequence is selected, the MRI image data of selected sequence is read into interior
In depositing, and every image is stacked into three-dimensional voxel architecture according to its correct position, then uses 3 d-dem Gaussian smoothing
Operator is filtered pretreatment to facial MRI image, reduces influence (as shown in Figure 1) of the noise in image to reconstructed results, place
The Gaussian kernel of the 3*3*3 used during reason is as follows:
Upper layer:
0.016254 0.031658 0.016254
0.031658 0.061662 0.031658
0.016254 0.031658 0.016254
Middle layer:
0.031658 0.061662 0.031658
0.061662 0.120101 0.061662
0.031658 0.061662 0.031658
Lower layer:
0.016254 0.031658 0.016254
0.031658 0.061662 0.031658
0.016254 0.031658 0.016254
The process for using the three-dimensional Gaussian core to filter is Gaussian kernel, which is placed in three-dimensional voxel, and is successively translated makes core
Each pixel is passed through at center, and the gray value of each pixel of Gauss kernel covering is multiplied with Gaussian kernel value, is replaced after product addition
Change the gray value of core central point.Analyzing three-dimensional Gaussian kernel filtering is it is seen that in addition to boundary point, and each point is by filtering
Value afterwards can all be amounted to the influence of 26 points (being free of itself), institute by each 9 points of adjacent 8 points and upper and lower level of its same layer
Can achieve denoising effect well.
2) based on the skin of face tissue reconstruction amendment of connection grid area criterion
Classics MC is carried out in the three-dimensional voxel of formation to rebuild, and obtains the reconstructed results of skin of face tissue.Reconstructed results
Observation is got up very complete and true to nature from the outside, but there is also reconstruction errors inside it.(as shown in Figure 2)
The appearance of reconstruction errors be on the one hand it is suitable with the brightness of skin histology because there are its hetero-organizations in MRI image,
On the other hand be because human body head invagination anatomical structure (ear canal, nostril) caused by, for former reason formed error,
The present invention, which devises, a kind of corrects reconstructed results based on the criterion of trellis connectivity.The concrete operation step of this method is such as
Under:
(2.1) reconstructed results are corrected using trellis connectivity criterion, first has to a reconstructed results and is divided according to connectivity
For several connected components.And the judgment criterion of connected region are as follows:
(2.1.1) A is directly connected to B≤and=> A and B have total side;
(2.1.2) A and C indirect communication ≤=> A can achieve C by the direct connected relation of n times (n > 1, n are integer);
(2.1.3) A is connected to D≤and=> A is directly connected to D or indirect communication.
(2.2) result is separated into and needs to solve its area after several connected regions, because data are the triangulation network
Lattice, so be calculated as summing after calculating each triangle gridding area to area, triangle area can by formula (2-1) and
Formula (2-2) calculates, wherein a, and b, c are triangle side length, and s is semi-perimeter, and S is triangle area.
SiFor the area of i-th of triangle, then grid gross area StBeing represented by formula (2-3), wherein n is triangular plate number
Final result SfTake StMaximum grid, i.e. Sf=Max (St)。
The flow chart of entire area criterion is as shown in Figure 3.
A) it is the reconstructed results without surface extraction, b in Fig. 4) it is to carry out surface extraction by connectivity criterion to remove
As a result the result after tiny chaotic configuration in.Compare a) figure in Fig. 4 and b) figure be not difficult to find out, through connection grid criterion
Result that treated has very big improvement, but because invagination structure causes reconstruction errors to be solved not yet at all in result,
Main elaboration causes invagination structure to the analysis and processing of error in next step.To make the reconstruction of skin more meet practical feelings
Condition.
3) layering ray method corrects three-dimensional reconstruction result
(3.1) to 2) obtained MC three-dimensional reconstruction result carries out hierarchical operations after step operation, the layering being layered in ray method
Refer to and MC three-dimensional reconstruction result is converted into a series of two dimension slicings, concrete operations are to remove cutting MC at certain intervals using plane
Reconstructed results, the cross surface cut each time are one layer of two dimension slicing, after plane completes all cutting operations, just
Generate a series of two dimensional slice data corresponding with three-dimensional reconstruction result.
(3.2) after layering obtains two dimensional slice data, the data in each layer are several lines data, next behaviour
Work is to get rid of the lines that surface texture is significantly not belonging in these lines.
The operation of this part is completed using lines boundary includes information criterion.This Method And Principle is if one
The boundary of lines includes that then the lines are not belonging to surface portion certainly by another lines, can be removed from current plane.
The definition on lines boundary: the lines being made of n point, coordinate representation are (Xi, Yi), 1≤i≤n.Then
Its boundary is (MAX (Xi), MAX (Yi)) and (MIN (Xi), MIN (Yi)).It is denoted as (Xma, Yma) and (Xmi, Ymi).
The definition that lines include: for lines L1 and L2, boundary information is denoted as { (Xma1, Yma1), (Xmi1, Ymi1) }
{ (Xma2, Yma2), (Xmi2, Ymi2) }.L1 includes that L2 is set up and if only if formula (3-1).
Xma1 >=Xma2 ∩ Yma1 >=Yma2 ∩ Xmi1≤Ymi1 ∩ Ymi1≤Ymi2 (3-1)
The processing step of this method is that data are divided into several lines, connection lines extraction method and a upper section by connectivity
Middle connection grid method is identical.The boundary information for solving and comparing these lines, obtain between lines comprising result.If one
Lines include another lines, then another lines are internal structure, should be removed from current section.Fig. 5 is illustrated by side
Boundary includes the processing result after method, and as can be seen from the figure the inside lines of upper part have obtained effectively going after treatment
It removes, but lower half portion is located at the internal structure of ear canal position and then keeps down as former state, because ear canal structure go these should
The internal structure removed is connected directly with outer profile, so boundary includes that criterion can be removed effectively apart from ear hole, nostril compared with
The internal structure of distant place.But to ear hole is close to, the structure redundant structure at nostril does not work then.
(3.3) pass through last point of processing, the lines for being significantly not belonging to surface have been had been removed in slice, next
Processing target be get rid of ear canal etc. invagination hole configurations tissue around non-surface texture, i.e., in Fig. 5 right half part under
The internal structure of half part.
Its base step of boundary extraction algorithm based on ray is first using level, and vertical and deflection is 45 ° and 135 °
Four cluster straight cuts two dimensional slice data, every straight line can intersect with contour line in section and be formed intersection point, and ray is cut
The case where face, sees the b in Fig. 3-10) figure.It is most marginal in the point that cut-off line intersects with lines after straight line intersects with contour line
Two points, marginal point take 135 ° of direction straight line referring to formula (3-2), and other three kinds of straight line situations are referring to formula (3-3), X in formula
Indicate the set of x coordinate value in all crosspoints, Y indicates the set of y-coordinate in all crosspoints, and P1, P2 are two obtained
Marginal point.The straight line in each direction is according to a certain distance uniformly by section, available a series of marginal point, difference
Extract and merge available four point sets of marginal point of each cut direction.
P1x=min (X), P1y=max (Y);P2x=max (X), P2y=min (Y) (3-2)
P1x=min (X), P1y=min (Y);P2x=max (X), P2y=max (Y) (3-3)
It obtains needing to carry out point set mutual analysis comparison after four point sets and merge, be pertaining only to boundary to finally obtain
The point of structure, the consolidation strategy used herein are as follows: if a point can be used as the profile point in final result, and if only if in addition
Three points, which are concentrated, at least has the point for being less than some given threshold at a distance from the point.Wherein distance is calculated referring to formula
(3-4), d in formulapqIndicate the distance of p, q point-to-point transmission.Minimum range d of the point to other three setpIt calculates referring to formula (3-
5), in formula n be other three centrostigmas number.Final result selection is referring to formula (3-6), and P is currently processed point in formula
Collection, D are selected threshold value.By such screening merging treatment, the point in final result is at least to differ 45 ° from two
The point that direction can be observed, is the point for belonging to surface texture, and the size of threshold value determines that maximum allowable invagination structure is deep
Degree, according to the feature of facial hole institutional framework, the threshold value chosen herein is 5 millimeters.
dp=min (dp1..., dpn) (3-7)
{p∈P|dp< D } (3-8)
By the extraction operation of this step, the point of obtained final result is all the point on three-dimensional reconstruction result surface, is had
The internal structure for eliminating recess inside in lines to effect and being formed.Fig. 6 is the schematic diagram of this processing.
The algorithm flow (as shown in Figure 7) for the ray boundary detection being entirely layered mainly comprises the following steps and uses plane first
MC reconstructed results are cut, cross surface two-dimensional result is obtained, the processing based on lines boundary comprising criterion is carried out to the result of cross surface
The internal structure in addition to the structure periphery that invaginates is removed, then remaining internal junction is removed using the surface extraction algorithm based on ray
Structure obtains the marginal information of cross surface.
By position of the mobile cutting planes in MC reconstructed results, can be obtained a series of after Boundary Extraction is handled
Point set converged as a result, result point set being merged, a point can be obtained, as shown in Figure 8.
Obtaining the work to be done after Boundary Extraction point converges is to be converged to rebuild come MC before correcting with this surface point
As a result, correction strategy is the point to each reconstructed results, if it converges big Mr. Yu at a distance from middle closest approach with the surface point
A threshold value, the then grid for abandoning the point and being made from it, the size of threshold value determine the maximum invagination constructional depth allowed.Fig. 9 is
Effect after converging amendment MC reconstructed results using boundary point.
4) facial chin hole repairing
As can be seen that the skin of face tissue reconstruction error as caused by facial hole invagination structure from the result figure of Fig. 9
Largely be improved, but because herein in processes in only the horizontal direction on cut, and cut
The distance between larger (1 millimeter), and the cut surface at facial chin almost with this paper is parallel, cause to sample it is less, finally
So that there is hole in revised result, so next having carried out the processing of holes filling to correction result herein.Hole
Hole filling algorithm is to find boundary edge, and then boundary edge head and the tail are connected, and the boundary edge of these connections just constitutes a hole,
To carry out triangle division to hole new tri patch is added to complete the filling of hole, because chin lacks in holes filling operation
Hole at mistake is smaller and is parallel to the horizontal plane substantially, so holes filling not will cause too big error.Final result with
Figure 10 is shown in the comparison of original reconstructed results.A) figure is that different directions observe final reconstructed results, b in figure) it is to be rebuild comprising inside
The reconstructed results of error, c) it is the redundant structure that amendment reconstructed results remove in the process.
The scheme of the present embodiment completes the three-dimensional reconstruction of skin of face using MC method, and uses three-dimensional Gaussian and filter
Wave and connected surface area criterion optimize reconstructed results, propose a kind of surface extraction method based on layering ray method,
MC algorithm for reconstructing can effectively be removed because of face ear canal, mouth etc. invaginates reconstruction error caused by structure.
Claims (3)
1. a kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image, it is characterised in that: the method includes following
Step:
1) facial MRI image pretreatment
Using 3 d-dem Gaussian smoothing operator, pretreatment is filtered to facial MRI image;
2) MC reconstruction is carried out in the three-dimensional voxel of formation, obtains the reconstructed results of skin of face tissue, based on connection grid surface
The skin of face tissue reconstruction amendment of product criterion, steps are as follows:
(2.1) reconstructed results are corrected using trellis connectivity criterion, is divided into if first having to a reconstructed results according to connectivity
Dry connected component;
(2.2) result is separated into and needs to solve its area after several connected regions, data are triangle gridding, so right
Area is calculated as summing after calculating each triangle gridding area, and triangle area is counted by formula (2-1) and formula (2-2)
It calculates, wherein a, b, c are triangle side length, and s is semi-perimeter, and S is triangle area;
SiFor the area of i-th of triangle, then grid gross area StBeing expressed as formula (2-3), wherein n is triangular plate number
Final result SfTake StMaximum grid, i.e. Sf=Max (St);
3) layering ray method corrects three-dimensional reconstruction result, and steps are as follows:
(3.1) MC three-dimensional reconstruction result is converted into a series of two dimension slicings, cutting MC is gone to rebuild at certain intervals using plane
As a result, the cross surface cut each time is one layer of two dimension slicing, after plane completes all cutting operations, just generate
A series of two dimensional slice data corresponding with three-dimensional reconstruction result;
The operation of this part is completed using lines boundary includes information criterion;
(3.2) process that human eye is observed from horizontal direction is simulated, the structure that can be observed from horizontal direction, process are only retained
Are as follows: first using level, the four cluster straight cuts two dimensional slice data that vertical and deflection is 45 ° and 135 °, every straight line meeting
Intersect and formed intersection point with contour line in section, after straight line intersects with contour line, most side in the point that cut-off line intersects with lines
Two points of edge, marginal point take 135 ° of direction straight line referring to formula (3-2), and other three kinds of straight line situations are referring to formula (3-3), formula
Middle X indicates the set of x coordinate value in all crosspoints, and Y indicates the set of y-coordinate in all crosspoints, and P1, P2 are two obtained
A marginal point;The straight line in each direction, uniformly by section, obtains a series of marginal point, mentions respectively according to a certain distance
It takes and the marginal point for merging each cut direction obtains four point sets;
P1x=min (X), P1y=max (Y);P2x=max (X), P2y=min (Y) (3-2)
P1x=min (X), P1y=min (Y);P2x=max (X), P2y=max (Y) (3-3)
It obtains needing to carry out point set mutual analysis comparison after four point sets and merge, be pertaining only to border structure to finally obtain
Point, the consolidation strategy of use an are as follows: if point can be used as the profile point in final result, and if only if in the other three point set
In at least exist one with the point at a distance from less than some given threshold point;By screening merging treatment, in final result
Point is the point that the direction at least differing 45 ° from two can be observed, is the point for belonging to surface texture, and the size of threshold value determines
Maximum allowable invagination constructional depth;
By the extraction operation of this step, the point of obtained final result is all the point on three-dimensional reconstruction result surface.
2. a kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image as described in claim 1, it is characterised in that:
The skin of face three-dimensional rebuilding method is further comprising the steps of: 4) facial chin hole repairing, process are as follows:
Holes filling method are as follows: find boundary edge, then boundary edge head and the tail are connected, the boundary edge of these connections just constitutes one
New tri patch is added to carry out triangle division to hole to complete the filling of hole in a hole, holes filling operation.
3. a kind of skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image as claimed in claim 1 or 2, feature exist
In: in the step 1), the Gaussian kernel of the 3*3*3 used in treatment process, the process for using three-dimensional Gaussian core to filter is will be high
This core is placed in three-dimensional voxel and successively translation makes core center by each pixel, and each pixel of Gauss kernel covering
Gray value be multiplied with Gaussian kernel value, the gray value of core central point is replaced after product addition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610628087.6A CN106447782B (en) | 2016-08-02 | 2016-08-02 | Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610628087.6A CN106447782B (en) | 2016-08-02 | 2016-08-02 | Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106447782A CN106447782A (en) | 2017-02-22 |
CN106447782B true CN106447782B (en) | 2019-05-28 |
Family
ID=58184501
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610628087.6A Active CN106447782B (en) | 2016-08-02 | 2016-08-02 | Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106447782B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107132245A (en) * | 2017-06-06 | 2017-09-05 | 浙江大学 | Crop root quick nondestructive three-dimensional reconstruction apparatus and method based on nmr imaging technique |
CN107730519A (en) * | 2017-09-11 | 2018-02-23 | 广东技术师范学院 | A kind of method and system of face two dimensional image to face three-dimensional reconstruction |
CN112215750A (en) * | 2020-09-14 | 2021-01-12 | 湖北工业大学 | Three-dimensional reconstruction method based on RGB-D data |
CN112288762B (en) * | 2020-10-15 | 2023-05-09 | 西北工业大学 | Discrete iteration reconstruction method for limited angle CT scanning |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103678847A (en) * | 2012-09-24 | 2014-03-26 | 邵雁 | Skin soft tissue expanding auxiliary system |
CN103678848A (en) * | 2012-09-24 | 2014-03-26 | 邵雁 | Setting method of skin soft tissue expander |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2999325A1 (en) * | 2012-12-07 | 2014-06-13 | Commissariat Energie Atomique | ITERATIVE METHOD OF DETERMINING A TWO-DIMENSIONAL OR THREE-DIMENSIONAL IMAGE FROM X-RAY TOMOGRAPHY SIGNALS |
-
2016
- 2016-08-02 CN CN201610628087.6A patent/CN106447782B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103678847A (en) * | 2012-09-24 | 2014-03-26 | 邵雁 | Skin soft tissue expanding auxiliary system |
CN103678848A (en) * | 2012-09-24 | 2014-03-26 | 邵雁 | Setting method of skin soft tissue expander |
Also Published As
Publication number | Publication date |
---|---|
CN106447782A (en) | 2017-02-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106447782B (en) | Skin of face three-dimensional rebuilding method based on nuclear magnetic resonance image | |
CN103679810B (en) | The three-dimensional rebuilding method of liver's CT image | |
US10713856B2 (en) | Medical imaging system based on HMDS | |
US9980691B2 (en) | Method and apparatus for three dimensional viewing of images | |
CN104573309B (en) | Device and method for computer-aided diagnosis | |
CN107895364B (en) | A kind of three-dimensional reconstruction system for the preoperative planning of virtual operation | |
CN106901834A (en) | The preoperative planning of minimally invasive cardiac surgery and operation virtual reality simulation method | |
CN107230206A (en) | A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data | |
CN104462650A (en) | Materialized heart 3D model manufacturing method capable of achieving internal and external structures | |
CN103489198A (en) | Method for partitioning brainstem areas automatically from MR (magnetic resonance) sequence images | |
CN110298804A (en) | One kind is based on generation confrontation network and the decoded medical image denoising method of 3D residual coding | |
CN103035009A (en) | Pulmonary nodule edge rebuilding and partitioning method based on computed tomography (CT) image | |
Chougule et al. | Development of patient specific implants for Minimum Invasive Spine Surgeries (MISS) from non-invasive imaging techniques by reverse engineering and additive manufacturing techniques | |
CN103985122B (en) | The dirtiest extracting method of based on heart CT image | |
Yang et al. | Improving catheter segmentation & localization in 3d cardiac ultrasound using direction-fused fcn | |
CN104899923A (en) | Method for constructing facial prosthesis optimized model based on smile expression geometrical characteristic modification | |
Guerroudji et al. | Automatic brain tumor segmentation, and 3d reconstruction and visualization using augmented reality | |
CN114365188A (en) | Analysis method and product based on VRDS AI inferior vena cava image | |
KR102227735B1 (en) | 3D Modeling method for organ and 3D organ model | |
CN108171659A (en) | A kind of image repair method based on K-SVD dictionaries | |
Zhao et al. | Research on 3D reconstruction algorithm of medical CT image based on parallel contour | |
Smith et al. | Automated torso contour extraction from clinical cardiac MR slices for 3D torso reconstruction | |
Yanamadala et al. | Segmentation of the visible human project®(VHP) female cryosection images within MATLAB® environment | |
CN104217432B (en) | A kind of intracranial vessel bounding box automatic planning and device | |
CN113570724B (en) | Grid sphere corner-protection parameterization method based on inverse sphere projection and application thereof |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |