CN109741265A - A kind of spine CT image bilateral filtering denoising method based on feature selecting - Google Patents
A kind of spine CT image bilateral filtering denoising method based on feature selecting Download PDFInfo
- Publication number
- CN109741265A CN109741265A CN201811386922.5A CN201811386922A CN109741265A CN 109741265 A CN109741265 A CN 109741265A CN 201811386922 A CN201811386922 A CN 201811386922A CN 109741265 A CN109741265 A CN 109741265A
- Authority
- CN
- China
- Prior art keywords
- point
- grid
- data
- image
- point cloud
- 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.)
- Withdrawn
Links
Abstract
The present invention discloses a kind of spine CT image bilateral filtering denoising method based on feature selecting, belong to medical image process field, the deficiency of actual clinical operation and teaching research demand cannot be fully met in order to make up picture quality after conventional denoising method is handled, method proposed by the present invention, enhance interest region while extraneous areas in removing backbone medical image, for the segmentation of later image, positioning, the operations such as three-dimensional reconstruction are laid a good foundation, it is first that backbone point cloud data is spatially horizontal, it is vertical, perpendicular three directions carry out grid, judge the point cloud quantity in each grid, noise spot cloud is divided into again and deviates backbone interest region farther out and sparse point cloud, with backbone interest region still intensive point cloud and the noise spot cloud three classes to mix with backbone interest region farther out, successively go again divided by reach preferably denoising effect.
Description
Technical field
The invention belongs to field of medical technology, in particular to image preprocessing in backbone clinical manipulation and teaching research
Denoising method.
Background technique
Spine CT image will receive scanning spinal shape, surface layer status consideration during tomographic apparatus obtains
Etc. influence and generate many noise jammings, therefore, spine CT image is split, is positioned, the behaviour such as three-dimensional reconstruction
Before work, backbone volume data should be subjected to denoising, it is therefore an objective to while guaranteeing backbone feature itself, improve the matter of image
The smoothness of image after amount and later period three-dimensional reconstruction and visualized operation.When carrying out denoising, need according to backbone number
According to specificity take different denoising methods, to make up the precision of images after the processing of conventional denoising method, quality cannot expire completely
The problem of full border clinical manipulation and teaching research demand, therefore a kind of denoising method for being more suitable for spine CT image is designed to closing
It is important.
Summary of the invention
For this purpose, present invention aims at cannot fully meet reality for the precision of images, quality after conventional denoising method processing
The problem of border clinical manipulation and teaching research demand, provides a kind of spine CT image bilateral filtering denoising side based on feature selecting
Method solves the technical problem in above-mentioned background.
To achieve the above object, the invention provides the following technical scheme: a kind of spine CT image based on feature selecting is double
Side filtering and noise reduction method, it is characterised in that:
The point cloud data of image is subjected to grid first, the point cloud amount in each grid is then counted, for image
The point cloud of non-region of interest is directly removed；
Secondly, successively calculating the bilateral filtering factor of characteristic point cloud and non-characteristic point cloud, then to non-spy in region of interest
Sign point cloud is removed.
Specific implementation step is:
1) CT image volumetric data is established
It by all spine CT image slices of scanning, is sequentially overlapped according to spine form, forms image volumetric data；
2) CT image pane networking
A cuboid is established, including all image datas are all won over by any means, if then cuboid is divided into dry lattice, formation lattice
Net；
3) grid quantity is calculated
Quantity of the computation grid on tri- directions cuboid X, Y, Z；
4) position where the amount of point cloud data in each grid, and confirmation grid is calculated；
5) CT image region of interest and non-region of interest are divided
Data, which are compared, concentrates the region that can highlight spinal shape tentatively to delimit as backbone region of interest, then passes through adaptive
It answers the Canny edge detection method of threshold value to carry out edge detection, the threshold of Canny operator is calculated according to the gray value of image itself
Value finally determines the boundary in backbone interest region；Region except interest zone boundary is uniformly set to non-region of interest；
6) it denoises
6.1) for non-region of interest, directly these point cloud datas are all removed；
6.2) for region of interest, the bilateral filter of characteristic point cloud and non-characteristic point cloud is successively calculated according to the point cloud of different range
The wave factor, the noise spot cloud for recycling the removal of bilateral filtering Denoising Algorithm to mix with region of interest.
Further, CT image pane networking comprises the concrete steps that:
A three-dimensional system of coordinate is initially set up, CT image volumetric data is all read under coordinate system, obtains each data point
Three-dimensional coordinate, while obtaining the maximum in data set in all directions, minimum value xmax、ymax、zmax, xmin、ymin、zmin；
Then according to the maximum of all directions, minimum value, cuboid is established, the cuboid all wins all data points over by any means
Inside；
Then further according to the density degree of data point, the cuboid is divided into several grids, the side of each grid
Length is customized.
Further, the method for calculating grid quantity is:
If each grid side length is d, then grid quantity A, B, C upward in cuboid X, Y, Z tripartite are as follows:
xmax、ymax、zmax, xmin、ymin、zminMaximum, minimum value for image data in each tri- directions X, Y, Z.
Further, the method for calculating the amount of point cloud data in each grid and the position where confirmation grid is:
4.1) coordinate of all data points in volumetric data set is obtained first；
4.2) it then calculates the number of all data points in same grid: appointing some data point M taken in data fields, with
The point is foundation, the neighbor point K on its periphery is found, if M and K is both less than M point to grid where it at a distance from tri- direction X, Y, Z
The obverse distance of lattice, then otherwise decision-point K and M is determined as in other grids in same grid；And so on, it calculates
Out in each grid data point number；
4.3) grid positions where data point are then calculated:
If coordinate of the M point in three directions is Mx、My、Mz, then the grid hash function where this point is shown in formula (2):
Wherein, grid saves the call number, just in the call number in 3 directions of X, Y, Z axis where O, P, Q respectively represent M point
It is the grid positions where M point, with other points in grid also in the grid of the call number, and so on, other grids
Call number is also so established, and the position of grid is determined according to call number.
Further, for region of interest, the specific method of denoising is:
6.2.1 characteristic point and non-characteristic point cloud) are first determined whether
Judge that the cloud genera in characteristic point cloud is also non-characteristic point cloud using neighborhood: calculating separately and put cloud in grid in the domain i
The average Euclidean distance of dataIt is averaged Euclidean distance with the point cloud in this cloud k neighborhoodIt obtainsSetting
One threshold value D0, the threshold value is according to the determination of the Canny edge detection method of adaptive threshold；
Work as Di> D0When, illustrate that i point cloud is characterized a cloud, point cloud in neighborhood is only used when calculating bilateral filtering factor-alpha i.e.
It can；
Work as Di< D0When, illustrate i point Yun Weifei characteristic point cloud, uses the point in whole grids when calculating bilateral filtering factor-alpha
Cloud；
6.2.2 the bilateral filtering factor) is calculated
Bilateral filtering is because of subformula are as follows:
Its formula ofWeight function is filtered for fairing,To keep feature to weigh letter
Number, σcPoint p to neighborhood point distance to its impact factor, value is the radius length of the point to neighborhood point, σsIt is arrived for point p
To the impact factor of p point, value is the standard deviation of neighborhood point for projection of the distance vector of neighborhood point on its normal vector；
<p-pj, two random vector p-p of n > expressionjAnd n, p-pjIndicate data point p to closing on data point pjDistance to
Amount, n indicate data point p to closing on data point pjDistance vector normal vector；
||<p-pj, n > | | indicate data point p to closing on data point pjDistance vector normal direction projection；
||p-pj| | indicate vector p-pjMould it is long；
6.2.3 it) denoises
According to bilateral filtering Denoising Algorithm p'=p- α n, the point cloud data after denoising is obtained, wherein p is raw data points,
P' is the data point after denoising, and n is data point p to closing on data point pjDistance vector normal vector, α be bilateral filtering because
Son.
Compared with prior art, the beneficial effects of the present invention are embodied in: the present invention provides a kind of ridge based on feature selecting
Column CT image bilateral filtering denoising method, is the pretreatment to image, when pre-processing to image, utilizes bilateral filtering
Method, it is ensured that remove extraneous areas while spine image characteristic information itself, while enhancing interest region, be later image
The operations such as segmentation, positioning, three-dimensional reconstruction lay the foundation；Compared with conventional denoising method, better effect is denoised；And do not changing
It is effective to remove uncorrelated noise region while becoming spine CT image data characteristic area, actual clinical is operated and is taught
It learns research and provides technical support.
Other features and advantages of the present invention will illustrate in the following description, and partial become from specification
It is clear that understand through the implementation of the invention.
Detailed description of the invention
Fig. 1 is picture point cloud Data grid illustraton of model.
Specific embodiment
Below in conjunction with the attached drawing in the present invention, the specific embodiment in the present invention is described in detail, it is very bright
Aobvious, it is not whole embodiments that embodiment described, which is a part in the present invention, for the embodiment in the present invention,
One skilled in the relevant art obtains not making the creative labor, and belongs to the range that the present invention is protected.
A kind of spine CT image bilateral filtering denoising method based on feature selecting provided by the invention, is to spine CT figure
A kind of early period of picture pre-processes.Spine CT image is pre-processed, it is therefore an objective to guarantee spine image characteristic information itself
Extraneous areas is removed simultaneously, while enhancing interest region, establishes base for operations such as segmentation, positioning, the three-dimensional reconstructions of later image
Plinth.Since backbone bone image data feature is neat, interest region integrated distribution, so being relatively suitble to denoise using bilateral filtering
Method pre-processes spine CT image.
In an embodiment of the present invention, a kind of spine CT image bilateral filtering denoising method based on feature selecting, base
This process is: the point cloud data of image is subjected to grid first, then counts the point cloud amount in each grid using grid,
For deviateing farther away of backbone interest region cloud, this kind of noise spot is removed by calculating the quantity of each grid midpoint cloud
Cloud.Secondly, then needing successively to be calculated according to the point cloud of different range for confusing some noise spots in backbone interest region
The bilateral filtering factor of characteristic point cloud and non-characteristic point cloud, then this noise like is removed.
Specific implementation step may is that
1) CT image volumetric data is established
By all spinal column sections (such as Fig. 1's opens from first to m) of scanning, it is sequentially overlapped according to backbone volume morphing,
Form image volumetric data.
2) CT image pane networking
The premise of spine CT image pane networking is all to win dispersion point clouds all in image in grid over by any means.The side of grid
Method is: as shown in Figure 1, initially setting up a three-dimensional system of coordinate, spine CT image volumetric data being all read under the coordinate system, is obtained
The three-dimensional coordinate of each point is taken, while obtaining data maximum in all directions, min coordinates xmax、ymax、zmax, xmin、ymin、
zmin；Then according to the maximum of all directions, min coordinates, a big cuboid is established, which all wins all data over by any means
Inside；Then further according to the density degree of data point, which is further subdivided into several small grids, forms grid, often
Each side side length of a grid can customize.Preferably, big cuboid, small grid are preferably both cube, it is in this way meter in next step
It is convenient to calculate.
3) grid quantity is calculated
If min coordinates of the point cloud data in tri- direction X, Y, Z are respectively xmin、ymin、zmin, maximum coordinates are respectively xmax、
ymax、zmax, each grid side length is d, then quantity A, B, the C of grid on tri- directions X, Y, Z are shown in formula (1):
4) point cloud quantity and the position in each grid are obtained
4.1) coordinate of all data points in volumetric data set is obtained first.
4.2) it then calculates the number of all data points in same grid: appointing some data point M taken in data fields, with
The point is foundation, the adjoint point K on its periphery is found, if where M and K is both less than M point to it at a distance from three directions of coordinate system
The obverse distance of grid, then otherwise decision-point K and M is determined as in other grids in same grid；M and K is in which side
To distance be greater than M point to the distance of direction grid face, decide that K point is located at M point in this direction in adjacent grid；With
This analogizes, then inquires neighbor point outward centered on other points.Of data point in each grid can thus be calculated
Number.
4.3) grid positions where data point are then calculated:
If coordinate of the M point in three directions is Mx、My、Mz, then the grid hash function where this point is shown in formula (2):
Wherein, grid saves the call number, just in the call number in 3 directions of x, y, z axis where O, P, Q respectively represent M point
It is the grid positions where M point, other points in same grid are also in the grid of the call number；Similarly, and so on, other
The call number of grid is also so established.
5) CT image region of interest and non-region of interest are divided
It is handled first by the volume data gridization of early period, data is compared and concentrate the region that can highlight spinal shape big
Cause is determined as backbone interest region, and by the regional planning primarily determined in a cuboid bounding box into grid；
Edge detection is carried out followed by the Canny edge detection method of adaptive threshold, according to the ash of image itself
Threshold value of the angle value feature to calculate Canny operator just (this is a kind of disclosed method), finally determines backbone interest region
Boundary, the region unified definition except interest zone boundary are non-region of interest.
6) it denoises
6.1) for non-region of interest, directly these point cloud datas are all removed；
6.2) for region of interest, the bilateral of characteristic point cloud and non-characteristic point cloud need to be successively calculated according to the point cloud of different range
Filtering factor, the noise spot cloud for recycling the removal of bilateral filtering Denoising Algorithm to mix with interest region, specific method is:
6.2.1) judging characteristic point and non-characteristic point cloud
Judge that the cloud genera in characteristic point cloud is also non-characteristic point cloud using neighborhood: calculating separately and put cloud in grid in the domain i
The average Euclidean distance of dataIt is averaged Euclidean distance with the point cloud in this cloud k neighborhoodIt obtainsIf
A fixed threshold value D0, which determines according to the Canny edge detection method of above-mentioned adaptive threshold, works as Di> D0When, illustrate i
Point cloud is characterized a cloud, and the point cloud in neighborhood is only used when calculating bilateral filtering factor-alpha；Work as Di< D0When, illustrate i point cloud
For non-characteristic point cloud, the point cloud in whole grids should be used when calculating bilateral filtering factor-alpha.Since experimental data source is different,
Different threshold values, threshold value D can be arranged according to different spine CT characteristics of image in we0It is usually set to 300-1000.
6.2.2 the bilateral filtering factor) is calculated
Bilateral filtering is because of subformula are as follows:
Its formula ofWeight function is filtered for fairing,To keep feature to weigh letter
Number, σcPoint p to neighborhood point distance to its impact factor, value is the radius length of the point to neighborhood point, σsIt is arrived for point p
To the impact factor of p point, value is the standard deviation of neighborhood point for projection of the distance vector of neighborhood point on its normal vector.
<p-pj, two random vector p-p of n > expressionjAnd n, p-pjIndicate data point p to closing on data point pjDistance to
Amount, n indicate data point p to closing on data point pjDistance vector normal vector.
||<p-pj, n > | | indicate data point p to closing on data point pjDistance vector normal direction projection.
||p-pj| | indicate vector p-pjMould it is long.
6.2.3 it) denoises
The point cloud data after denoising is obtained, wherein p according to bilateral filtering Denoising Algorithm p'=p- α n for characteristic point cloud
For raw data points, p' is the data point after denoising, and n is data point p to closing on data point pjDistance vector normal vector, α
For the bilateral filtering factor.
It summarizes to obtain by many experiments: calculating the bilateral filtering factor by using the point cloud of different range, reuse
Bilateral filtering algorithm denoising based on feature selecting can be avoided and the phenomenon that excessive fairing occurs.
Since backbone bone image data feature is neat, interest region integrated distribution is suitble to select bilateral filtering denoising side
Method carries out image preprocessing, and when being denoised, point cloud is carried out grid first, farther away for deviateing backbone interest region
Point cloud, this kind of noise spot cloud is removed by calculating the quantity of each grid midpoint cloud, for being mixed in backbone interest region
Noise spot together successively calculates the bilateral filtering factor of characteristic point cloud and non-characteristic point cloud according to the point cloud of different range, then
This noise like is removed.And backbone point cloud data is spatially horizontal, vertical, perpendicular three directions carry out grid processing, with
Facilitate the point cloud quantity calculated in each grid.
Foregoing description clearly illustrates technical solution of the present invention, process and advantage, and those skilled in the art is obvious
Understand, the present invention is not restricted because of above-described embodiment, and the embodiment and specification of foregoing description are technology of the invention
Scheme and principle do not represent whole, and under the premise of not abandoning spirit of that invention and content, the present invention carries out respective algorithms
It improves, all within the scope of protection of present invention, realizes that experimental result of the invention, the present invention are protected in the form of distinctive
The range of shield is limited by appended claims and equivalency.
Claims (6)
1. a kind of spine CT image bilateral filtering denoising method based on feature selecting, it is characterised in that:
The point cloud data of image is subjected to grid first, then counts the point cloud amount in each grid, it is non-for image emerging
The point cloud in interesting area is directly removed；
Secondly, successively calculating the bilateral filtering factor of characteristic point cloud and non-characteristic point cloud, then to non-characteristic point in region of interest
Cloud is removed.
2. the spine CT image bilateral filtering denoising method according to claim 1 based on feature selecting, which is characterized in that
Specific implementation step is:
1) CT image volumetric data is established
It by all spine CT image slices of scanning, is sequentially overlapped according to spine form, forms image volumetric data；
2) CT image pane networking
A cuboid is established, including all image datas are all won over by any means, if then cuboid is divided into dry lattice, formation grid；
3) grid quantity is calculated
Quantity of the computation grid on tri- directions cuboid X, Y, Z；
4) position where the amount of point cloud data in each grid, and confirmation grid is calculated；
5) CT image region of interest and non-region of interest are divided
Data, which are compared, concentrates the region that can highlight spinal shape tentatively to delimit as backbone region of interest, then passes through adaptive thresholding
The Canny edge detection method of value carries out edge detection, and the threshold value of Canny operator is calculated according to the gray value of image itself,
Finally determine the boundary in backbone interest region；Region except interest zone boundary is uniformly set to non-region of interest；
6) it denoises
6.1) for non-region of interest, directly these point cloud datas are all removed；
6.2) for region of interest, according to the point cloud of different range successively calculate the bilateral filtering of characteristic point cloud and non-characteristic point cloud because
Son, the noise spot cloud for recycling the removal of bilateral filtering Denoising Algorithm to mix with region of interest.
3. the spine CT image bilateral filtering denoising method according to claim 1 or 2 based on feature selecting, feature exist
In CT image pane networking comprises the concrete steps that:
A three-dimensional system of coordinate is initially set up, CT image volumetric data is all read under coordinate system, obtains the three of each data point
Coordinate is tieed up, while obtaining the maximum in data set in all directions, minimum value xmax、ymax、zmax, xmin、ymin、zmin；
Then according to the maximum of all directions, minimum value, cuboid is established, the cuboid all wins all data points over by any means
It is interior；
Then further according to the density degree of data point, the cuboid is divided into several grids, the side length of each grid is certainly
Definition.
4. the spine CT image bilateral filtering denoising method according to claim 2 or 3 based on feature selecting, feature exist
In the method for calculating grid quantity is:
If each grid side length is d, then grid quantity A, B, C upward in cuboid X, Y, Z tripartite are as follows:
xmax、ymax、zmax, xmin、ymin、zminMaximum, minimum value for image data in each tri- directions X, Y, Z.
5. the spine CT image bilateral filtering denoising method according to claim 2 based on feature selecting, which is characterized in that
The method for calculating the amount of point cloud data in each grid and the position where confirmation grid is:
4.1) coordinate of all data points in volumetric data set is obtained first；
4.2) it then calculates the number of all data points in same grid: appointing some data point M taken in data fields, with the point
For foundation, the neighbor point K on its periphery is found, if M and K is both less than M point to grid pair where it at a distance from tri- direction X, Y, Z
The distance in face is answered, then otherwise decision-point K and M is determined as in other grids in same grid；And so on, it calculates every
The number of data point in one grid；
4.3) grid positions where data point are then calculated:
If coordinate of the M point in three directions is Mx、My、Mz, then the grid hash function where this point is shown in formula (2):
Wherein, grid saves the call number, is exactly M in the call number in 3 directions of X, Y, Z axis where O, P, Q respectively represent M point
Grid positions where point, with other points in grid also in the grid of the call number, and so on, the index of other grids
It number also so establishes, the position of grid is determined according to call number.
6. the spine CT image bilateral filtering denoising method according to claim 2 based on feature selecting, which is characterized in that
For region of interest, the specific method of denoising is:
6.2.1 characteristic point and non-characteristic point cloud) are first determined whether
Judge that the cloud genera in characteristic point cloud is also non-characteristic point cloud using neighborhood: calculating separately in grid point cloud data in the domain i
Average Euclidean distanceIt is averaged Euclidean distance with the point cloud in this cloud k neighborhoodIt obtainsSetting one
Threshold value D0, the threshold value is according to the determination of the Canny edge detection method of adaptive threshold；
Work as Di> D0When, illustrate that i point cloud is characterized a cloud, the point cloud in neighborhood is only used when calculating bilateral filtering factor-alpha；
Work as Di< D0When, illustrate i point Yun Weifei characteristic point cloud, uses the point cloud in whole grids when calculating bilateral filtering factor-alpha；
6.2.2 the bilateral filtering factor) is calculated
Bilateral filtering is because of subformula are as follows:
Its formula ofWeight function is filtered for fairing,To keep feature weight function, σc
Point p to neighborhood point distance to its impact factor, value is the radius length of the point to neighborhood point, σsFor point p to neighborhood
To the impact factor of p point, value is the standard deviation of neighborhood point for projection of the distance vector of point on its normal vector；
< p-pj, two random vector p-p of n > expressionjAnd n, p-pjIndicate data point p to closing on data point pjDistance vector, n
Indicate data point p to closing on data point pjDistance vector normal vector；
||<p-pj, n > | | indicate data point p to closing on data point pjDistance vector normal direction projection；
||p-pj| | indicate vector p-pjMould it is long；
6.2.3 it) denoises
According to bilateral filtering Denoising Algorithm p'=p- α n, the point cloud data after denoising is obtained, wherein p is raw data points, and p' is
Data point after denoising, n are normal vector of the data point p to the distance vector for closing on data point pj, and α is the bilateral filtering factor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811386922.5A CN109741265A (en) | 2018-11-20 | 2018-11-20 | A kind of spine CT image bilateral filtering denoising method based on feature selecting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811386922.5A CN109741265A (en) | 2018-11-20 | 2018-11-20 | A kind of spine CT image bilateral filtering denoising method based on feature selecting |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109741265A true CN109741265A (en) | 2019-05-10 |
Family
ID=66356995
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811386922.5A Withdrawn CN109741265A (en) | 2018-11-20 | 2018-11-20 | A kind of spine CT image bilateral filtering denoising method based on feature selecting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109741265A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110349094A (en) * | 2019-06-12 | 2019-10-18 | 西安工程大学 | It is peeled off based on statistics and the 3D point cloud denoising method of adaptive bilateral mixed filtering |
| CN110633385A (en) * | 2019-09-24 | 2019-12-31 | 天津医科大学 | Retrieval and compression method of medical image |
| WO2022141373A1 (en) * | 2020-12-31 | 2022-07-07 | 西安大医集团股份有限公司 | Medical image processing method and apparatus, and computer device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846272A (en) * | 2017-01-18 | 2017-06-13 | 西安工程大学 | A kind of denoising compressing method of point cloud model |
| CN107845079A (en) * | 2017-11-15 | 2018-03-27 | 浙江工业大学之江学院 | 3D shearlet medicine CT video denoising methods based on compact schemes |
| CN108537750A (en) * | 2018-03-29 | 2018-09-14 | 哈尔滨理工大学 | A kind of quick spine CT 3-dimensional reconstruction method |
| US20180322618A1 (en) * | 2017-05-02 | 2018-11-08 | Color Enhanced Detection, Llc | Methods for color enhanced detection of bone density from ct images and methods for opportunistic screening using same |
-
2018
- 2018-11-20 CN CN201811386922.5A patent/CN109741265A/en not_active Withdrawn
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846272A (en) * | 2017-01-18 | 2017-06-13 | 西安工程大学 | A kind of denoising compressing method of point cloud model |
| US20180322618A1 (en) * | 2017-05-02 | 2018-11-08 | Color Enhanced Detection, Llc | Methods for color enhanced detection of bone density from ct images and methods for opportunistic screening using same |
| CN107845079A (en) * | 2017-11-15 | 2018-03-27 | 浙江工业大学之江学院 | 3D shearlet medicine CT video denoising methods based on compact schemes |
| CN108537750A (en) * | 2018-03-29 | 2018-09-14 | 哈尔滨理工大学 | A kind of quick spine CT 3-dimensional reconstruction method |
Non-Patent Citations (1)
| Title |
|---|
| XIA LIU, HUAN LIU, BO WANG, MIAOMIAO WANG: "Vertebrae CT Images 3D Reconstruction with Improved Marching Cubes Algorithm", 《JOURNAL OF HARBIN INSTITUTE OF TECHNOLOGY(NEW SERIES)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110349094A (en) * | 2019-06-12 | 2019-10-18 | 西安工程大学 | It is peeled off based on statistics and the 3D point cloud denoising method of adaptive bilateral mixed filtering |
| CN110633385A (en) * | 2019-09-24 | 2019-12-31 | 天津医科大学 | Retrieval and compression method of medical image |
| CN110633385B (en) * | 2019-09-24 | 2023-05-12 | 天津医科大学 | Medical image retrieval and compression method |
| WO2022141373A1 (en) * | 2020-12-31 | 2022-07-07 | 西安大医集团股份有限公司 | Medical image processing method and apparatus, and computer device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993783B (en) | Roof and side surface optimization reconstruction method for complex three-dimensional building point cloud | |
| JP6660313B2 (en) | Detection of nuclear edges using image analysis | |
| CN104809723B (en) | The three-dimensional CT image for liver automatic division method of algorithm is cut based on super voxel and figure | |
| CN105761252B (en) | The method and device of image segmentation | |
| CN107945179A (en) | A kind of good pernicious detection method of Lung neoplasm of the convolutional neural networks of feature based fusion | |
| CN109741265A (en) | A kind of spine CT image bilateral filtering denoising method based on feature selecting | |
| CN104143190B (en) | The dividing method and system organized in CT images | |
| CN105678806B (en) | A kind of live pig action trail automatic tracking method differentiated based on Fisher | |
| CN105828691B (en) | Image processing apparatus, image processing method | |
| AU2005244641A1 (en) | Nodule detection | |
| CN110544300B (en) | Method for automatically generating three-dimensional model based on two-dimensional hand-drawn image characteristics | |
| CN109300113A (en) | A kind of Lung neoplasm assisted detection system and method based on improvement Convex Hull Method | |
| CN107506795A (en) | A kind of local gray level histogram feature towards images match describes sub- method for building up and image matching method | |
| CN107463933A (en) | A kind of image object detection method | |
| CN113313107B (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
| CN110443793A (en) | A kind of asphalt mixture gap distributing homogeneity evaluation method | |
| CN114926407A (en) | Steel surface defect detection system based on deep learning | |
| CN109636810A (en) | Lung neoplasm dividing method, system and the region growing method of CT images, segmentation terminate judgment method, cutting adhesion region method | |
| CN104182965B (en) | The method of chest muscle is split in a kind of galactophore image | |
| CN103164843B (en) | A kind of medical image colorize method | |
| CN114299490A (en) | Tumor microenvironment heterogeneity evaluation method | |
| JP2014106713A (en) | Program, method, and information processor | |
| CN110930330B (en) | Image segmentation and region growth based salt and pepper noise reduction algorithm | |
| CN112801970A (en) | Breast ultrasound image tumor segmentation method | |
| CN115631194B (en) | Method, device, equipment and medium for identifying and detecting intracranial aneurysm |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190510 |