CN105303533B - A kind of ultrasound image filtering method - Google Patents
A kind of ultrasound image filtering method Download PDFInfo
- Publication number
- CN105303533B CN105303533B CN201510736092.4A CN201510736092A CN105303533B CN 105303533 B CN105303533 B CN 105303533B CN 201510736092 A CN201510736092 A CN 201510736092A CN 105303533 B CN105303533 B CN 105303533B
- Authority
- CN
- China
- Prior art keywords
- image
- ultrasound image
- target area
- point
- filtered
- 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
Abstract
The invention discloses a kind of ultrasound image filtering methods, include the following steps：(1) ultrasound image to be filtered is split using Otsu algorithm, obtains bianry image；(2) to bianry image, erosion operation is first carried out, carries out dilation operation afterwards；(3) it carries out connected region tracking and several independent communication regions is divided by neighborhood search for all bright spots in bianry image；(4) according to preset threshold value, retain the independent communication region more than threshold value as target area；(5) using the target area of ultrasound image to be filtered as filtered ultrasound image, other regions are set to background.The method of the present invention can effectively be filtered ultrasound image, cancelling noise, keep the feature of initial data, and any deformation will not occur for target, meanwhile, this method calculation amount is moderate, and algorithm difficulty is small, realizes and is easy.
Description
Technical field
The invention belongs to ultrasonic imaging fields, more particularly, to a kind of ultrasound image filtering method.
Background technique
Ultrasonic imaging is the important content of medical imaging field, its advantage is that radiationless, equipment and use cost are low, ultrasound
Imaging provides detailed information more abundant, intuitively feels to doctor and the person of being diagnosed, reduces the professional technique of doctor
It is required that and improving the reliability of judgement.Meanwhile ultrasound image has serious noise, noise not only influences visual effect,
The judgement of doctor is interfered to a certain extent, and therefore, the filtering of ultrasound image is always that a hot spot in medical ultrasound image is asked
Topic.Current filtering method is based on Mean Shift principle, by being averaged in a certain range to original image, or
Person and some smooth templates do convolution, achieve the purpose that smooth.In general, these conventional methods the problem is that：Hold
The details of image easy to be lost, reason are that its filtering is by smoothly realizing；Second is to easily cause the whole of data model to become
Shape；Last poor robustness, it is closely related with used parameter.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of ultrasound image filtering method,
Purpose is thus to solve existing ultrasound filtering by the way that Threshold segmentation and the cooperation of connected domain track algorithm are filtered out speckle noise
Method loses the technical issues of image detail, overall deformation, poor robustness.
To achieve the above object, according to one aspect of the present invention, ultrasound image filtering method, including following step are provided
Suddenly：
(1) ultrasound image to be filtered is split using Otsu algorithm, obtains bianry image；
(2) for the bianry image obtained in step (1), erosion operation is first carried out, dilation operation is carried out afterwards, obtains target
The bianry image that region and noise separate；
(3) bianry image that the target area obtained in step (2) and noise separate is subjected to connected region tracking, specifically
Steps are as follows：
Target area tracking：For all bright spots in bianry image, several are divided into solely by neighborhood search
Vertical connected region；
(4) independence more than threshold value is retained according to preset threshold value for the independent communication region obtained in step (3)
Connected region is as target area；
(5) using the target area of ultrasound image to be filtered as filtered ultrasound image, other regions are set to background.
Preferably, the ultrasound image filtering method, further includes step：(6) step (5) is obtained filtered super
The edge of acoustic image carries out smoothing operation.
Preferably, the ultrasound image filtering method, step (6) smoothing operation use gaussian filtering.
Preferably, the ultrasound image filtering method, specific step is as follows for step (3) described neighborhood search：
(3-1) arbitrarily chooses a bright spot as seed point, is marked；
(3-2) searches for its neighborhood, if bright spot, then these target points is marked；
(3-3) constantly searches for labeled neighborhood of a point, until the point in all labeled neighborhoods of a point has been marked
Note is dim spot, using the bright spot of these labels as an independent communication region；
(3-4) repeats step (3-1)-(3-3) in remaining bright spot, until all bright spots are labeled, i.e., it is all
Bright spot be divided in some independent communication region.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect：
1, compared with noise, artifact, target in ultrasound image it is opposite be it is brighter, then, by Threshold segmentation,
Substantially noise and artifact and target can be separated.
2, "ON" operation is carried out to bianry image, as far as possible by some small targets or close to the noise and target of target
Area disconnects.
3, by the search to target vertex neighborhood, by it is all be connected to target point o'clock gather as one, thus by mesh
Mark, noise and artifact separate, it is believed that the biggish several targets of volume be for actual ultrasonic imaging target, it is other to be considered as making an uproar
Sound and artifact.
4, in original image, the gray scale of non-imaged target is assigned to 0, the bright spot of imageable target assigns its original gray scale
Value, the dim spot of imageable target then take the mean value of non-zero point in its neighborhood, obtain filtered image.Therefore, this filtering method is not
There is the gray scale for changing target in image, but as far as possible distinguishes background, artifact and noise and target, therefore filtered image
The improper feature that can keep initial data, and the deformation of target will not occur.
To sum up, the method for the present invention can effectively be filtered ultrasound image, cancelling noise, keep the spy of initial data
Sign, and any deformation will not occur for target, meanwhile, this method calculation amount is moderate, and algorithm difficulty is small, realizes and is easy.
Detailed description of the invention
Fig. 1 present invention filters flow chart；
Three-dimensional medical ultrasonic image volume rendered images before Fig. 2 is filtered；
Three cross-section images at three-D ultrasound data center before Fig. 3 is filtered；
Three-dimensional medical ultrasonic image volume rendered images after Fig. 4 filtering；
Three cross-section images at three-D ultrasound data center after Fig. 5 filtering.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
Ultrasonic imaging has very extensive clinical value, and a large amount of noise not only affects vision in image
Effect, and judgement of the doctor to imageable target may be will affect.Therefore the filtering of ultrasound image is always the research heat of related fields
Point, the method for this method are no longer based on Mean Shift principle, but by the target and background, noise and puppet in ultrasound image
Shadow separates, and non-targets of interest is then considered as background, is finally handled using traditional gaussian filtering image.
Ultrasound image filtering method provided by the invention, includes the following steps：
(1) ultrasound image to be filtered is split using Otsu algorithm, obtains bianry image；In general, target
Presentation is highlighted, and background is relatively darker, therefore bright point is considered the point on target image.It is preferred that using following steps, carry out
Segmentation：
Raw ultrasound image approximate is divided into two classes, target and background：If the gray level of image is L, gray scale is the pixel of i
Points be set as Ni.Gray value is that the probability that i occurs is Pi=Ni/N.Then, a thresholding t is found, is divided the image into so that two
Gray variance between class target is maximum.The point that gray value is greater than threshold value in image is assigned a value of 255, remaining is 0.
(2) for the bianry image obtained in step (1), erosion operation is first carried out, carries out dilation operation afterwards, it will be different
Target separates, and obtains the bianry image of target area and noise separation；
Small target is separated from big target using corrosion dilation operation, wherein erosion operation is a kind of elimination boundary point
Method, do with operation with structural element template and bianry image, the Er value Tu Xiang Of-thin mono- made is enclosed.Dilation operation and corrosion are transported
It is similar, it is a difference in that and structural element template and image is done into OR operation, expand the boundary of object to outside.Due to this
The purpose of invention is to be filtered to ultrasound image, therefore take the strategy for first corroding and expanding afterwards, in mathematical morphology also referred to as
To be opening operation.
(3) bianry image that the target area obtained in step (2) and noise separate is subjected to connected region tracking, specifically
Steps are as follows：
Target area tracking：For all bright spots in bianry image, several are divided into solely by neighborhood search
Vertical connected region.
To realize above process, following steps can be taken：
(3-1) arbitrarily chooses a bright spot as seed point, is marked；
(3-2) searches for its neighborhood, if bright spot, then these target points is marked；
(3-3) constantly searches for labeled neighborhood of a point, until the point in all labeled neighborhoods of a point has been marked
Note is dim spot, using the bright spot of these labels as an independent communication region；
(3-4) repeats step (3-1)-(3-3) in remaining bright spot, until all bright spots are labeled, i.e., it is all
Bright spot be divided in some independent communication region.
(4) independence more than threshold value is retained according to preset threshold value for the independent communication region obtained in step (3)
Connected region is as target area；
Threshold value setting can be arranged threshold value according to size or set according to area ranking according to the property of image itself
Set threshold value.
(5) using the target area of ultrasound image to be filtered as filtered ultrasound image, other regions are set to background.
(6) edge for the filtered ultrasound image for obtaining step (5) carries out smoothing operation, it is preferred to use Gauss filter
Wave.The template that a certain size is generated using Gaussian function is carried out convolution by template and above-mentioned image, obtains the knot of gaussian filtering
Fruit image.
The following are embodiments：
A kind of ultrasound image filtering method is applied to three-dimensional ultrasound pattern and filters：
(1) Threshold segmentation of three-dimensional ultrasound pattern, if the size of input 3-D image is nwidth*nheight*
Ndepth, the gray level of image are L=256, and gray scale is that the points of the pixel of i are set as Ni, nwidth*nheight*ndepth=
N0+N1+...+NL-1.Gray value is that the probability that i occurs is Pi=Ni/N.Then, a thresholding t is found, 3-D image is divided into
Bright target c1With dark background c2, then the relationship of inter-class variance σ and t is σ=a1*a2(u1-u2) ^2, in formula, a1、a2For class c1、c2Always
Area ratio, a1=sum (Pi)i>T, a2=1-a1；u1、u2The mean value of respectively two classes, u1=sum (i*Pi)/a1i<T, u2
=sum (i*P2)/a2,i>=t selects an optimum thresholding t, and the inter-class variance made is maximum.It is former after threshold value is calculated
Gray scale is assigned to 255 greater than threshold value in beginning image, other to be assigned to 0, generation bianry image.
(2) corrosion and expansive working are carried out to bianry image.In the method, the operation of corrosion is：With two-value volume data
In each voxel centered on 3 × 3 × 3 windows in, judge whether there is gray value be 0 point, if so, then current point is assigned
Value is 0；The operation of expansion is：In 3 × 3 × 3 windows centered on each voxel in two-value volume data, ash is judged whether there is
The point that angle value is 255, if so, current point is then assigned a value of 255.Corrosion and expansive working are constantly carried out, the three of binaryzation
It ties up in image, it is the region of 255 multiple connections that being formed, which has a certain number of gray values,.
(3) in bianry image, arbitrarily select the target point that one gray value is 255, (general recommendations selection approaches three-dimensional
The point of picture centre), it is added to a set, search is secondly whether having gray value in 16 neighborhoods is 255 point, if there is then
It is added in point set, meanwhile, the point in set is marked, if 26 neighborhoods of some point are searched in set
Rope mistake is then labeled as 1, is otherwise 0；In next step be exactly arbitrarily selected in point set one label for point, search for secondly ten
Whether there is bright spot in six neighborhoods, and bright spot is added in point set, is then marked as 1；Then, constantly from point set
Selected marker value is 0 point, and searches for its neighborhood constantly to repeat the above process, until point all in point set is labeled
Until 1.
(4) target collection that step (3) are equivalent to obtain a connection is completed, and the image of practical binaryzation, Ke Nengyou
The target collection being much connected to then arbitrarily selects a point (not to be labeled in the point for being 255 from remaining gray value
Point), the process of step (3) is repeated, corresponding connectivity points set is obtained.(3)-(4) constantly are executed, until owning in bianry image
Point it is labeled until.
(5) in the point set that step (4) obtains, if the volume of point set is less than certain quantity, by these points
Gray scale is assigned to 0.
(6) bianry image of original three-dimensional image and step (5), gray scale of the voxel outside bianry image targets of interest are compared
Value is 0, then the gray value of respective pixel in original image is assigned to 0；Gray value of the voxel in bianry image targets of interest be
255, then the gray value of corresponding position in its original image is assigned, gray value of the voxel in bianry image targets of interest is 0, then
Assign the mean value of non-zero point in its 26 neighborhood of original image.
(7) step (6) image is filtered using Gaussian template, specially：First with dimensional Gaussian template be 1,2,
3,2,1；2,5,6,5,2；3,6,8,6,3；2,5,6,5,2；1,2,3,2,1 }, two dimension is carried out to the 5*5 image in 5*5*5 window
Filtering, generates the data of 5*1；Then, { 3,6,8,6,3 } are recycled to carry out convolution again；Obtain the filter result of window center point.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (2)
1. a kind of three-dimensional ultrasound pattern filtering method, which is characterized in that follow the steps below：
(1) ultrasound image to be filtered is split using Otsu algorithm, obtains bianry image；
(2) for the bianry image obtained in step (1), erosion operation is first carried out, dilation operation is directly carried out afterwards, obtains target
The bianry image that region and noise separate；
(3) bianry image that the target area obtained in step (2) and noise separate is subjected to connected region tracking, specific steps
It is as follows：
Target area tracking：For all bright spots in bianry image, several are divided by neighborhood search and is independently connected
Logical region；Specific step is as follows for the neighborhood search：
(3-1) arbitrarily chooses a bright spot as seed point, is marked；
(3-2) searches for its neighborhood, if bright spot, then these bright spots is marked；
(3-3) constantly searches for labeled bright neighborhood of a point, until the point in all labeled bright neighborhoods of a point has been marked
Note is dim spot, using the bright spot of these labels as an independent communication region；
(3-4) repeats step (3-1)-(3-3) in remaining bright spot, until all bright spots are labeled, i.e., all is bright
Point is divided in some independent communication region；
(4) independent communication more than threshold value is retained according to preset threshold value for the independent communication region obtained in step (3)
Region is as target area；
Threshold value setting is arranged threshold value according to size or threshold value is arranged according to area ranking；By the maximum target view of volume
It is other to be considered as noise and artifact for actual ultrasonic imaging target；
(5) using the target area of ultrasound image to be filtered as filtered ultrasound image, other regions are set to background；Specifically
For：The target area for comparing initial three-dimensional ultrasound image and step (4), if gray value of the voxel outside bianry image target area
It is 0, then the gray value of respective pixel in initial three-dimensional ultrasound image is assigned to 0；If voxel is in bianry image target area
Gray value is 255, then assigns the gray value of its corresponding position in initial three-dimensional ultrasound image to the voxel；If voxel is in two-value
Gray value in image target area is 0, then assigning the mean value of non-zero point in its 26 neighborhood of initial three-dimensional ultrasound image should
Voxel；
(6) edge for the filtered ultrasound image for obtaining step (5) carries out smoothing operation.
2. three-dimensional ultrasound pattern filtering method as described in claim 1, which is characterized in that step (6) described smoothing operation is adopted
Use gaussian filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510736092.4A CN105303533B (en) | 2015-11-03 | 2015-11-03 | A kind of ultrasound image filtering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510736092.4A CN105303533B (en) | 2015-11-03 | 2015-11-03 | A kind of ultrasound image filtering method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105303533A CN105303533A (en) | 2016-02-03 |
| CN105303533B true CN105303533B (en) | 2018-11-30 |
Family
ID=55200757
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510736092.4A Active CN105303533B (en) | 2015-11-03 | 2015-11-03 | A kind of ultrasound image filtering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105303533B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961532B (en) * | 2017-05-26 | 2020-11-17 | 深圳怡化电脑股份有限公司 | Method, device and equipment for processing crown word number image and storage medium |
| CN107545568B (en) * | 2017-08-07 | 2021-08-20 | 东方财富信息股份有限公司 | Processing method and system for 3D binary image |
| EP3843034A4 (en) | 2018-08-22 | 2021-08-04 | GeneMind Biosciences Company Limited | Method and device for detecting bright spots on image, and computer program product |
| US20210217186A1 (en) * | 2018-08-22 | 2021-07-15 | Genemind Biosciences Company Limited | Method and device for image registration, and computer program product |
| EP3843033A4 (en) | 2018-08-22 | 2021-11-24 | GeneMind Biosciences Company Limited | Method for constructing sequencing template based on image, and base recognition method and device |
| CN115294605B (en) * | 2022-08-05 | 2023-05-16 | 杭州电子科技大学 | Millimeter wave image strong noise airspace eliminating method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020605A (en) * | 2012-12-28 | 2013-04-03 | 北方工业大学 | Bridge identification method based on decision-making layer fusion |
| CN103914843A (en) * | 2014-04-04 | 2014-07-09 | 上海交通大学 | Image segmentation method based on watershed algorithm and morphological marker |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007017765A (en) * | 2005-07-08 | 2007-01-25 | Ricoh Co Ltd | Spray head, fixing device using same, image forming apparatus, and toner removal device |
-
2015
- 2015-11-03 CN CN201510736092.4A patent/CN105303533B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020605A (en) * | 2012-12-28 | 2013-04-03 | 北方工业大学 | Bridge identification method based on decision-making layer fusion |
| CN103914843A (en) * | 2014-04-04 | 2014-07-09 | 上海交通大学 | Image segmentation method based on watershed algorithm and morphological marker |
Non-Patent Citations (2)
| Title |
|---|
| Objective evaluation method of fabric pilling based on image analysis;Zhu Shuangwu等;《 Proceedings of 2006 China International Wool Textile Conference & IWTO Wool Forum》;20061101;第312-317页 * |
| 基于聚类算法的边缘点集连接方法;张彩仙等;《武汉大学学报（工学版）》;20151001;第48卷(第5期);第723-726页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105303533A (en) | 2016-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105303533B (en) | A kind of ultrasound image filtering method | |
| Revaud et al. | Epicflow: Edge-preserving interpolation of correspondences for optical flow | |
| Tosi et al. | Beyond local reasoning for stereo confidence estimation with deep learning | |
| CN106204555B (en) | A kind of optic disk localization method of combination Gbvs model and phase equalization | |
| US8170304B2 (en) | Modeling cerebral aneurysms in medical images | |
| CN110910405B (en) | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network | |
| CN104794721B (en) | A kind of quick optic disk localization method based on multiple dimensioned spot detection | |
| CN106157303A (en) | A kind of method based on machine vision to Surface testing | |
| CN104616308A (en) | Multiscale level set image segmenting method based on kernel fuzzy clustering | |
| CN103455984A (en) | Method and device for acquiring Kinect depth image | |
| Lo et al. | Joint trilateral filtering for depth map super-resolution | |
| CN110245600B (en) | Unmanned aerial vehicle road detection method for self-adaptive initial quick stroke width | |
| CN109598738A (en) | A kind of line-structured light center line extraction method | |
| Letscher et al. | Image segmentation using topological persistence | |
| CN106096491A (en) | The automatic identifying method of the microaneurysm in color fundus image | |
| CN104268893A (en) | Method for segmenting and denoising lung parenchyma through lateral scanning and four-corner rotary scanning | |
| CN110503637A (en) | A kind of crack on road automatic testing method based on convolutional neural networks | |
| CN101425140A (en) | Encephalic angioma image recognizing and detecting method based on framework characteristic | |
| CN104036481A (en) | Multi-focus image fusion method based on depth information extraction | |
| CN1261910C (en) | Automatic generating method for colour multi-window CT image | |
| CN106408533A (en) | Card image extraction method and card image extraction system | |
| CN105513055A (en) | Method and apparatus for segmenting tissue in CTA image | |
| CN108924434A (en) | A kind of three-dimensional high dynamic-range image synthesis method based on exposure transformation | |
| CN104504711A (en) | Vascular image processing method based on circular contour polarity | |
| CN108537802A (en) | A kind of blood vessel segmentation method and device |
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 |