CN111583286A - Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator - Google Patents
Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator Download PDFInfo
- Publication number
- CN111583286A CN111583286A CN202010276311.6A CN202010276311A CN111583286A CN 111583286 A CN111583286 A CN 111583286A CN 202010276311 A CN202010276311 A CN 202010276311A CN 111583286 A CN111583286 A CN 111583286A
- Authority
- CN
- China
- Prior art keywords
- edge
- image
- xdog
- flow
- operator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 86
- 210000001015 abdomen Anatomy 0.000 title claims abstract description 20
- 238000002595 magnetic resonance imaging Methods 0.000 title description 30
- 230000003187 abdominal effect Effects 0.000 claims abstract description 20
- 210000000056 organ Anatomy 0.000 claims abstract description 8
- 230000002708 enhancing effect Effects 0.000 claims abstract description 5
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000009499 grossing Methods 0.000 claims abstract description 3
- 238000001914 filtration Methods 0.000 claims description 34
- 238000000034 method Methods 0.000 claims description 32
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 abstract description 8
- 230000007246 mechanism Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 16
- 230000008569 process Effects 0.000 description 11
- 238000012545 processing Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000002146 bilateral effect Effects 0.000 description 7
- 238000005481 NMR spectroscopy Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000009877 rendering Methods 0.000 description 4
- 238000002591 computed tomography Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 210000001519 tissue Anatomy 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 239000002356 single layer Substances 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000002600 positron emission tomography Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G06T5/70—
-
- G06T5/77—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
Abstract
The invention discloses an abdomen MRI image contour extraction method based on a Flow-XDoG operator, which comprises the following steps: preprocessing an abdominal MRI image, reducing interference caused by noise and a deviation field, and enhancing and smoothing the image; extracting the edges of tissues or organs, respectively extracting the edges of the preprocessed MRI images by using two Flow-XDoG operators with different parameters, and obtaining a preliminary result through edge connection; removing false edge interference; and removing the false edge, thinning the edge and finishing the extraction of the abdominal MRI image contour. Compared with the classic edge extraction operator and the edge extraction method of non-photorealistic drawing, the abdomen MRI image contour extraction method based on the Flow-XDoG operator has the advantages that the extracted edge is clearer, more complete and more accurate, the network training is not depended on, the problem of the data volume of a training set does not need to be solved, and the abdomen MRI image contour extraction method based on the Flow-XDoG operator is very suitable for contour extraction of abdomen MRI images or other images with complex graphic mechanisms and texture structures.
Description
Technical Field
The invention relates to the field of medical image processing, in particular to an abdominal nuclear Magnetic Resonance (MRI) image contour extraction method based on a Flow-based extended Difference-of-Gaussians (Flow-based extended Difference-of-Gaussians) operator.
Background
With the advent of medical Imaging equipment such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET), the revolutionary development of modern computer science and technology has brought about a dramatic progress in medical image processing and analysis technology. Today, medical image processing and analysis techniques are widely used to assist physicians in performing accurate and clear clinical diagnosis. At present, the technology for analyzing and processing MRI images or CT images of brain, lung, blood vessels, bones and other parts is mature, but the research on abdominal MRI images is less, so the technology for processing and analyzing abdominal MRI images has very important practical significance.
Edge extraction is an important component of computer graphics processing techniques. The clear and complete contour is extracted, and the method can provide help for subsequent image processing and analyzing operations such as image segmentation and target recognition. The abdominal nuclear magnetic resonance image has the characteristic of complex graphic structure and texture structure, and the common edge extraction method cannot meet the requirement of practical application. The classical image edge extraction algorithm mainly utilizes detection operators such as a Sobel operator, a Laplace operator and a Canny operator to extract edge features. The detection operators have the advantages of small calculation amount, high calculation speed and strong real-time performance, but are sensitive to illumination and noise, and have poor detection effect on image edges under the conditions of poor illumination conditions and noise pollution. The contour extraction method based on the neural network and the deep learning is also a popular research direction in recent years, but because a large number of training samples are needed, and a large amount of manpower is needed for calibrating the medical images, how to solve the problem of the number of training sets becomes a big problem. Therefore, the method for extracting the edge without depending on the neural network and the deep learning and capable of extracting the complete and clear outline has very important significance.
In many practical applications, people pay attention to only a part of an image, and other details are ignored, such as in the fields of medicine, video communication, archaeology and the like. Non-photorealistic rendering (NPR) is a method of representing a scene by an artistic approach without pursuing realism, which generally does not show real and specific details of a target, but shows an image in a structured and simplified style focused by an observer, which is easier to show non-real details, emphasizes certain information capable of representing image characteristics, and ignores secondary information. The edge extraction of the image is a process of extracting the outline and the shape of the image, and therefore can be regarded as extraction of image skeletons or extraction of image structural features. Therefore, non-photorealistic rendering (NPR) is of great significance for realizing contour extraction of images. Commonly used non-photorealistic rendering (NPR) methods include Flow-Based Difference of gaussian (FDoG) and Flow-Based Extended Difference of gaussian (Flow-XDoG). However, many false edges appear in the extraction result of the edge extraction method of NPR, so an improved contour extraction method based on a non-photorealistic rendering technology is urgently needed.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides an abdominal nuclear Magnetic Resonance (MRI) image contour extraction method based on a Flow-based extended Difference-of-Gaussians (Flow-based extended Difference) operator.
The technical purpose of the invention is realized by the following technical scheme.
A Flow-XDoG operator-based abdominal MRI image contour extraction method comprises the following steps:
in the step 1, bilateral filtering is carried out on the image, so that the interference of noise and a deviation field to contour extraction is reduced; enhancing the texture details of the image by using fractional order differential filtering, and reserving a smooth area in the image;
in step 1, the fractional order differential template used by the fractional order differential filtering is a Tiansi operator.
in step 2, the Flow-XDoG operator is an anisotropic filter that extracts a set of smooth and abstracted edge curves from an image; the edge extraction process for the preprocessed MRI image using the Flow-XDoG operator is as follows:
(1) calculating the image gradient to obtain a gradient image;
(2) constructing Edge Tangential Flow (ETF);
(3) extracting a line profile using extended gaussian differential filtering (XDoG);
(4) dividing a threshold value;
(5) filtering a connected domain threshold;
in step 2, the Flow-XDoG operators with different parameters have different edge extraction effects, and in order to reduce adjustable parameters as much as possible, the Flow-XDoG operators with two different parameters ρ are respectively used for performing edge extraction on the preprocessed MRI image, so as to obtain an image a corresponding to a smaller ρ value and an image B corresponding to a larger ρ value; the parameter rho is used for controlling the number of lines extracted by an operator, and the larger the rho value is, the more lines are extracted;
in step 2, the edge connection means that if each edge pixel (x) of the image A is connectedi,yi) Edge points (x ') exist in the neighborhood of the corresponding position of image B'i,y′i) Then it is located at (x 'in graph A'i,y′i) Is set as an edge point.
Step 3, removing error edge interference;
in step 3, by obtaining difference information between images and connected domain threshold filtering, erroneous edges due to noise and other interference factors and erroneous edges located inside organs are removed.
in step 4, the neighborhood total variation can reflect the spectral variation degree in the neighborhood, the edge has a larger response value, and considering that each pixel has different contributions to the neighborhood total variation, the Weighted neighborhood total variation (WTV) is used for further filtering out a false edge generated in the process of edge extraction, and the edge is preliminarily refined;
in step 4, the calculation formula of the weighted neighborhood total variation is as follows:
wherein (x, y) represents the coordinates of a pixel point in the image, (x)0，y0) Indicating the coordinates of a certain pixel in the image, D indicating the point (x)0,y0) The weighted neighborhood total variation value of (x) is expressed by Ω0,y0) A neighborhood space of the center, h (x, y) denotes a gaussian function, ▽ f (x, y) denotes a gradient;
in step 4, the edge is further refined through a skeleton extraction algorithm to obtain a final edge refinement result.
Compared with the prior art, the method has the advantages that the influence of noise and image self-texture interference on the extraction result is reduced by utilizing bilateral filtering and fractional order differential filtering, the number of parameters to be adjusted is reduced by adopting two Flow-XDoG operators with given different parameters, the extracted error edge is removed through connected domain threshold analysis, image difference information and weighted total variation edge judgment, and the edge extraction quality is improved; compared with the classic edge extraction operator and the edge extraction method of non-photorealistic drawing, the abdomen MRI image contour extraction method based on the Flow-XDoG operator has the advantages that the extracted edge is clearer, more complete and more accurate, the method does not depend on network training, and the problem of the data volume of a training set does not need to be solved; the method of the invention is very suitable for contour extraction of abdominal MRI images or other images with complex graphical mechanisms and textures.
Description of the drawings:
FIG. 1 is a flow chart of the contour extraction of the present invention;
FIG. 2 is a schematic diagram of a four-direction Tiansi template decomposed according to coordinate axes;
FIG. 3 is a template of a sobel in eight orientations;
FIG. 4 is a graph showing the effect of pretreatment according to the present invention; fig. 4(a) is an original image, fig. 4(b) is a bilateral filtering effect diagram, and fig. 4(c) is a fractional differential filtering effect diagram;
FIG. 5 is a graph illustrating the effect of connected domain threshold filtering according to the present invention; fig. 5(a) is a Flow-XDoG operator extraction effect graph with ρ ═ 0.99, fig. 5(b) is a connected domain threshold value filtering effect graph with ρ ═ 0.99, fig. 5(c) is a Flow-XDoG operator extraction effect graph with ρ ═ 1.0, and fig. 5(d) is a connected domain threshold value filtering effect graph with ρ ═ 1.0;
FIG. 6 is a diagram illustrating the effect of edge connection according to the present invention;
FIG. 7 is a diagram illustrating the effect of the present invention after obtaining an image difference and removing false edge interference;
FIG. 8 is a Zhang-Suen refinement algorithm pseudo-code; wherein, fig. 8(a) is a schematic diagram of 8 neighborhoods, and fig. 8(b) is pseudo code of Zhang-Suen refinement algorithm;
FIG. 9 is a graph of edge refinement results of the present invention; wherein, fig. 9(a) is a weighted total variation edge determination process, and fig. 9(b) is a Zhang-Suen refinement algorithm effect diagram;
FIG. 10 is a comparison of different contour extraction methods; wherein, fig. 10(a) is the Sobel operator edge extraction result, fig. 10(b) is the fuzzy inference-based edge extraction result, fig. 10(c) is the single Flow-XDoG operator edge extraction result, and fig. 10(d) is the edge extraction result of the method of the present invention.
Detailed Description
The method of the present invention will be described in detail with reference to specific embodiments.
In consideration of the influence of Image noise and complex Image texture on contour extraction, bilateral filtering and fractional differential filtering are adopted to preprocess an abdominal nuclear Magnetic Resonance Image (MRI), two Flow-XDoG operators with different parameters are used to extract edge lines and perform edge connection on the Image, residual noise and error edges caused by Image texture are removed by obtaining difference information between the images and analyzing a connected domain, Weighted Total Variation (WTV) and skeleton extraction are used to realize false edge removal and edge refinement, and finally the extraction of the abdominal nuclear Magnetic Resonance Image (MRI) contour is realized.
The data of the invention is derived from the public data set provided by the SPIE-AAPM-NCI PROSTATex competition, and the extraction flow chart is shown in figure 1.
The invention discloses an abdomen MRI image contour extraction method based on a Flow-XDoG operator, which comprises the following steps:
carrying out bilateral filtering on the image to reduce the interference of noise and a deviation field to contour extraction; using fractional order differential filtering to enhance the image texture details and retain the smooth area in the image, the specific process is as follows:
(1) the image is subjected to bilateral filtering, so that the interference of noise and a deviation field to contour extraction is reduced, and the calculation formula is as follows:
wherein (x)0，y0) Coordinates, f' (x), representing a certain pixel point in the image0,y0) Is filtered at (x)0,y0) At pixel value, f (x, y) is (x) before filtering0,y0) The pixel value, k and l, of a certain pixel point in the neighborhood are respectively expressed by (x)0，y0) The length and width of the central neighborhood window; w (x)0,y0X, y) is the weight of a point (x, y) that depends on the spatial domain kernel d (x, y)0,y0X, y) and a value range kernel r (x)0,y0X, y), the calculation formula is as follows:
wherein σdAnd σrFiltering parameters for the spatial domain and the value domain, respectively.
(2) The most widely applied fractional order differential template, namely the Tiansi operator, is decomposed into four small templates of 3 × 3 along the coordinate axis direction, the pixel points to be processed are respectively enhanced by using each small template, and the four templates decomposed along the coordinate axis direction are shown in FIG. 2. in FIG. 2, a0＝1，a1＝-v，
v is the fractional order of differentiation;
in the processing process, the pixel point (x, y) to be processed is a constant coefficient 2a0At the position, performing convolution operation on each pixel point in the image by using four 3 × 3 templates in fig. 2 respectively to obtain four convolution values M corresponding to the coordinate axis directions respectively1、M2、M3、M4(ii) a For a pixel point (x, y) to be processed in an image, the pixel value of the point is s, and s can be M1、M2、M3、M4To indicate that:
respectively carrying out edge extraction on the preprocessed MRI images by using two Flow-XDoG operators with different rho to obtain an image A (smaller rho) with less edge information and an image B (larger rho) with more edge information, and then carrying out edge connection on the image A by taking the image B as a guide image, wherein the specific process is as follows:
(1) the Flow-XDoG operator is an anisotropic filter that extracts a set of smooth and abstract edge curves from an image; the edge extraction process for the preprocessed MRI image using the Flow-XDoG operator is as follows:
1) calculating image gradient to obtain gradient image
And (3) respectively performing convolution calculation on the image by using the sobel operators in the eight directions, wherein the direction with the maximum gradient assignment is taken as the gradient of the current point, and the sobel operator templates in the eight directions are shown in FIG. 3.
2) Construction Edge tangential Flow (Edge indexed Flow, ETF)
For a given input image, defining the vector perpendicular to the image gradient as the edge tangent vector, denoted by t (x, y), which can be regarded as the local edge curve direction in a sense, the specific construction formula is as follows:
wherein omegaμIs used to denote by (x)0,y0) A centered, μ radius kernel, (x, y) is used to denote (x)0,y0) The pixels in the neighborhood, k, are used to represent the normalization factor, tn(. cndot.) denotes an iteratively determined local edge tangent vector with a period of 2 pi, and specifies t0(. is) an image gradient vector, phi (x)0,y0X, y) denotes two tnAngle of inclination, ws(x0,y0X, y) is a spatial weighting function, wm(x0,y0X, y) is a magnitude weighting function, wd(x0,y0X, y) represents a directional weight function; their calculation formulas are shown in formulas (6) to (9):
wd(x0,y0,x,y)＝|tn-1(x0,y0)·tn-1(x,y)| (9)；
wherein the content of the first and second substances,
represents a normalized gradient magnitude; to ensure smoothness of the edge, the number of iterations is set to 3.
3) Contour line extraction using extended Gaussian differential filtering (XDoG)
And (3) applying extended Gaussian difference filtering (XDoG) to the obtained edge tangential flow graph to extract a contour line, wherein the application formula is as follows:
wherein G isσRepresenting a Gaussian function, x representing a pixel (x)0,y0) S represents an arc length parameter, T represents the size of the Gaussian function kernel, + -S represents the upper and lower integral limits of the arc length parameter S, I (l)x,s(t)) is represented byx,sA pixel value at (t); lx,s(t) represents a tangent line segment perpendicular to and intersecting a certain curve at an intersection point of (x)0,y0)；Dσ,k,ρ(t) represents extended Gaussian difference Filter (XDoG), and the calculation formula is shown in formula (11), σmThe size of S is determined:
wherein the content of the first and second substances,
and
representing a Gaussian operator, σcAnd σsDenotes the scale size of a Gaussian template, let σ denotes＝1.6σc，σcFor later edge join, two different values of ρ are used, where ρ controls the number of lines extracted by the operator, the larger the value of ρ the more lines are extracted, where ρ is equal to 1.01＝0.99、ρ2Each image is subjected to contour extraction, and the extraction result is subjected to threshold processing by using a threshold function of formula (12) to obtain a final contour extraction graph:
wherein, T(u) represents a pixel value output after thresholding, u represents a pixel value input to the thresholding function, represents an error coefficient,
representing a luminance parameter.
4) Threshold segmentation
Because the extracted line profile image is not a binary image and cannot be subjected to connected domain threshold filtering, the image is subjected to binarization processing by using an Otsu threshold method; for a given image, T is marked as a segmentation threshold value of the foreground and the background, and the ratio of the number of foreground points in the image is p0Average gray of u0The background point number is p in the proportion of the image1Traversing T from the minimum gray value to the maximum gray value, wherein when the variance value is maximum due to T, the T is the optimal threshold value for segmentation, and the variance value calculation formula is as follows:
＝p0·p1·(u0-u1)2(13)。
5) connected-domain threshold filtering
Due to the influence of residual noise and complex texture information, the extracted contour has many false interference edges, and a part of interference items can be removed by analyzing the connected domain and setting an area threshold, and an effect schematic diagram is shown in fig. 5.
(2) Edge connection, the specific process is as follows:
in FIG. 5(b) and FIG. 5(d), respectively, are ρ1＝0.99、ρ2Outline when 1.0Taking a graph, the error edges in fig. 5(b) are fewer but have an edge missing phenomenon, and the edge details in fig. 5(d) are rich but have a large number of error edges, so an edge connection scheme is applied to process fig. 5(b), and the specific process is as follows: for each edge pixel (x) in FIG. 5(b)i,yi) Finding whether there is an edge point (x ') in the neighborhood of the corresponding position of FIG. 5 (d)'i,y′i) And if so, is located at (x 'in FIG. 5 (b)'i,y′i) Until the complete image is traversed, the join effect is as shown in fig. 6.
Step 3, removing error edge interference:
from the results shown in fig. 6, it can be seen that there are still false edges inside the organ, and to solve this problem, image subtraction is applied to obtain the difference image of fig. 6 and fig. 5 (b); the erroneous edges that need to be removed are then selected by connected component analysis and removed in fig. 6, resulting in the results shown in fig. 7.
And 4, removing the false edge, performing edge thinning, and finishing the extraction of the abdominal MRI image contour, wherein the specific process is as follows:
(1) weighted neighborhood Total variation (WTV)
The neighborhood Total variation reflects the spectral change degree in the neighborhood, the edge has a larger response value, the contribution of each pixel to the neighborhood Total variation is considered to be different, the Weighted neighborhood Total variation (WTV) is used for further filtering the false edge generated in the process of edge extraction, and the edge is preliminarily refined, and the calculation formula is as follows:
wherein D represents a point (x)0,y0) The weighted neighborhood total variation value of (x) is expressed by Ω0,y0) A central neighborhood space, h (x, y) represents a Gaussian function, ▽ f (x, y) represents a gradient, and in the neighborhood, if the weighted total variation of the point is greater than that of more than half of the points in the neighborhood, the point can be retained as an edge pointOtherwise, removing the residual liquid; because the relation between the current point and the window neighboring point and the relation between the window and the window are considered, the method not only can preliminarily refine the edge, but also can better filter the false edge caused by noise and tiny ground feature details.
(2) Image skeleton extraction
Actually, the central pixel outline of the target on the image is extracted, and the main idea is that the target is continuously corroded and refined from the periphery of the target to the center of the target by utilizing the characteristic of a 3-by-3 pixel window taking the pixel to be detected as the center until the target is corroded to be incapable of being corroded (the width of a single-layer pixel), so that the edge of the single-layer pixel width is obtained; here, Zhang-Suen refinement algorithm (pseudo code is shown in FIG. 8) is applied to achieve edge refinement, where A (P)1) Is from P9Go back to P1From 0 to 1, B (P)1) Is P1The number of non-zero neighborhood pixels.
As can be seen from fig. 4, the bilateral filtering can not only smooth the image noise but also retain the edge information, and the fractional differential filtering can further smooth the texture structure inside the organ or tissue, thereby reducing the influence on the extraction result.
As can be seen from fig. 5, as ρ increases, the extracted edge information is richer, and the number of false edges is also larger, and a connected domain threshold filtering is introduced, so that a part of false edges can be removed, and the number of false edges in the image is reduced.
As shown in fig. 6, the method using neighborhood search can connect the fracture edges in fig. 5(b), but adds many error edges, which indicates that the Flow-XDoG operator using only two parameters cannot meet the requirement of practical application, and further processing is required on the extraction result.
As can be seen from fig. 7, the edge extraction result after the post-processing is introduced is clearer and more complete, the number of false edges is less, which illustrates the necessity of the method of the present invention, but the edge is too coarse and has too large difference with the actual edge, so that the steps of edge discrimination and edge refinement of weighted total variation are introduced, and as can be seen from fig. 9(b), the edge of the refined result is clearer and more complete, and the number of false edges is less.
As can be seen from fig. 10, the edge of the image contour extracted by the Flow-XDoG operator-based abdominal MRI image contour extraction method is clearer and more complete, and has less interference, and especially when the edge in an area where gradient change is not significant is extracted, the effect is better than that of other prior art. Therefore, the method provided by the invention is suitable for extracting the outline of the abdominal Magnetic Resonance Image (MRI).
Compared with the classic edge extraction operator and the edge extraction method of non-photorealistic drawing, the abdomen MRI image contour extraction method based on the Flow-XDoG operator has the advantages that the extracted edge is clearer, more complete and more accurate, the network training is not depended on, the problem of the data volume of a training set does not need to be solved, and the abdomen MRI image contour extraction method based on the Flow-XDoG operator is very suitable for contour extraction of abdomen MRI images or other images with complex graphic mechanisms and texture structures.
Although the methods and techniques of the present invention have been described in terms of preferred embodiments, it will be apparent to those of ordinary skill in the art that variations and/or rearrangements of the methods and techniques described herein may be made without departing from the spirit and scope of the invention. It is expressly intended that all such similar substitutes and modifications apparent to those skilled in the art are deemed to be within the spirit, scope and content of the invention.
Claims (10)
1. A Flow-XDoG operator-based abdominal MRI image contour extraction method is characterized by comprising the following steps:
step 1, abdominal MRI image preprocessing: reducing the interference caused by noise and deviation fields, and enhancing and smoothing the image;
step 2, extracting the edges of the tissues or organs: respectively carrying out edge extraction on the preprocessed MRI image by using two Flow-XDoG operators with different parameters, and obtaining a preliminary result through edge connection;
step 3, removing error edge interference;
and 4, removing the false edge, and performing edge thinning to finish the extraction of the abdominal MRI image contour.
2. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to the claim 1, characterized in that, in the step 1, the image is filtered bilaterally, and the interference of noise and deviation field to the contour extraction is reduced; and (3) enhancing the texture details of the image and reserving a smooth area in the image by using fractional order differential filtering.
3. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 2, wherein in step 1, the fractional differential template used by the fractional differential filtering is Tiansi operator.
4. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 1, wherein in the step 2, the edge extraction process of the preprocessed MRI image by using the Flow-XDoG operator is as follows:
(1) calculating the image gradient to obtain a gradient image;
(2) constructing Edge Tangential Flow (ETF);
(3) extracting a line profile using extended gaussian differential filtering (XDoG);
(4) dividing a threshold value;
(5) and filtering a connected component threshold.
5. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 1, characterized in that in step 2, the Flow-XDoG operator with two different parameters ρ is used to perform edge extraction on the preprocessed MRI image, so as to obtain an image A corresponding to a smaller ρ value and an image B corresponding to a larger ρ value; the parameter rho is used for controlling the number of lines extracted by an operator, and the larger the rho value is, the more lines are extracted.
6. Flow-XDoG based operator according to claim 5The method for extracting the contour of the abdomen MRI image according to (1), wherein in the step 2, the edge connection means that if each edge pixel (x) of the image a is connectedi,yi) Edge points (x ') exist in the neighborhood of the corresponding position of image B'i,y′i) Then it is located at (x 'in graph A'i,y′i) Is set as an edge point.
7. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 1, characterized in that, in step 3, by obtaining the difference information between the images and the connected domain threshold filtering, the false edges generated due to noise and other interference factors and the false edges located inside the organ are removed.
8. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 1, characterized in that in step 4, the weighted neighborhood total variation is used to filter out the false edges generated in the edge extraction process, and the edges are preliminarily refined.
9. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 8, wherein in the step 4, the calculation formula of the weighted neighborhood total variation is as follows:
wherein (x, y) represents the coordinates of a pixel point in the image, (x)0，y0) Indicating the coordinates of a certain pixel in the image, D indicating the point (x)0,y0) The weighted neighborhood total variation value of (x) is expressed by Ω0,y0) A neighborhood space of the center, h (x, y) represents a Gaussian function,
the gradient is indicated.
10. The abdomen MRI image contour extraction method based on the Flow-XDoG operator according to claim 8, wherein in step 4, the edge is further refined by a skeleton extraction algorithm to obtain a final edge refinement result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010276311.6A CN111583286B (en) | 2020-04-09 | 2020-04-09 | Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010276311.6A CN111583286B (en) | 2020-04-09 | 2020-04-09 | Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111583286A true CN111583286A (en) | 2020-08-25 |
| CN111583286B CN111583286B (en) | 2023-01-20 |
Family
ID=72113007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010276311.6A Active CN111583286B (en) | 2020-04-09 | 2020-04-09 | Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111583286B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113129235A (en) * | 2021-04-22 | 2021-07-16 | 深圳市深图医学影像设备有限公司 | Medical image noise suppression algorithm |
| CN113436212A (en) * | 2021-06-22 | 2021-09-24 | 广西电网有限责任公司南宁供电局 | Extraction method for inner contour of circuit breaker static contact meshing state image detection |
| CN115100304A (en) * | 2022-04-24 | 2022-09-23 | 南通海隆磁业有限公司 | Nuclear magnetic resonance image enhancement method based on image processing |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102800063A (en) * | 2012-07-12 | 2012-11-28 | 中国科学院软件研究所 | Image enhancement and abstraction method based on anisotropic filtering |
| CN103914863A (en) * | 2014-03-10 | 2014-07-09 | 西藏民族学院 | Method for abstractly drawing color image |
| CN108320799A (en) * | 2018-03-17 | 2018-07-24 | 北京化工大学 | Image analysis and recognition method for lateral flow paper strip disease diagnosis |
| CN108961278A (en) * | 2018-06-20 | 2018-12-07 | 深圳市旭东数字医学影像技术有限公司 | The method and its system of abdominal wall muscle segmentation based on image data |
| CN110097626A (en) * | 2019-05-06 | 2019-08-06 | 浙江理工大学 | A kind of basse-taille object identification processing method based on RGB monocular image |
-
2020
- 2020-04-09 CN CN202010276311.6A patent/CN111583286B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102800063A (en) * | 2012-07-12 | 2012-11-28 | 中国科学院软件研究所 | Image enhancement and abstraction method based on anisotropic filtering |
| CN103914863A (en) * | 2014-03-10 | 2014-07-09 | 西藏民族学院 | Method for abstractly drawing color image |
| CN108320799A (en) * | 2018-03-17 | 2018-07-24 | 北京化工大学 | Image analysis and recognition method for lateral flow paper strip disease diagnosis |
| CN108961278A (en) * | 2018-06-20 | 2018-12-07 | 深圳市旭东数字医学影像技术有限公司 | The method and its system of abdominal wall muscle segmentation based on image data |
| CN110097626A (en) * | 2019-05-06 | 2019-08-06 | 浙江理工大学 | A kind of basse-taille object identification processing method based on RGB monocular image |
Non-Patent Citations (4)
| Title |
|---|
| HAMEDA ABD EL-FATTAH EL-SENNARY等: "Edge Detection of an Image Based on Extended Difference of Gaussian", 《AMERICAN JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY》 * |
| RU XUE等: "An extended flow-based difference-of-Gaussians method of line drawing for polyhedral image", 《OPTIK》 * |
| 李刚等: "应用启发式边缘生长的遥感影像分割", 《应用科学学报》 * |
| 李阳等: "多变量高斯差分的风格化方法", 《云南大学学报(自然科学版)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113129235A (en) * | 2021-04-22 | 2021-07-16 | 深圳市深图医学影像设备有限公司 | Medical image noise suppression algorithm |
| CN113436212A (en) * | 2021-06-22 | 2021-09-24 | 广西电网有限责任公司南宁供电局 | Extraction method for inner contour of circuit breaker static contact meshing state image detection |
| CN115100304A (en) * | 2022-04-24 | 2022-09-23 | 南通海隆磁业有限公司 | Nuclear magnetic resonance image enhancement method based on image processing |
| CN115100304B (en) * | 2022-04-24 | 2024-04-19 | 江苏中勤通信科技有限公司 | Nuclear magnetic resonance image enhancement method based on image processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111583286B (en) | 2023-01-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111583286B (en) | Abdomen MRI (magnetic resonance imaging) image contour extraction method based on Flow-XDoG operator | |
| CN107644420B (en) | Blood vessel image segmentation method based on centerline extraction and nuclear magnetic resonance imaging system | |
| CN109300113B (en) | Pulmonary nodule auxiliary detection system and method based on improved convex hull method | |
| US7233329B2 (en) | Rendering for coronary visualization | |
| Pang et al. | Automatic lung segmentation based on texture and deep features of HRCT images with interstitial lung disease | |
| US7711167B2 (en) | Fissure detection methods for lung lobe segmentation | |
| JP6570145B2 (en) | Method, program, and method and apparatus for constructing alternative projections for processing images | |
| WO2004049923A1 (en) | Image processing device and image processing method | |
| CN107292835B (en) | Method and device for automatically vectorizing retinal blood vessels of fundus image | |
| CN112102385B (en) | Multi-modal liver magnetic resonance image registration system based on deep learning | |
| WO2022141145A1 (en) | Object-oriented high-resolution remote sensing image multi-scale segmentation method and system | |
| CN111292307A (en) | Digestive system gallstone recognition method and positioning method | |
| CN116580068B (en) | Multi-mode medical registration method based on point cloud registration | |
| Li et al. | A visual saliency-based method for automatic lung regions extraction in chest radiographs | |
| Lee et al. | Unsupervised segmentation of lung fields in chest radiographs using multiresolution fractal feature vector and deformable models | |
| CN113223004A (en) | Liver image segmentation method based on deep learning | |
| Liu et al. | Extracting lungs from CT images via deep convolutional neural network based segmentation and two-pass contour refinement | |
| CN111383759A (en) | Automatic pneumonia diagnosis system | |
| CN113269029B (en) | Multi-modal and multi-characteristic finger vein image recognition method | |
| Yu et al. | Biomedical image segmentation via constrained graph cuts and pre-segmentation | |
| CN111062909A (en) | Method and equipment for judging benign and malignant breast tumor | |
| CN108921860B (en) | Full-automatic segmentation method for prostate magnetic resonance image | |
| Jiang et al. | Model-based automated extraction of microtubules from electron tomography volume | |
| Wei et al. | An algorithm for segmentation of lung ROI by mean-shift clustering combined with multi-scale HESSIAN matrix dot filtering | |
| CN112017132A (en) | Vein image enhancement method based on maximum curvature method and multi-scale Hessian matrix |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |