TW201722357A - Three-dimensional median filtering method applied to computerized tomographic image eliminating the discontinuity occurred in the combination of two-dimensional section tomographic images - Google Patents

Abstract

A three-dimensional median filtering method applied to a computerized tomographic image comprises the steps of: (A) reading a set of two-dimensional section computerized tomographic images which can be utilized to construct a three-dimensional image; (B) calibrating all coordinate positions of noises in the set of two-dimensional section computerized tomographic images; and (C) analyzing a distribution state of each noise in a two-dimensional section tomographic image and the adjacent two-dimensional section tomographic image where the noise is located; (D) performing a filtering algorithm according to the distribution state of the noise to obtain a median image datum that is utilized to replace the noise; and (E) processing the noise of the set of two-dimensional section computerized tomographic images to construct the three-dimensional image. In this way, the discontinuity occurred in the combination of the two-dimensional section tomographic images can be eliminated when the three-dimensional image is processed for stacking imagining.

Description

Three-dimensional median filtering method applied to computer tomographic images

The invention relates to an image processing method for computer tomographic images, in particular to a three-dimensional median filtering method for computer tomographic images.

Computer Tomography uses X-rays to generate images. X-rays are generated by X-tubes. When X-rays penetrate different densities, X-rays will be attenuated to varying degrees due to differences in tissue density. X-rays with different degrees of intensity are absorbed by the scintillator (CsI) and converted into visible light. The visible light is absorbed by a one-dimensional or two-dimensional CCD or CMOS image sensor, and finally the X-ray projection image of the object is obtained. The X-ray tube and the detector will continue to rotate on the cross-section of the object to be tested, and one image will be taken at a certain angle until the multi-directional image is rotated by more than 180 degrees; the one-dimensional detection is used. When the X-tube is rotated in a spiral manner, the object to be detected is rotated by a plurality of turns. Using a two-dimensional detector and a cone-shaped X-tube, it is only necessary to rotate the object for one rotation, and then the computer will obtain an algorithm through the algorithm. The X-ray image is reconstructed into a two-dimensional computerized tomographic image. The two-dimensional image is stacked to form a three-dimensional image. The three-dimensional image is generated by two-dimensional image synthesis, so it is often mixed due to the accuracy of the image. Speckle noise (impulsive noise).

At present, many image processing technologies provide a variety of methods to solve the noise problem, but if the purpose is only to remove the noise, such a process can effectively visually eliminate the spot noise of each slice image (two-dimensional image). When these individual image-processed two-dimensional slice images are stacked into a three-dimensional image, the two-dimensional slice image will cause discontinuity of the object slice combination when the three-dimensional image is imaged. These discontinuities are caused by the image slice after the image processing. The edges cannot be aligned, so the 3D image reconstructed by the detected object 3D will produce contour distortion.

Therefore, there is a great need in the industry to develop an image processing method for three-dimensional computed tomography images, which can eliminate the discontinuity of the object slice combination in the three-dimensional image stacking imaging, so that the noise can be eliminated simultaneously and the three-dimensional computer can be avoided. The problem of contour distortion occurs when the tomographic image is constructed.

In view of the above-mentioned shortcomings of the prior art, the main object of the present invention is to provide a three-dimensional median filtering method for computerized tomographic images, which integrates a set of two-dimensional sliced computer tomographic images and a noise distribution state in which three-dimensional images can be constructed. A filter algorithm and a median image data are used to construct a three-dimensional computed tomography image that eliminates noise and avoids contour distortion.

In order to achieve the above object, according to one aspect of the present invention, a three-dimensional median filtering method for computer tomographic images is provided, the steps comprising: (A) reading a set of two-dimensional slice computed tomography images capable of constructing three-dimensional images (B) calibration of the coordinate position of all noise of the two-dimensional section computer tomography image (C) analyze the noise distribution of the two-dimensional section computer tomographic image of each point of noise and the adjacent two-dimensional section computer tomogram; (D) perform a filtering algorithm according to the distribution state of the noise A median image data is obtained, and the noise is replaced by the median image data; (E) the noise of the two-dimensional sliced computer tomographic image is processed to construct a three-dimensional image.

The calibration of the coordinate position of the noise in the step (B) is to give a Cartesian coordinate value (x, y, z) of the noise in the Z-dimensional two-dimensional section computer tomographic image, wherein the z system represents the Z-dimensional two-dimensional section Computer tomographic image, x, y represents the two-dimensional coordinates of the noise in the plane of the Z-dimensional two-dimensional computerized tomographic image, and the noise of the Cartesian coordinates (x, y, z) of the noise can accurately locate the noise. The group can construct the position of the two-dimensional slice computer tomographic image of the three-dimensional image.

The two-dimensional section computer tomographic image of the noise in step (C) and its adjacent two-dimensional section computer tomographic image refer to the Z-dimensional two-dimensional section computer tomographic image with noise and the third Z, Z +1 piece of two-dimensional section computer tomography image; the noise of the analysis and processing of the present invention does not simply deal with two-dimensional noise information, but analyzes the noise in the Z-1, Z and Z+1 two-dimensional section computer tomography image The 3-dimensional distribution state on the top.

According to the three-dimensional distribution state of the noise, the noise can be divided into three types (but not limited thereto) for subsequent image processing, and the present invention proposes three types of regions (but not limited thereto). The first-order Descartes coordinate area, the second-order Descartes coordinate area, and the third-order Descartes coordinate area; when the three-dimensional distribution of the noise points falls into the first-order Descartes coordinate area The range of the filtering algorithm proposed by the present invention is: I ₁ (0,0,0)=median({I(x,y,x))|(x,y,z) S ₁ }) where S ₁ represents the first equal Cartesian coordinate region, including (0, 0, 0), ( ₁ , 0, 0), (0, ₁ , 0), (0, 0, 1 ), (-1,0,0), (0,-1,0), (0,0,-1), where S ₁ : first equal Cartesian coordinate region, I ₁ : perform filtering calculation After the law, the median image data of the noise at the coordinate position, (0,0,0) is the coordinate of the calibrated noise point in the coordinate area of the first-order Descartes; when the 3D of the noise When the distribution state falls within the range of the second-order Cartesian coordinate region, the filtering algorithm proposed by the present invention is: I ₂ (0,0,0)=median({I(x,y,x))|( x,y,z) S ₂ }) wherein S ₂ represents a second equal Cartesian coordinate region, including (0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1 ), (-1,0,0), (0,-1,0), (0,0,-1), (0,1,1), (0,-1,1), (0,1 ,-1), (0,-1,-1), (1,1,0), (1,-1,0), (-1,1,0), (-1,-1,0) , (1,0,1), (-1,0,1), (1,0,-1), (-1,0,-1), where S ₂ : second-order Descartes coordinates Area, I ₂ : After performing the filtering algorithm, the median image data of the noise at the coordinate position, (0, 0, 0) is the coordinate of the already calibrated noise point in the second-order Cartesian coordinate area. When the 3-dimensional distribution state of the noise falls within the range of the third-order Cartesian coordinate region, the filtering algorithm proposed by the present invention is: I ₃ (0, 0, 0) = median ({I(x) ,y,x))|(x,y,z) S ₃ }) where S ₃ represents a third-order Cartesian coordinate region including (0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1 ), (-1,0,0), (0,-1,0), (0,0,-1), (0,1,1), (0,-1,1), (0,1 ,-1), (0,-1,-1), (1,1,0), (1,-1,0), (-1,1,0), (-1,-1,0) , (1,0,1), (-1,0,1), (1,0,-1), (-1,0,-1), (1,1,1), (-1,1 , 1), (1, -1, 1), (-1, -1, 1), (1, 1, -1), (1, -1, -1), (-1, 1, -1 ), (-1,-1,-1), where S ₃ : third-order Cartesian coordinate region, I ₃ : median image data of the noise at the coordinate position after performing the filtering algorithm, ( 0,0,0) is the coordinate of the calibrated noise point in the third-order Cartesian coordinate area.

The above summary and the following detailed description and drawings are intended to further illustrate the manner, means and effects of the present invention in achieving its intended purpose. Other purposes and advantages of this creation will be explained in the following description and drawings.

S101-S105‧‧‧Steps

S ₁ ‧‧‧First Descartes Coordinate Area

S ₂ ‧‧‧Second Descartes Coordinate Area

S ₃ ‧‧‧ Third Descartes Coordinate Area

The first figure is a flow chart of a three-dimensional median filtering method applied to computer tomographic images according to the present invention; the second figure is a schematic diagram of a first-order Cartesian coordinate area of the present invention; Schematic diagram of the Descartes coordinate area; the fourth figure is a schematic diagram of the third-order Cartesian coordinate area of the present invention.

The embodiments of the present invention are described by way of specific examples, and those skilled in the art can readily understand the advantages and effects of the present invention from the disclosure of the present disclosure.

Various stereoscopic medical images (such as: CT, MRI, PET, etc.) tend to cause three-dimensional synthetic images to be doped with spots due to the accuracy of image acquisition. (impulsive noise), the method for eliminating speckle noise in the present invention is to use a median filtering method to perform noise filtering on each two-dimensional computed tomography image, but if only the median filtering method is used, each two-dimensional computerized tomographic image is mixed. The filtering can obviously eliminate the speckle noise of each tomographic image visually, but it will cause the discontinuity of the object slice combination in the 3D image imaging, so that the edges between the filtered object slices cannot be aligned.

Please refer to the first figure, which is a flow chart of a three-dimensional median filtering method applied to computer tomographic images according to the present invention. As shown in the figure, the present invention provides a three-dimensional median filtering method for computer tomographic images, the steps comprising: (A) reading a set of two-dimensional slice computer tomography image S101 capable of constructing a three-dimensional image; (B) calibration The coordinate position of all the noises of the two-dimensional section computerized tomographic image S102; analyzing the distribution state of the two-dimensional section computer tomographic image of each point of noise and the adjacent two-dimensional section computer tomographic image S103; according to the distribution of noise The state performs a filtering algorithm to obtain a median image data, and replaces the noise S104 with the median image data; and processes the noise of the two-dimensional slice computer tomographic image to construct a three-dimensional image S105.

The filtering algorithm of the present invention uses a median filter for noise filtering, and a median filter is particularly useful when impulse noises appear, because the pulse noise appears to be superimposed on The white point and the black point on the image are also called salt-and-pepper noise. The filtering algorithm using the median filter of the present invention is a sorting operation using a mask, and the mask is used. Sort operation is a kind of non Linear filtering method, first sorting the grayscale value of all the pixels in the mask, and then selecting the image data of the grayscale value (or the median grayscale value, the median grayscale value) sorted in the middle, Then, the median value (grayscale) image data is substituted into the corresponding pixel (noise image data), so that a pixel filtering operation is completed, and the method can not only filter out the high frequency noise portion of the image in the image. The edge of the image can also be appropriately reserved, and then the value of the pixel (the image data of the noise) is replaced by the median image data of the pixel neighbor gray scale (filtered algorithm operation), and the present invention is Depending on the size of the noise, choose a median filter of different masks in the form of three equal Cartesian coordinates.

Please refer to the second figure, which is a schematic diagram of the first-order Cartesian coordinate area of the present invention. Please refer to the third figure, which is a schematic diagram of the second-order Cartesian coordinate area of the present invention. Please refer to the fourth figure, which is the first embodiment of the present invention. Schematic diagram of the third-order Descartes coordinate area. The steps of the embodiment of the present invention are as follows:

Step (1) first reads a set of two-dimensional slice computer tomographic images that can construct a three-dimensional image, and (x, y, z) represents the (x, y) coordinate value of the z-th two-dimensional slice image, and the two-dimensional slice image The plane is digitized to facilitate position calibration and digital processing of image data.

Step (2) defines three Cartesian coordinate regions of different area sizes, including a first-order Cartesian coordinate region, a second-order Cartesian coordinate region, and a third-order Cartesian coordinate region. The coordinates in the Cartesian coordinate region of three different region sizes are different from the coordinates (x, y, z) of step (1), wherein the first equal Cartesian coordinate region (S ₁ ) includes (0, 0, 0), (1,0,0), (0,1,0), (0,0,1), (-1,0,0), (0,-1,0), (0,0, -1) (as shown in Figure 2), the second-order Cartesian coordinate region (S ₂ ) includes (0,0,0), (1,0,0), (0,1,0), ( 0,0,1), (-1,0,0), (0,-1,0), (0,0,-1), (0,1,1), (0,-1,1) , (0,1,-1), (0,-1,-1), (1,1,0), (1,-1,0), (-1,1,0), (-1, -1,0), (1,0,1), (-1,0,1), (1,0,-1), (-1,0,-1) (as shown in Figure 3), The third-order Cartesian coordinate region (S ₃ ) includes (0,0,0), (1,0,0), (0,1,0), (0,0,1), (-1,0) , 0), (0, -1, 0), (0, 0, -1), (0, 1, 1), (0, -1, 1), (0, 1, -1), (0 ,-1,-1),(1,1,0), (1,-1,0), (-1,1,0), (-1,-1,0), (1,0,1 ), (-1,0,1), (1,0,-1), (-1,0,-1), (1,1,1), (-1,1,1), (1, -1,1), (-1,-1,1), (1,1,-1) (1, -1, -1), (- 1,1, -1), (- 1, -1, -1) (as shown in Figure IV).

Step (3) Check the noise distribution of the three-dimensional two-dimensional tomographic image of the z-th slice and the front and rear slices (the z+1th, the z-1th), and record the coordinate points of all the noise coordinates (x, y , z), where the noise is usually not an individual, the distribution of a noise may contain many noise particles, so the coordinate point (x, y, z) of the noise coordinate position can use the centroid point (not The calculation method of this limit is to calculate the centroid point of a certain community noise particle, and the centroid point coordinates the coordinate point (x, y, z) of the noise coordinate position to facilitate calibration of each point noise. In this group, the position of the two-dimensional image computerized tomographic image of the three-dimensional image can be constructed, but in the subsequent filtering algorithm (median filtering), the coordinates of each noise point in the Cartesian coordinate region to which it belongs are Set to (0,0,0).

Step (4) selects the mask size of the median filter and the position of the adjacent point according to the noise distribution. When the noise distribution is scattered in the first isochronous Cartesian coordinate region, the mask size of the median filter is 7 The position of the adjacent point (the noise distribution area) is selected as S ₁ ; when the noise is scattered in the second equal Cartesian coordinate area, the mask size of the median filter is 19, and the adjacent point position The choice of (noise distribution area) is S ₂ ; when the noise is scattered in the third-order Cartesian coordinate area, the mask size of the median filter is 27, and the adjacent point position (noise distribution area) is selected to S _3.

Step (5) performs a different filtering algorithm (median filtering) according to which isochronous region of the noise distribution region is located, if the noise distribution is located in the first isochronous Cartesian coordinate region When using equation (1) to complete the filtering: I ₁ (0,0,0)=median({I(x,y,x))|(x,y,z) S ₁ })... In the formula (1), I ₁ is the middle gray scale value (median image data) of the noise at the coordinate position after the execution of the filtering algorithm, and the coordinates (0, 0, 0) is the coordinate of the noise coordinate position in the first-order Cartesian coordinate area (S ₁ ), and the other coordinates are other noise particles (noise distribution) in the first-order Cartesian coordinate area (S ₁ ) Coordinates; if the noise distribution is in the first-order Cartesian coordinate region, use (2) to complete the filtering: I ₂ (0,0,0)=median({I(x,y,x))| (x,y,z) S ₂ })... In equation (2) (2), I ₂ is the middle gray scale value (median image data) of the noise at the coordinate position after the execution of the filtering algorithm, and the coordinates (0, 0, 0) is the coordinate of the noise coordinate position in the second-order Cartesian coordinate area (S ₂ ), and the remaining coordinates are other noise particles (noise distribution) in the second-order Cartesian coordinate area (S ₂ ) Coordinates; if the noise distribution is in the third-order Cartesian coordinate region, use (3) to complete the filtering: I ₃ (0,0,0)=median({I(x,y,x))| (x,y,z) S ₃ })... In equation (3) (3), I ₃ is the middle gray scale value (median image data) of the noise at the coordinate position after the execution of the filtering algorithm, and the coordinates (0, 0, 0) is the coordinate of the noise coordinate position in the second-order Cartesian coordinate area (S ₃ ), and the remaining coordinates are other noise particles (noise distribution) in the third-order Cartesian coordinate area (S ₃ ) The coordinates of the coordinates.

Step (6) repeat steps (4)-(5) until all noise is processed.

Step (7) After all the noises of the two-dimensional sliced computer tomographic image are processed, the three-dimensional image is constructed in sequence.

The above-described embodiments are merely illustrative of the features and functions of the present invention and are not intended to limit the scope of the technical content of the present invention. Any person skilled in the art can modify and change the above embodiments without departing from the spirit and scope of the creation. Therefore, the scope of protection of this creation should be as listed in the scope of the patent application described later.

S101-S105‧‧‧Steps

Claims (7)

A three-dimensional median filtering method for computer tomographic images, the steps comprising: (A) reading a set of two-dimensional slice computer tomographic images capable of constructing three-dimensional images; (B) calibrating the two-dimensional slice computer tomographic images of the group Coordinate position of the signal; (C) analyze the distribution of the two-dimensional section computer tomographic image of each point of noise and the adjacent two-dimensional section computer tomogram; (D) perform a filtering algorithm according to the distribution state of the noise A median image data is obtained, and the noise is replaced by the median image data; (E) the noise of the two-dimensional sliced computerized tomographic image is processed to construct a three-dimensional image. The three-dimensional median filtering method for computer tomographic image according to claim 1, wherein the coordinates of the coordinates of the noise in the step (B) are based on the Z-dimensional two-dimensional computerized tomographic image. The signal is given a Cartesian coordinate value (x, y, z). The three-dimensional median filtering method for computer tomographic images according to the second aspect of the patent application, wherein the distribution state of the noise is the Z-1, Z and Z+1 two-dimensional section computer tomography of the noise. The distribution state on the image. The three-dimensional median filtering method for computer tomographic image according to claim 2, wherein the distribution state of the noise in step (C) is one The first-order Descartes coordinate area, a second-order Descartes coordinate area, and a third-order Descartes coordinate area are classified. The three-dimensional median filtering method for computer tomographic images according to claim 4, wherein the filtering algorithm performed in the noise distribution of the first isochronous coordinate region in step (D) is performed. Is: I ₁ =median({I(x,y,x))|(x,y,z) S ₁ })S ₁ : First-order Cartesian coordinate area, I ₁ : Median image data of the noise at the coordinate position after performing the filtering algorithm. The three-dimensional median filtering method for computer tomographic image according to claim 4, wherein the filtering algorithm performed in the noise distribution of the second Cartesian coordinate in step (D) Is: I ₂ =median({I(x,y,x))|(x,y,z) S ₂ }).S ₂ : second-order Cartesian coordinate region, I ₂ : median image data of the noise at the coordinate position after performing the filtering algorithm. The three-dimensional median filtering method for computer tomographic image according to claim 4, wherein the filtering algorithm performed in the noise distribution of the third-order Cartesian coordinate in step (D) Is: I ₃ =median({I(x,y,x))|(x,y,z) S ₃ }) S ₃ : third-order Cartesian coordinate region, I ₃ : median image data of the noise at the coordinate position after performing the filtering algorithm.