CN103208106A - Method and device for detecting collimation side and X-ray imaging device - Google Patents

Abstract

The invention relates to the technical field of X-ray imaging and discloses a method for detecting a collimation side in an X-ray image, a device for detecting the collimation side and an X-ray imaging device. The method for detecting the collimation side in the X-ray image comprises the steps of dividing the image to obtain an initial boundary, extracting characteristics in the image to form a characteristic image and obtaining a final boundary from the initial border according to the characteristic image. By means of the method for detecting the collimation side in the X-ray image, the device for detecting the collimation side and the X-ray imaging device, the stability of collimation side detection is improved.

Description

Detect method, device and the x-ray imaging equipment on collimation limit

Technical field

The present invention relates to x-ray imaging art (radiography) technical field, particularly in radioscopic image, detect the device on collimation limit (collimating apparatus border), method and the x-ray imaging equipment on detection collimation limit.

Background technology

For most of radioscopic images, be not that the whole zone of image is all useful.Therefore, x-ray imaging equipment generally includes collimating apparatus, and described collimating apparatus is used for covering useless zone, makes these zones can not arrived by x-ray bombardment.But, in the radioscopic image that x-ray imaging equipment directly obtains, still comprise the zone that these are useless usually.So, how cutting out useful zone from radioscopic image, is a technical matters of being badly in need of solution.

The cutting radioscopic image generally needs to detect earlier the collimation limit.Detecting the collimation limit has had many methods, and these methods can be divided into two kinds substantially: rely on the method for hardware and do not rely on the method for hardware.For the method that relies on hardware, detection method utilizes the input of collimating apparatus geological information to calculate the position of the collimating apparatus in the image of obtaining.For the method that does not rely on hardware, there is parallel and paired vertical collimation limit in pairs in supposition in the image that obtains usually.In fact, the number on collimation limit is less, and may be blocked by the anatomy object.Therefore, under this prerequisite, existing detection collimation limit method may be failed.Therefore, need to improve the stability of collimation frontier inspection survey technology.

The application people is that University Of Ningbo, application number are 200910152622.5, Granted publication number discloses mammary gland method for extracting region in a kind of breast molybdenum target radioscopic image for the Chinese patent of CN101667297B.CN101667297B specifically discloses: when calculating the segmentation threshold of mammary gland zone to be extracted and background area, utilized the ratio that mammary gland is regional and the background area occupies in the breast molybdenum target radioscopic image of breast molybdenum target radioscopic image in a reasonable range, and the density in mammary gland zone is greater than the characteristics of the density of background area, make that like this segmentation threshold that calculates is more accurate, helps to improve the extraction precision in final mammary gland zone.

Summary of the invention

In view of this, the present invention proposes a kind of method that detects the collimation limit in radioscopic image, it does not rely on hardware, and has better stability.The invention allows for a kind of device that detects the collimation limit, and proposed a kind of x-ray imaging equipment.

The invention provides a kind of method that in radioscopic image, detects the collimation limit, comprising:

Image is cut apart, and obtained initial boundary;

Extract the feature in the image, form characteristic pattern;

According to characteristic pattern, obtain final border from described initial boundary growth.

Alternatively, this method further comprises: detect described final border; And, from described image, cut out the part in the final border.

Preferably, this method further comprises: the feature in extracting image and image cut apart before, image is carried out down-sampling, and/or, image is carried out Gauss's smoothing processing.

According to an embodiment of the invention, comprise described cutting apart:

The histogram of computed image;

Select when the class internal variance be hour or the possible threshold value when inter-class variance is maximum;

With pixel in the described image less than or split smaller or equal to the point of this threshold value.

Another embodiment according to the present invention selects to cut apart outermost starting point in the image of back, and forms described initial boundary with these starting points.

Preferably, the feature in the described extraction image comprises: calculate the gradient information in the described image, form outline line.

The present invention also provides a kind of device that detects the collimation limit in radioscopic image, comprising:

A cutting unit is used for image is cut apart, and obtains initial boundary;

A feature extraction unit for the feature of extracting image, forms characteristic pattern;

A border growing element according to characteristic pattern, obtains final border from described initial boundary growth.

Alternatively, this device further comprises: a cutting unit for detection of described final border, and cuts out part in the final border from described image.

Preferably, this device further comprises: a pretreatment unit, carried out pre-service to described image before cutting apart in the feature of extracting image with to image.

The present invention also provides a kind of x-ray imaging equipment, comprises the device on collimation limit in the aforesaid detection radioscopic image.

As can be seen, on the one hand the present invention does not rely on hardware from such scheme, and utilization of the present invention is cut apart and obtained initial boundary and obtain final border according to characteristic pattern from the initial boundary growth on the other hand.So, even the collimation limit is blocked by the anatomy object, the present invention also can grow and obtain final border, thereby has improved the stability that the collimation frontier inspection is surveyed.

In addition, the present invention is particularly suitable for the manual collimating apparatus of entry level system, and is adopting more traditional method of the non-self-adapting of Hough (Hough) conversion to have better stability than those aspect the processing picture noise.

Description of drawings

To make clearer above-mentioned and other feature and advantage of the present invention of those of ordinary skill in the art by describing the preferred embodiments of the present invention in detail with reference to accompanying drawing below, in the accompanying drawing:

Fig. 1 is a width of cloth radioscopic image.

Fig. 2 is the schematic flow sheet of method according to an embodiment of the invention.

Fig. 3 is the histogram when cutting apart in one embodiment of the invention.

Fig. 4 (a) is for cutting apart the direct exposure area that obtains; Fig. 4 (b) has shown outmost four points that extract in Fig. 4 (a); Fig. 4 (c) has shown initial boundary B ₀

Fig. 5 is for showing the characteristic pattern of gradient information.

Fig. 6 is for having shown final boundary B _fImage.

Fig. 7 is images cut.

Fig. 8 is the schematic representation of apparatus according to the detection collimation limit of further embodiment of this invention.

Wherein, Reference numeral comprises:

10: background

20: anatomical object

30: prospect

210-270: the step in the method

T: threshold value P ₁: threshold value starting point P ₂: the threshold value emphasis

B ₀: initial boundary B _f: final border

300: the device that detects the collimation limit

310: pretreatment unit

320: cutting unit

330: feature extraction unit

340: the border growing element

350: the cutting unit

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in more detail by the following examples.

As shown in Figure 1, generally, radioscopic image comprises direct exposure area and Indirect exposure zone, and wherein directly exposure area (background 10) gray-scale value is bigger, and Indirect exposure zone (anatomical object 20 and prospect 30) gray-scale value is lower.And the difference of anatomical object 20 and prospect 30 is homogeneity, and there is more rich gradient information in the zone of anatomical object 20, and there is less gradient information in the zone of prospect 30.

As shown in Figure 2, the method according to the detection of embodiment of the invention collimation limit comprises the steps:

Step 210 is carried out pre-service to image, to reduce noise and to save calculated amount.After this step, it is littler and through the image after level and smooth to obtain data volume.

Cut apart result with feature extraction because noise may influence image, therefore can preferably utilize the level and smooth or similar means of Gauss to reduce noise in the image in the present embodiment.On the other hand, cutting only is the part of whole x-ray imaging process flow diagram, and cutting must have very high efficient.Therefore, can adopt the reduction hits in the present embodiment, reduce calculated amount.

Smoothly be the noise reduction that example illustrates present embodiment with Gauss below, those skilled in the art also can adopt other means to realize this purpose.At two-dimensional space, the Gauss operator of isotropic is:

$G(x,y)=\frac{1}{{2\mathrm{\π\σ}}^2}{e}^{-\frac{x^2+y^2}{{2\mathrm{\σ}}^2}}$

Wherein, (x y) is Gauss operator to G, and σ is the standard deviation that just too distributes, and x, y are the coordinate of pixel in the image.Original image calculates convolution through following formula, is convolution kernel with top Gauss operator wherein.

$I_1(x,y)=G(x,y)\⊗I_0(x,y)$

Wherein, I ₁(x y) is pretreated image, I ₀(x y) is original image.

Step 220 after the pretreated image that acquisition step 210 is exported, is cut apart pretreated image, to obtain initial boundary.

In the present embodiment, to come split image based on histogrammic dividing method, image is divided into direct exposure area (background 10) and Indirect exposure zone (anatomical object 20 and prospect 30).So, by the histogram of split image, thereby realized cutting apart radioscopic image.

Below with a kind of based on histogrammic dividing method---Otsu algorithm---introduce how to cut apart entire image.Certainly, it will be appreciated by those skilled in the art that and also can adopt other similar method to come split image, for example iteration mean algorithm, C means clustering algorithm etc.

The Otsu algorithm is used to determine histogrammic image threshold usually.This algorithm will use prospect and background pixel simultaneously, and the threshold value that calculates optimization is opened two class region separation.

By making the class internal variance minimize to seek this threshold value, wherein, the class internal variance refers to a weighted sum of two kind variablees:

${\mathrm{\σ}}_w^2\left(t\right)={\mathrm{\ω}}_1\left(t\right){\mathrm{\σ}}_1^2\left(t\right)+{\mathrm{\ω}}_2\left(t\right){\mathrm{\σ}}_2^2\left(t\right)$

Wherein, σ _w ²(t) be the class internal variance; ω ₁(t) be weighting coefficient, representing two kinds by threshold value t and variable σ _i ²(t) probability that separates, wherein i=1 or 2.

In addition, the Otsu algorithm shows to make the class internal variance minimize and makes that the maximized effect of inter-class variance is identical, therefore, and also can be by making inter-class variance maximize to seek above-mentioned threshold value:

${\mathrm{\σ}}_b^2\left(t\right)={\mathrm{\σ}}^2-{\mathrm{\σ}}_w^2\left(t\right)={\mathrm{\ω}}_1\left(t\right){\mathrm{\ω}}_2\left(t\right){[{\mathrm{\μ}}_1\left(t\right)-{\mathrm{\μ}}_2\left(t\right)]}^2$

σ wherein _b ²(t) be inter-class variance, this expression formula is by class probability ω _iWith class average μ _iDetermine (wherein i=1 or 2) that wherein, the class average can be upgraded conversely in iterative process.

As shown in Figure 3, in order to remove useless information from histogram, adaptive threshold is limited at starting point p ₁With terminal point p ₂Between.Wherein,

p _i＝iHeight*iWidth*histogram_ratio

p ₂＝iHeight*iWidth*(1-histogram_ratio)

Wherein, iHeight and iWidth represent height and the width of radioscopic image respectively, and histogram_ration is the histogram ratio.

According to above-mentioned algorithm, cutting apart in the step 220 can comprise following substep:

Substep 221, the probability of compute histograms and each intensity level;

Substep 222 is set initial weighting coefficient ω _i(0) and class average μ _i(0);

Substep 223, all possible threshold value t=p1 of substitution ... p2 upgrades ω _iAnd μ _i, and compute classes internal variance or inter-class variance;

Substep 224 finds the threshold value corresponding with inter-class variance maximal value or class internal variance minimum value, as cutting apart used threshold value;

Substep 225, the threshold value that obtains according to substep 224, with pixel in the image greater than or split more than or equal to the point of this threshold value, obtain direct exposure area.

So far, finished the process of cutting apart.

Step 230, after from pretreatment image, being partitioned into direct exposure area quilt, extract outmost four points from the direct exposure area shown in Fig. 4 (a), namely on each direction up and down outmost point (from left to right four arrows point to respectively in Fig. 4 b: leftmost some left ₀, uppermost some top ₀, rightmost some right ₀With nethermost some bottom ₀), thereby form an initial boundary B as shown in Fig. 4 (c) ₀

Step 240 can also be used to generating feature figure through pretreated image.Extract outline line along four direction up and down from characteristic pattern, the eigenwert of outline line forms a series of point, detects the most violent points that descend from these some the insides and constitutes a rectangle.Then, extract the feature in the image, form characteristic pattern.

The difference in anatomical object 20 zones and prospect 30 zones is homogeneity, and in other words, there is more gradient information in the anatomical object zone than foreground area.Concerning anatomical object, necessarily there are many gradient informations in some directions, this direction may be vertical direction, horizontal direction, diagonal or other direction.

In the present embodiment, be that example illustrates how to find out this gradient information with the Kirsch operator.For each pixel of entire image, use eight different operation core to carry out convolution, find the maximal value in these eight convolution algorithm formulas then.This computing spread all over entire image from top to bottom, each pixel from left to right.

So, characteristic pattern is:

$I_{\mathrm{2,1}}(x,y)={\max }_{z=1,...,8}\underset{i=-1}{\overset{1}{\mathrm{\Σ}}}\underset{j=-1}{\overset{1}{\mathrm{\Σ}}}{g}_{\mathrm{ij}}^{\left(z\right)}\⊗I_1(x+i,y+j)$

Wherein, I _2,1(x y) is characteristic pattern, I ₁(x y) is pretreated image. Be operation core, have following eight kinds altogether:

$g^{\left(1\right)}=\left[\begin{array}{ccc}+5& +5& +5\\ -3& 0& -3\\ -3& -3& -3\end{array}\right]$

$g^{\left(2\right)}=\left[\begin{array}{ccc}+5& +5& -3\\ +5& 0& -3\\ -3& -3& -3\end{array}\right]$

$g^{\left(3\right)}=\left[\begin{array}{ccc}+5& -3& -3\\ +5& 0& -3\\ +5& -3& -3\end{array}\right]$

$g^{\left(4\right)}=\left[\begin{array}{ccc}-3& -3& -3\\ +5& 0& -3\\ +5& +5& -3\end{array}\right]$

$g^{\left(5\right)}=\left[\begin{array}{ccc}-3& -3& -3\\ -3& 0& -3\\ +5& +5& +5\end{array}\right]$

$g^{\left(6\right)}=\left[\begin{array}{ccc}-3& -3& -3\\ -3& 0& +5\\ -3& +5& +5\end{array}\right]$

$g^{\left(7\right)}=\left[\begin{array}{ccc}-3& -3& +5\\ -3& 0& +5\\ -3& -3& +5\end{array}\right]$

$g^{\left(8\right)}=\left[\begin{array}{ccc}-3& +5& +5\\ -3& 0& +5\\ -3& -3& -3\end{array}\right]$

Be that example has described how to extract feature with the Kirsch operator in the above example, yet those skilled in the art also can adopt other operators such as Canny operator, compass operator to extract feature, form characteristic pattern.

It should be noted that step 220 to step 230 and step 240, can carry out simultaneously, also can one in front and one in back carry out.Above numbering and order just clear and convenient for what describe, not in order to limit sequencing.

Step 250, according to characteristic pattern, the initial boundary that obtains from the step 230 final boundary B of looking next life _f, namely collimate the limit.

As previously mentioned, initial boundary B ₀Come from the result of cutting apart, based on characteristic pattern I _2,1To grow at four direction, when running into discrete point, make boundary B _fTo stop growing this boundary B _fBe exactly final border, the position on this final border of storage (being last rectangle) then.

For each point on the outline line among Fig. 5, pixel value is from the average on the horizontal or vertical direction of characteristic pattern.

Wherein, from left to right the pixel value on the direction is:

$P_{x={\mathrm{right}}_0,...\mathrm{iWidth}}=\frac{1}{\mathrm{iHeight}}\underset{y=1}{\overset{\mathrm{iHeight}}{\mathrm{\Σ}}}{I}_{\left(\mathrm{2,1}\right)}$

Pixel value on the direction from right to left is:

$P_{x={\mathrm{left}}_0,...1}=\frac{1}{\mathrm{iHeight}}\underset{y=1}{\overset{\mathrm{iHeight}}{\mathrm{\Σ}}}{I}_{\left(\mathrm{2,1}\right)}$

Pixel value on the direction is from top to bottom:

$P_{y={\mathrm{bottom}}_0,...,\mathrm{iHeight}}=\frac{1}{\mathrm{iWidth}}\underset{x=1}{\overset{\mathrm{iWidth}}{\mathrm{\Σ}}}{I}_{\left(\mathrm{2,1}\right)}$

Pixel value on the direction is from top to bottom:

$P_{y={\mathrm{iTop}}_0,...,1}=\frac{1}{\mathrm{iWidth}}\underset{x=1}{\overset{\mathrm{iWidth}}{\mathrm{\Σ}}}{I}_{\left(\mathrm{2,1}\right)}$

Wherein, I _(2,1)It is the pixel value of characteristic pattern.

Step 260 detects above-mentioned final border.

Before detecting four the most violent points of four direction decline, can carry out the noise reduction of one dimension earlier with level and smooth, for example carry out medium filtering at outline line.Medium filtering is a kind of non-linear, digital filtering technique, is commonly used to remove the data that peel off.

Given one group of array N that takes from the pixel intensity of outline line notes by abridging and is P _i, i=1 wherein, 2 ..., N, and P _iArrange according to the order that increases progressively, for example:

P _i＝{P ₁，P ₂，…，P _N}

Wherein, P _i＜P _i+ 1.

One group of sample sequence is carried out median operation, can be expressed as:

$\mathrm{median}\left(p\left(i\right)\right)={\mathrm{Rank}}^{\frac{N+1}{2}}\left(P\left(i\right)\right)$

Wherein, median () represents median operation, and Pi is $P_{{x=\mathrm{right}}_0,...\mathrm{iWidth}},$ $P_{{x=1,...,\mathrm{left}}_0},$ $P_{{y=\mathrm{bottom}}_0,...,\mathrm{iHeight}}$ Perhaps $P_{{y=1,...,\mathrm{iTop}}_0}.$ Central point.

Outline line behind the medium filtering carries out the derivation of one dimension derivative then, has detected peaked point, and these corresponding point in a series of outline lines are marked as detected position.

In Fig. 6, last boundary B f will form top by four points _Final, bottom _Final, left _Final, right _Final

Step 270 is with last boundary B _fBe mapped to original image I ₀Coordinate system in, and cutting obtains output image I from original image ₃, as shown in Figure 7.

According to another embodiment of the present invention, showed a kind of device 300 that in radioscopic image, detects the collimation limit as shown in Figure 8.

This device 300 comprises: a cutting unit 320, a feature extraction unit 330 and a border growing element 340.Wherein, described cutting unit 320 is used for image is cut apart, and obtains initial boundary.Described feature extraction unit 330 is used for extracting the feature of image, and forms characteristic pattern.Described border growing element 340 is used for according to characteristic pattern, obtains final border from described initial boundary growth.

Alternatively, this device 300 further comprises: a cutting unit 350 for detection of described final border, and cuts out part in the final border from described image.

Alternatively, this device 300 further comprises: a pretreatment unit 310, before cutting apart in the feature of extracting image with to image, described image is carried out pre-service, comprise image is carried out down-sampling and/or image is carried out Gauss's smoothing processing.

According to an embodiment more of the present invention, a kind of x-ray imaging equipment also is provided, comprise the device 300 on collimation limit in the above-mentioned detection radioscopic image.

From above-described embodiment as can be seen, the technology according to the embodiment of the invention can also be used in the scene that does not have such as available additional informations such as collimating apparatus blade position or collimating apparatus blade geometric shapes.Therefore this method work does not rely on hardware, is particularly suitable for the manual collimating apparatus of entry level system.And adopting more traditional method of the non-self-adapting of Hough conversion to have better stability than those aspect the processing picture noise.

The present invention relates to x-ray imaging art technical field, and disclose a kind of method on collimation limit, a kind of device and a kind of x-ray imaging equipment that detects the collimation limit of in radioscopic image, detecting.Wherein, the described method that detects the collimation limit in radioscopic image comprises: image is cut apart, and obtained initial boundary; Extract the feature in the image, form characteristic pattern; According to characteristic pattern, obtain final border from described initial boundary growth.The present invention has improved the stability that detects the collimation limit.

The above only is preferred embodiment of the present invention, and is in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of doing, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. one kind is detected the method that collimates the limit in radioscopic image, comprising:

Image is cut apart, and obtained initial boundary;

Extract the feature in the image, form characteristic pattern;

According to characteristic pattern, obtain final border from described initial boundary growth.

2. method according to claim 1 is characterized in that, this method further comprises: detect described final border; And, from described image, cut out the part in the final border.

3. method according to claim 1 is characterized in that, this method further comprises: the feature in extracting image and image cut apart before, image is carried out down-sampling, and/or, image is carried out Gauss's smoothing processing.

4. method according to claim 1 is characterized in that, comprises described cutting apart:

The histogram of computed image;

Select when the class internal variance be hour or the possible threshold value when inter-class variance is maximum;

With pixel in the described image less than or split smaller or equal to the point of this threshold value.

5. method according to claim 4 is characterized in that, selects to cut apart outermost starting point in the image of back, and forms described initial boundary with these starting points.

6. method according to claim 1 is characterized in that, the feature in the described extraction image comprises: calculate the gradient information in the described image, form outline line.

7. one kind is detected the device that collimates the limit in radioscopic image, comprising:

A cutting unit (320) is used for image is cut apart, and obtains initial boundary;

A feature extraction unit (330) for the feature of extracting image, forms characteristic pattern;

A border growing element (340) according to characteristic pattern, obtains final border from described initial boundary growth.

8. device according to claim 7 is characterized in that, this device further comprises: a cutting unit (350) for detection of described final border, and cuts out part in the final border from described image.

9. device according to claim 7 is characterized in that, this device further comprises: a pretreatment unit (310), carried out pre-service to described image before cutting apart in the feature of extracting image with to image.

10. an x-ray imaging equipment comprises the device according to collimation limit in the described detection radioscopic image of claim 7-9.

Images