JP6002324B2 - Radiation image generating apparatus and image processing method - Google Patents

Description

  The present invention relates to a technique for reducing noise in a radiographic image.

  In the X-ray diagnostic imaging apparatus, when a fluoroscopic image is obtained, the signal component becomes minute with the low dose of X-rays, and noise is easily manifested as compared with imaging. As this type of noise removal method, there is a case where recursive filter processing is used in which an image obtained by weighting an image one frame before the current image is added to the current image. However, when there is a moving object such as a guide wire in the X-ray irradiation field, the movement of the guide wire or the like may appear as a blurred image by this recursive filter processing.

  With regard to the technology, Patent Document 1 describes, “In the image processing apparatus according to the present invention, recursive filter means and means for controlling the recursive filter coefficient for each pixel from the image data of the current frame according to the average luminance in the vicinity of the pixel. And the means for spatially filtering the image data of the current frame, and using the spatial filter processing data according to the motion component for each pixel obtained from the relationship between the image data of the current frame and the image data of the previous frame. The technology of “having a means” is disclosed.

JP-A-6-47036

  In the technique disclosed in Patent Document 1, a local motion component is detected for each pixel and used for processing. However, radiographic images such as X-ray fluoroscopic images are noisy. For such noisy images, local motion component detection is likely to be erroneously detected, and it is not easy to properly use the recursive filter and the spatial filter, and the desired effect cannot be obtained sufficiently. was there.

  One embodiment for solving the above-described problem is, for example, that a plurality of images having different imaging times are input with respect to a fluoroscopic image captured using radiation, and the temporally preceding image among the plurality of input images. And motion detection using the later image to detect one motion information for one image, and temporally change the position of the previous image by the amount of motion information. The difference from the subsequent image is calculated for each local region, and the region for performing noise reduction processing in the time direction and the region for performing noise reduction processing in the spatial direction are determined based on the difference for each local region. , Perform at least noise reduction processing in the time direction, do not perform spatial noise reduction processing, perform spatial noise reduction processing, do not perform temporal noise reduction processing, temporal noise reduction processing and space Direction Area for both the noise reduction processing, to set the three regions may be configured to perform noise reduction processing different for each region.

  According to the present invention, it is possible to more suitably reduce noise in a radiographic image.

It is a block diagram of an example of the radiographic image diagnostic apparatus of 1st embodiment. It is a block diagram explaining the process of the image process part which concerns on 1st embodiment. It is explanatory drawing for demonstrating the example of the similarity histogram display of 1st embodiment. (A)-(c) is explanatory drawing for demonstrating an example of the area | region where the spatial direction noise reduction process and time direction noise reduction process of 1st embodiment are applied. (A) And (b) is explanatory drawing for demonstrating the example of a similarity histogram when an image | video is still and the image | video is moving and deform | transforming, respectively, of 1st embodiment. . It is a block diagram explaining the process of the image process part which concerns on 2nd embodiment. It is explanatory drawing for demonstrating the example of a setting of the 1st LUT of 2nd embodiment. It is explanatory drawing for demonstrating the example of a setting of the 2nd LUT of 2nd embodiment. It is explanatory drawing for demonstrating the example of a display of a similarity histogram and LUT of 2nd embodiment. (A)-(c) is explanatory drawing for demonstrating an example of the area | region where the spatial direction noise reduction process and time direction noise reduction process of 2nd embodiment are applied. It is a flowchart of the image processing of 1st embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The radiographic image diagnostic apparatus (also referred to as a radiographic image generation apparatus, a radiographic image generation apparatus, or a radiographic image generation apparatus) according to an embodiment of the present invention may use radiation other than X-rays. Will be described using an X-ray image diagnostic apparatus as an example.

<< First Embodiment >>
FIG. 1 is a schematic diagram showing the configuration of the X-ray image diagnostic apparatus according to the present embodiment.
The X-ray diagnostic imaging apparatus 1 according to the present embodiment includes an X-ray tube 2 that generates and irradiates X-rays, a high voltage generator 4 that is electrically connected to the X-ray tube 2, and a high voltage generator 4. Are electrically connected to the X-ray controller 3, the diaphragm 5 disposed in the X-ray irradiation direction of the X-ray tube 2, the X-ray compensation filter 6, the X-ray compensation filter 6 and the diaphragm 5. A diaphragm / filter controller 7, a table 8, an X-ray plane detector 9 disposed opposite to the X-ray tube 2 through the diaphragm 5, the X-ray compensation filter 6 and the table 8, and an X-ray plane An image processing unit 10 electrically connected to the detector 9, a display output unit 11 electrically connected to the image processing unit 10, and a mechanism electrically connected to the table 8 and the X-ray flat panel detector 9. Control unit 12, X-ray control unit 3, high voltage generation unit 4, diaphragm / filter control unit 7, X-ray flat panel detector 9, image Having a processing unit 10 and the mechanism control section 12 and the central processing unit 13 which is electrically connected, the.

  Here, the high voltage generator 4 generates a high voltage to be applied to the X-ray tube 2. The X-ray tube 2 irradiates the subject with X-rays. The X-ray controller 3 controls the high voltage generator 4 and controls the quality of X-rays emitted from the X-ray tube 2. The diaphragm 5 controls an area irradiated with X-rays generated by the X-ray tube 2 by opening and closing a metal having a high X-ray absorption rate. The X-ray compensation filter 6 is made of a substance having a high X-ray absorption rate, and reduces halation by attenuating X-rays that reach a portion of the subject having a low X-ray absorption rate. Table 8 is a bed on which a subject is placed.

  The X-ray flat detector 9 outputs image data corresponding to the intensity distribution of X-rays irradiated from the X-ray tube 2 and transmitted through the subject. The X-ray plane detector 9 can also generate the image data as a still image. In this case, X-ray imaging still image data is obtained. It is also possible to generate a plurality of image data captured at different timings and generate a moving image. In this case, X-ray imaging moving image data is obtained.

  The image processing unit 10 performs a correction process on the image data output from the X-ray plane detector 9. The display output unit 11 displays the X-ray captured image data after the correction process. The mechanism control unit 12 controls the table 8 and the X-ray flat detector 9 to move to a position suitable for fluoroscopy or imaging of the subject. The central processing unit 13 is a computer that controls the electrically connected components.

  The X-ray tube 2, the X-ray control unit 3, and the high voltage generation unit 4 constitute a radiation irradiation unit that irradiates the subject with radiation. Further, the X-ray flat panel detector 9 constitutes a detection unit that detects radiation irradiated by the radiation irradiation unit and generates a radiation image. The image processing unit 10 performs image processing on the image generated by the detection unit.

  The noise removal processing in the present invention is performed by a program executed by the central processing unit 13 in the image processing unit 10. That is, the central processing unit 13 includes a CPU, a memory, and a storage device. Each function of the central processing unit 13 is realized by the CPU loading a program stored in advance in the storage device into the memory and executing it. All or some of the functions may be realized by hardware such as ASIC (Application Specific Integrated Circuit) or FPGA (field-programmable gate array), but may be realized by GPU (Graphics Processing Unit). May be. Various data used for processing of each function and various data generated during the processing are stored in the storage device.

  FIG. 2 is a processing block diagram of noise removal processing executed by the image processing unit 10. As shown in FIG. 2, the image processing unit 10 includes a frame memory 200 that stores input images, and a motion detection unit that performs motion detection using a plurality of temporally different input images read from the frame memory 200. 201 and a similarity calculation unit 202 that calculates a similarity between images for each local region while performing motion correction on a plurality of images read from the frame memory 200 using motion information detected by the motion detection unit 201. The first region determining unit 203 that determines the first region from the similarity calculated by the similarity calculating unit 202 and the second region determining unit that determines the second region from the similarity calculated by the similarity calculating unit 202 205 and time direction noise for performing noise reduction processing in the time direction on the first region determined by the first region determination unit 203 in the plurality of input images read from the frame memory 200 Similar to the spatial noise reduction unit 206 that performs the spatial noise reduction process on the second region determined by the second region determination unit 205 in the image in which the noise reduction unit 204 and the temporal direction noise reduction unit 204 reduce noise. A similarity histogram calculation unit 207 that calculates a histogram of the similarity calculated by the degree calculation unit 202.

  The frame memory 200 stores each frame of the input image in a form that retains the temporal context, and outputs it according to a request from another block described later. The number of input images to be stored may be as many as the number of images used in the time direction noise reduction unit 204 described later.

  The motion detection unit 201 receives a plurality of images stored in the frame memory 200, performs motion detection using the plurality of images, detects one motion information per sheet, and outputs it. That is, a global motion in the image is detected and output as motion information of the image. More specifically, the motion detection unit 201 moves a predetermined region of a temporally subsequent image out of the two images at a predetermined interval such as a pixel unit, and performs SAD (Sum of Absolute) between the images. Difference) is calculated.

Since this process is intended to detect a global movement in the image, it is desirable that the predetermined area is, for example, a range of 50% or more of the image size. The region is assumed to be a region larger than the local region to be processed by the similarity calculation unit 202. Formula (1) is a calculation formula for SAD, where d is the direction of motion, I _a and I _b are input images acquired at different timings, and Ω is a set of coordinates in the SAD calculation range. Since the motion detection unit 201 detects global motion, a set of coordinates of the entire image is designated for Ω.

Here, it is considered that the input images I _a and I _b are more similar as the value of SAD calculated by Expression (1) is smaller. In other words, the direction SAD is the smallest within the search range, the motion direction between the input image I _a and I _b relative to the image I _a. Equation (2), the input image I _a input image relative to the I _a and I _b between the motion information V _a, a calculation formula of _b, [psi denotes the set of coordinates of the search range. The search range typically specifies the coordinates of a rectangular area of about 20 horizontal pixels and 20 vertical pixels. However, an area determined in consideration of the speed of the maximum movement of the subject assumed in the input image and the speed of calculating Formula (1) and Formula (2) may be specified.

  The similarity calculation unit 202 receives the two input images stored in the frame memory 200 and the motion information calculated by the motion detection unit 201 as input. After the position of the input image is shifted by the amount of the motion information, the similarity between the images is calculated and output for each local region.

More specifically, the SAD of the local area is calculated at each coordinate, and this is used as the similarity. Therefore, the smaller the similarity is, the more similar the images are. Formula (3) is a calculation formula for the similarity S _{a, b} at the coordinate x between the input images I _a and I _b , Ω _x is a set of local area coordinates at the coordinate x, and V _{a, b} is motion detection. This is motion information between the images I _a and I _b calculated by the unit 201. Ω _x typically designates a set of coordinates of 7 horizontal pixels and 7 vertical pixels centered on the coordinate x. However, the range and shape may be changed in consideration of the fineness of the pattern of the object in the image, the noise level of the image, the calculation speed, and the like.

  The first region determination unit 203 and the second region determination unit 205 are based on the similarity calculated by the similarity calculation unit 202, and perform a noise reduction process in the time direction on the input image and a noise reduction in the spatial direction. An area determination unit that performs area determination to determine an area to be processed. Here, (1) an area where noise reduction processing in the time direction is performed and noise reduction processing in the spatial direction is not performed, and (2) noise reduction processing in the spatial direction is performed and noise reduction processing in the time direction is performed on the input image. And (3) at least two of the three areas that perform both the noise reduction process in the time direction and the noise reduction process in the spatial direction.

The first region determination unit 203 performs threshold processing using the first threshold on the similarity S _{a, b} calculated by the similarity calculation unit 202 and determines the first region classification. The first threshold is set in advance in view of the noise level of the image. Formula (4) is a calculation formula for the first region segment R ¹ _{a, b at} the coordinate x, where S _{a, b} is the similarity and T ¹ is the first threshold value.

A set of coordinates in which the first region segment R ¹ _{a, b} is “1” represents a region having a low similarity (images are similar). In other words, this is a region close to a still state when motion compensation is performed on the input images I _a and I _b using the motion information V _{a, b} detected by the motion detection unit 201. That is, the first threshold T ¹ is set so that a region close to a stationary state is extracted.

The second region determination unit 205 performs threshold processing using the first threshold on the similarity calculated by the similarity calculation unit 202 and determines the second region classification. The second threshold is set in advance in view of the noise level of the image. Formula (5) is a formula for calculating the ^second region segment R ² _{a, b at} the coordinate x, where S _{a, b} is the similarity and T ² is the second threshold value.

A set of coordinates in which the second region segment R ² _{a, b} is “1” represents a region having a high degree of similarity (images are not similar). That is, this is a region where the motion information V _{a, b} detected by the motion detection unit 201 is used for motion or deformation when the input images I _a and I _b are compensated for motion. That is, the second threshold value T ² is set so that a region that moves or deforms is extracted.

  The time direction noise reduction unit 204 and the spatial direction noise reduction unit 206 are noise reduction processing units that perform noise reduction processing on each region according to the determination results of the first region determination unit 203 and the second region determination unit 205.

  The time direction noise reduction unit 204 inputs a plurality of input images stored in the frame memory, the motion information calculated by the motion detection unit 201, and the first region classification determined by the first region determination unit 203. While performing motion compensation on a plurality of input images using motion information, noise reduction is performed by calculating a weighted average between frames of the input image for a region where the first region segment is 1. Outputs an image with reduced noise.

Formula (6) is a formula for calculating the image I ^TIME in which the time-direction noise reduction at the coordinate x is performed, where I _n (1 ≦ n ≦ A) is the input image and c _n (1 ≦ n ≦ A) is the frame. The weight coefficient for each frame, V _{1, n} indicates the motion between the latest frame and each frame, and R ¹ _{1, n} indicates the similarity between the latest frame and each frame.

Here, if the input images I ₁ to I _A are temporally continuous frames, motion information V _1,2 , V _2,3 ,..., V _{(A -1) When A} is calculated, V _{1, n} is obtained by accumulating them. Equation (7) is a calculation formula for the motion information V _{1, n} between the input image I ₁ and I _n.

Also, the weighting coefficient c _n is set so that the sum is 1. Equation (8) is a constraint of the weighting coefficients c _n.

  As described above, the time-direction noise reduction unit 204 according to the present embodiment performs frame compensation on the input image while performing motion compensation on the input images using the global motion information detected by the motion detection unit 201. Reduce noise by a weighted average. Therefore, it is possible to reduce noise more suitably than in the case of performing weighted averaging in the time direction without using motion information.

  The spatial direction noise reduction unit 206 includes an image obtained by performing temporal direction noise removal on the region in which the first region division is 1 by the time direction noise reduction unit 204 and the second region determination unit 205 determines the second region. Is input to the region where the second region segment is 1 in the input image, noise in the spatial direction is reduced, and an image with reduced noise is output.

  As the noise reduction method in the spatial direction, a method is employed in which blurring due to the motion of an object that may occur in the noise reduction processing in the time direction is less likely to occur. That is, if a method of removing noise using only one input image is used, this blur does not occur in principle. For example, methods such as a Gaussian filter, a bilateral filter, a non-local means, and a wavelet degeneration are suitable. However, other methods such as using a plurality of frames may be used as long as the noise reduction method is resistant to blur caused by the motion of an object.

As an example, an example in which noise in the spatial direction is reduced using a Gaussian filter will be described below. The noise reduction in the spatial direction is applied to an area where the second area division is 1 in the image subjected to the noise reduction process in the time direction. Formula (9) is a calculation formula for generating an image I ^SPACE in which the noise in the spatial direction is reduced at the coordinate x, I ^TIME is an image from which noise in the time direction has been removed, and R ² _{1 and 2} are input. A second region segment between the images I ₁ and I ₂ , α is the size of the Gaussian kernel (a parameter for controlling the strength of noise removal), and Σ is a set of coordinates in a range to which the Gaussian kernel is applied. The Gaussian kernel size α and the Gaussian kernel application range Σ are determined in view of the noise level of the image and the calculation speed of Equation (9).

  The similarity histogram calculation unit 207 calculates the frequency distribution of the similarity calculated by the similarity calculation unit 202 and outputs it as a similarity histogram. The number of histogram bins is about the number of gradations of the input image. However, it may be changed in view of the size (number of pixels) of a histogram window, which will be described later, and the setting accuracy of the first (second) threshold value.

  The display output unit 11 visualizes and displays the similarity histogram and the first threshold value or the second threshold value in addition to the image from which noise has been removed.

  FIG. 3 is a display example when the similarity histogram is visualized.

  In the window 300, a vertical line 302 representing the first threshold value or the second threshold value is superimposed on the bar graph 301 representing the similarity histogram.

  The bar graph 301 represents a similarity histogram with the horizontal axis representing similarity and the vertical axis representing frequency. A vertical line 302 represents the first (second) threshold value in accordance with the position of the horizontal axis (similarity) of the bar graph 301.

  With the above configuration, the image processing unit 10 of the present embodiment performs (1) noise reduction processing in the time direction on the image generated by the detection unit by adjusting the first threshold value and the second threshold value, and the spatial direction (2) Area where noise reduction processing in the spatial direction is performed and noise reduction processing in the time direction is not performed, (3) Both noise reduction processing in the time direction and noise reduction processing in the spatial direction It is possible to set three areas for performing.

  FIG. 4A to FIG. 4C are schematic diagrams illustrating the first region segment and the second region segment, and regions to which temporal direction noise removal and spatial direction noise removal are applied. For example, as shown in the figure, the first area segment 400 is detected as “0” near the center and “1”, and the second area segment 401 is detected as “1” near the center and “0” elsewhere. It shall be assumed.

  At this time, the range 402 to which each noise reduction process is applied is as follows. The area 411 in which the area detected as “0” in the first area section 400 and the area detected as “1” in the second area section 401 overlap is applied with the spatial noise reduction processing, and the time direction Noise reduction processing is not applied. In the region 413 where the region detected as “1” in the first region segment 400 and the region detected as “0” in the second region segment 401 overlap, temporal direction noise reduction is applied, and spatial direction noise is applied. Reduction processing is not applied. A region 412 in which the region detected as “1” in the first region segment 400 and the region detected as “1” in the second region segment 401 overlap is a time direction noise reduction process and a spatial direction noise reduction process. Applies.

  That is, in the processing according to the present embodiment, a region where noise reduction in the time direction is effective (such as a stationary region) or a region where noise reduction in the spatial direction is effective (such as a region that is moving or deforming). ) And areas where it is effective to apply both temporal and spatial noise reduction (such as areas that are slightly deformed) are selected appropriately by the first threshold and the second threshold. Therefore, it is possible to apply an appropriate noise reduction process to each region.

  Further, the first threshold value and the second threshold value may be set so as to be more visually understandable. FIG. 5A and FIG. 5B illustrate a setting state in this case, and are a part (window) of a setting screen output to the display output unit 11 of the X-ray image diagnostic apparatus 1. The setting screen may be superimposed on the display of the X-ray captured image, or may be displayed side by side. Alternatively, the X-ray captured image may be displayed alone without being displayed.

  For example, the window 500 is an example of a histogram when a still image is input. At this time, the similarity histogram 501 is expected to be a type in which the similarity is concentrated in a small area. This mountain represents a noise component included in the video. If the image is stationary and no noise is present, all the similarities are zero. In other words, if time direction noise reduction is applied to this mountain, noise can be effectively removed. Therefore, the user who sets the X-ray image diagnostic apparatus 1 may set the first threshold 502 near the right end of the mountain while viewing the window 500.

  In this case, the X-ray image diagnostic apparatus 1 of the present embodiment further includes an operation input unit 14 as shown in FIG. The setting is performed via the operation input unit 14 and the like. That is, in the X-ray image diagnostic apparatus 1 according to the present embodiment, the setting change of the process of the image processing unit 10 from the user can be input while presenting the histogram from the display output unit 11.

  A window 510 is an example of a histogram when a moving / deforming video is input. The similarity histogram 511 is expected to be concentrated even in a region where the similarity is large. This mountain indicates a region where the similarity has increased due to the movement and deformation of the image. In other words, if spatial noise reduction is applied to this mountain, noise can be effectively removed. Therefore, the user may set the second threshold 512 near the left end of the mountain while viewing the window 510.

As described above, the radiological image diagnostic apparatus 1 according to the present embodiment includes a radiation irradiation unit that irradiates radiation, a detection unit that detects radiation irradiated by the radiation irradiation unit, and generates an image, and the detection unit. An image processing unit 10 that performs image processing on the image generated in Step 1, wherein the image processing unit 10 is a temporally previous image among a plurality of images generated by the detection unit and having different imaging times. And a motion detection unit 201 that detects one piece of motion information for one image by performing motion detection using a temporally subsequent image and the temporally previous image by the amount of the motion information. A similarity calculation unit 202 that calculates the difference between the image that has been changed and the temporally subsequent image as a similarity for each local region, and on the image based on the similarity, the time direction Noise reduction processing area and spatial direction Of comprising a region determination unit that performs area determination for determining an area to perform noise reduction processing according to the region determination result of the area determination unit, and a noise reduction unit which performs the noise reduction process to each region.
The region determination unit determines a region for performing noise reduction processing in the time direction and a region for performing noise reduction processing in the spatial direction by performing threshold processing on the similarity.
The area for performing the noise reduction process in the time direction and the area for performing the noise reduction process in the spatial direction may partially or entirely overlap.

  According to the radiological image diagnostic apparatus 1 and the image processing method thereof according to the present embodiment described above, it is more preferable to use the time direction noise reduction process and the spatial direction noise reduction process, or a region in which both are performed. Therefore, it is possible to achieve both noise reduction and suppression of blurring more suitably. Thereby, it is possible to realize a more preferable noise reduction of the radiographic image.

  In the first embodiment described above, only one piece of global motion information is detected by detecting the global motion in the previous stage of performing the noise reduction processing. However, this embodiment does not prevent the radiological image diagnostic apparatus 1 from performing a process in which a local motion is detected before and after the above-described process. In addition, the present embodiment does not prevent the radiological image diagnosis apparatus 1 from having a mode for performing another noise reduction process using local motion detection.

<< Second Embodiment >>
The X-ray image diagnostic apparatus of the present embodiment has basically the same configuration as the X-ray image diagnostic apparatus 1 of the first embodiment. However, the processing of the image processing unit 10 is different. Hereinafter, the present embodiment will be described focusing on the configuration different from the first embodiment.

  FIG. 6 is an example of a processing block diagram illustrating the image processing unit 10 according to the second embodiment. The image processing unit 10 according to the second embodiment performs conversion processing using a LUT (Look Up Table) on the similarity calculated by the similarity calculation unit 202 in the region determination of the first region and the second region. Do. The LUT may be stored in advance by each of the first region determination unit 600 or the second region determination unit 601 of the image processing unit 10 or may be stored in a storage unit (not shown) in the image processing unit 10. Good.

In the image processing unit 10 in FIG. 6, the description of the components having the same functions as those in FIG. The first region determination unit 600 performs conversion using the first LUT on the similarity calculated by the similarity calculation unit 202 and determines the first region. The first LUT is set in advance in view of the noise level of the image. Formula (10) is a calculation formula for the first region R ¹ _{a, b at} the coordinate x, S _{a, b} is the similarity, and L ¹ is the conversion by the first LUT.

In the first embodiment, the first region R ¹ _{a, b} takes a binary value, but in this embodiment takes a multi-value (a real number) from 0 to 1. The first LUT specifies a shape having a negative slope as a whole. FIG. 7 is an example of the first LUT. The LUT setting method is arbitrary, and it may be a polygonal line approximation by n-point designation used for general LUT setting.

If it is set more simply, it is preferable to create an LUT shape using a sigmoid function. Expression (11) is an expression indicating the shape of L ¹ which is the first LUT, x is an input value of the LUT, and β ₁ and β ₂ are parameters for controlling the shape of the LUT.

The second area determination unit 601 performs conversion using a second LUT (Look Up Table) on the similarity calculated by the similarity calculation unit 202 to determine a second area. The second LUT is set in advance in consideration of the noise level of the image. Formula (12) is a calculation formula for the second region R ² _{a, b at} the coordinate x, S _{a, b} is the similarity, and L ² is the conversion by the second LUT.

In the first embodiment, the second region R ² _{a, b} takes a binary value, but in the present embodiment takes a multivalue (real number) from 0 to 1. The second LUT designates a shape having a positive slope as a whole. FIG. 8 is an example of the second LUT. The LUT setting method is arbitrary, and it may be a polygonal line approximation by n-point designation used for general LUT setting.

If it is set more simply, it is preferable to create an LUT shape using a sigmoid function. Expression (13) is an expression showing the shape of L ² as the second LUT, x is an input value of the LUT, and β ₁ and β ₂ are parameters for controlling the shape of the LUT.

  The display output unit 11 displays an image from which noise has been removed. Instead of or in addition to this display, the similarity histogram and the first LUT or the second LUT may be visualized and displayed.

  FIG. 9 is a display example when the similarity histogram and the first LUT are visualized.

In the window 900, a bar graph 901 expressing a similarity histogram and a curve 902 expressing the first LUT or the second LUT are arranged.
As described above, by using the LUT, the first area section and the second area section can be expressed more finely based on the similarity. An example of the first area section and the second area section in the present embodiment is shown in FIGS. 10 (a) to 10 (c).

  FIG. 10A to FIG. 10C are schematic diagrams illustrating the first region segment and the second region segment, and regions to which temporal direction noise removal and spatial direction noise removal are applied. For example, as shown in the figure, the first area section 1000 detects “0” in the vicinity of the center and “1” in other areas, and the second area section 1001 detects “1” in the vicinity of the center and “0” in other areas. It shall be assumed.

  At this time, the range 1002 to which each noise reduction process is applied is as follows. The area 1011 in which the area detected as “0” in the first area section 1000 and the area detected as “1” in the second area section 1001 are overlapped, the spatial direction noise reduction processing is applied, and the time direction Noise reduction processing is not applied. The area 1013 in which the area detected as “1” in the first area section 1000 and the area detected as “0” in the second area section 1001 overlap is applied with time direction noise reduction, and spatial direction noise is applied. Reduction processing is not applied. A region 1012 in which a region detected as “1” in the first region segment 1000 and a region detected as “1” in the second region segment 1001 overlap is a time direction noise reduction process and a spatial direction noise reduction process. Applies. In addition, noise reduction is not performed in a region 1014 in which the region detected as “0” in the first region segment 1000 and the region detected as “0” in the second region segment 1001 overlap.

  As in the example shown in FIGS. 10A to 10C, in the processing of this embodiment, the first area section and the second area section may be set more finely than in the first embodiment. For example, (1) a region where noise reduction processing in the time direction is performed and noise reduction processing in the spatial direction is not performed, and (2) a region where noise reduction processing in the spatial direction is performed and noise reduction processing in the spatial direction is not performed , (3) four areas: a region where both noise reduction processing in the time direction and noise reduction processing in the spatial direction are performed, and (4) a region where both noise reduction processing in the time direction and noise reduction processing in the spatial direction are not performed It becomes possible to set.

As described above, the radiation image diagnostic apparatus 1 according to the present embodiment includes the radiation irradiation unit, the detection unit, and the image processing unit 10 as in the first embodiment. The region determination unit uses a lookup table stored in advance in the radiation image generation device 1 for the similarity, so that a region for performing noise reduction processing in the time direction and a region for performing noise reduction processing in the spatial direction are used. And decide.
The area for performing the noise reduction process in the time direction and the area for performing the noise reduction process in the spatial direction may partially or entirely overlap.

  With the above configuration, the first area and the second area are expressed in multiple values, and the values can be finely adjusted by the first LUT and the second LUT. Therefore, by setting the range and intensity to which the time direction noise reduction and the spatial direction noise reduction are applied independently, it becomes possible to more effectively reduce the noise of the fluoroscopic image.

  Further, as described in the first embodiment, it is appropriate to set the first threshold value and the second threshold value while observing the similarity histogram. By superimposing the first LUT and the second LUT on the similarity histogram, the user can visually change the shape of each LUT, and can select an appropriate noise reduction process for the image. It becomes.

  Finally, the flow of image processing by the image processing unit 10 in each of the above embodiments will be described. FIG. 11 is a processing flow of image processing of each of the above embodiments. Here, the configuration of the first embodiment will be described as an example.

  The image processing unit 10 receives an input of a plurality of images having different imaging times for an image captured using radiation (step S1101).

  The motion detection unit 201 performs motion detection using a temporally preceding image and a temporally subsequent image among a plurality of images received by the image processing unit 10, and performs one motion for one image. Information is detected (step S1102).

  The similarity calculation unit 202 calculates the difference between the temporally previous image whose position is changed by the amount of the motion information and the temporally subsequent image as the similarity for each local region (step). S1103).

  The first region determination unit 203 and the second region determination unit 205 determine the first region and the second region of the captured image based on the similarity (step S1104). Then, using the information on the first region and the second region, the input image is subjected to (1) a noise reduction process in the time direction and no noise reduction process in the spatial direction, and (2) a spatial direction. At least two of the three areas: the area that performs noise reduction processing and does not perform noise reduction processing in the time direction, and (3) the area that performs both noise reduction processing in the time direction and noise reduction processing in the spatial direction. (Step S1105).

  Specifically, as illustrated in FIGS. 4A to 4C, the region determination unit applies a time direction noise reduction process to the first region determined to have a low similarity. Is determined. In addition, the spatial direction noise reduction process is applied to the second region determined to have a high degree of similarity. A region determined as both the first and second regions is determined as a region to which both the time direction noise reduction processing and the spatial direction noise reduction processing are applied.

  Then, the time direction noise reduction unit 204 and the spatial direction noise reduction unit 206 perform noise reduction processing on each region according to the determinations of the first region determination unit 203 and the second region determination unit 205 (step S1106).

  Here, first, the time direction noise reduction unit 204 performs the time direction noise reduction processing on the first region with respect to the input image, and then the space direction noise reduction unit 206 performs the second region of the processed image. On the other hand, a spatial noise reduction process is performed.

  Note that the image processing flow of the second embodiment is basically the same. However, after determining the first region and the second region in step S1104, the first region determination unit 600 and the second region determination unit 601 perform (1) time-direction noise reduction processing on the input image to determine the space. An area where noise reduction processing in the direction is not performed, (2) An area where noise reduction processing in the spatial direction is performed and noise reduction processing in the time direction is not performed, and (3) Noise reduction processing in the time direction and noise reduction processing in the spatial direction Are divided into four areas, and (4) an area where both the noise reduction process in the time direction and the noise reduction process in the spatial direction are not performed.

  Note that it is possible to take two or more still images and reduce the noise of the still image using the method described in the first embodiment or the second embodiment. In general, when a still image is taken, the exposure dose increases because a high X-ray dose is used to obtain an image with less noise. According to the method of each of the above embodiments, a still image with less noise while reducing exposure than by shooting a single image with a high dose by capturing two or more still images with a lower dose of exposure. Can be obtained.

  The embodiments of the present invention are not limited to the above-described embodiments, and various additions and changes can be made without departing from the spirit of the invention. Further, the image processing unit 10 and the display output unit 11 may not be included in the X-ray image diagnostic apparatus 1. For example, it may be constructed on an independent information processing apparatus capable of transmitting / receiving data to / from the X-ray image diagnostic apparatus 1.

  1: X-ray diagnostic imaging apparatus, 2: X-ray tube, 3: X-ray controller, 4: High voltage generator, 5: Aperture, 6: X-ray compensation filter, 7: Aperture / filter controller, 8: Table 9: X-ray plane detector, 10: Image processing unit, 11: Display output unit, 12: Mechanism control unit, 13: Central processing unit, 14: Operation input unit, 200: Frame memory, 201: Motion detection unit, 202: similarity calculation unit, 203: first region determination unit, 204: time direction noise reduction unit, 205: second region determination unit, 206: spatial direction noise reduction unit, 207: similarity histogram calculation unit, 300: window , 301: similarity histogram, 302: vertical line, 400: first region section, 401: second region section, 402: noise reduction processing application range, 411: spatial direction noise reduction processing application region, 412: time direction And space Directional noise reduction processing application region, 413: Time direction noise reduction processing application region, 500: Window, 501: Similarity histogram, 502: First threshold, 510: Window, 511: Similarity histogram, 512: Second threshold , 600: first region determination unit, 601: second region determination unit, 900: window, 901: similarity histogram, 902: LUT, 1000: first region segment, 1001: second region segment, 1002: noise Reduction processing application range, 1011: Spatial direction noise reduction processing application region, 1012: Time direction and spatial direction noise reduction processing application region, 1013: Time direction noise reduction processing application region, 1014: Noise reduction processing non-application region

Claims (16)

A radiation irradiator for irradiating radiation;
A detection unit that detects radiation irradiated by the radiation irradiation unit and generates an image;
An image processing unit that performs image processing on the image generated by the detection unit,
The image processing unit
Among a plurality of images having different imaging times generated by the detection unit, motion detection is performed using a temporally previous image and a temporally subsequent image, and one motion information is detected for one image. A motion detector to perform,
A similarity calculator that calculates a difference between the temporally previous image and the image whose position has been changed by the amount of motion information and the temporally subsequent image as a similarity for each local region;
An area determination unit that performs area determination on the image based on the similarity and determines an area for performing noise reduction processing in the time direction and an area for performing noise reduction processing in the spatial direction;
A radiological image generation apparatus comprising: a noise reduction processing unit that performs the noise reduction processing on each region in accordance with the region determination result of the region determination unit.   The radiological image generation apparatus according to claim 1, wherein the region determination unit performs the region determination by performing threshold processing on the similarity.   The region determination unit includes a region for performing noise reduction processing in the time direction and a region for performing noise reduction processing in the spatial direction by using a look-up table for the similarity stored in advance in the radiation image generation device. The radiation image generating apparatus according to claim 1, wherein:   The radiographic image generation apparatus according to claim 1, wherein a part of the area for performing the noise reduction process in the time direction and a part of the area for performing the noise reduction process in the spatial direction overlap.   The radiation according to claim 1, wherein the motion detection unit performs the motion detection based on a result of a difference calculation process for a range of 50% or more of a size of an image generated by the detection unit. Image generation device. A display unit;
The radiographic image generation apparatus according to claim 1, wherein the display unit displays an image subjected to image processing by the image processing unit.   The said display part displays the distribution of the said similarity calculated by the said similarity calculation part instead of the image which performed the image process in the said image process part, or the said image. Radiation image generation device. It further includes an operation input unit,
The operation input unit receives an input of a setting change operation for processing of the image processing unit while the display unit displays the distribution of the similarity. Radiation image generation device. An image processing method for a fluoroscopic image,
An input step of inputting a plurality of images having different imaging times for an image captured using radiation,
Motion information detection for detecting one piece of motion information for one image by performing motion detection using a temporally previous image and a temporally subsequent image among the plurality of images input in the input step. Steps,
A similarity calculation step of calculating the difference between the image that has been changed in position by the amount of the motion information and the image that has been changed in time as the similarity for each local region;
A region determination step for performing region determination for determining a region for performing noise reduction processing in the time direction and a region for performing noise reduction processing in the spatial direction on the image based on the similarity,
According to the region determination result of the region determination step, a noise reduction processing step for performing the noise reduction processing on each region;
An image processing method comprising:   The image processing method according to claim 9, wherein in the region determination step, the region determination is performed by performing threshold processing on the similarity.   In the region determining step, a region for performing noise reduction processing in the time direction and a region for performing noise reduction processing in the spatial direction are determined by using a lookup table for the similarity stored in advance. The image processing method according to claim 9.   10. The image processing method according to claim 9, wherein a part of the area for performing the noise reduction process in the time direction and a part of the area for performing the noise reduction process in the spatial direction overlap partly.   The motion detection is performed in the motion information detection step based on a result of a difference calculation process for a range of 50% or more of the size of the image input in the input step. Image processing method.   The image processing method according to claim 9, further comprising a display step of displaying the image data that has undergone the noise reduction processing step.   The distribution of the similarity calculated in the similarity calculation step can be displayed in the display step, instead of or together with the image data that has undergone the noise reduction processing step. An image processing method described in 1.   The operation input step of receiving an input of a setting change operation for the processing of the region setting step while displaying the distribution of the similarity in the display step. Image processing method.

Images