JP6108575B2 - Image processing apparatus and X-ray imaging apparatus - Google Patents

Description

  The present invention relates to an image processing apparatus that processes an image captured by a general camera or an image captured using radiation, and in particular, an image processing apparatus that processes an image acquired by general medical or dental imaging, and The present invention relates to an X-ray imaging apparatus.

  Conventionally, various techniques have been researched and developed for obtaining images with a high resolution, that is, images that are easy to view or images with good contrast, in images taken with a general camera or images taken with radiation. It was. In general, in an image detector that detects X-rays such as a flat panel detector (FPD), the smaller the size of the pixel, that is, the larger the number of pixels per unit area, the obtained image. The resolution becomes higher. One method for obtaining a more easily viewable image is to reduce the pixel size of the image detector. However, an image detector with a small pixel size is expensive. Further, if the pixel size of the image detector is reduced, the light receiving sensitivity is lowered. Therefore, there is a limit to reducing the pixel size for use in medical X-ray imaging due to the problem of exposure dose.

  Another method is to increase the resolving power by image processing, in other words, to increase the resolving limit or cutoff frequency. That is, there is a method of adding some processing to the signal obtained by the image detector. 2. Description of the Related Art Conventionally, a technique for improving resolution of a panoramic tomographic image obtained by photographing a subject with a digital panoramic X-ray imaging apparatus by signal processing is known (for example, see Patent Documents 1 to 3).

Japanese Patent No. 4939287 Japanese Patent No. 4883618 Japanese Patent No. 4806433

  However, there is room for improvement in the techniques described in Patent Documents 1 to 3, and it has been desired to improve the image quality of the obtained monochrome image.

  In various conventionally known image processing methods, when a filtering process is performed on an original image, the spatial frequency domain of the image becomes narrower than that of the original image. This spatial frequency domain is obtained by Fourier transform of a signal of two-dimensional spatial coordinates (two-dimensional image data). Specifically, a region surrounded by a curved line, a vertical axis, and a horizontal axis representing the MTF frequency characteristics of the image data, with the spatial frequency as the horizontal axis and the modulation transfer function (MTF) as the vertical axis. When attention is paid to the above, the area of the region after the filter processing is reduced so that this region becomes narrower toward the lower spatial frequency side than the original image. Therefore, the conventional image processing method has a problem that the resolution is greatly reduced. Here, the reduction in the resolution of an image means that the image appears to be deteriorated by a filter even if the resolution determined by the pixel size of an image detector such as an FPD does not change. Note that the resolving power of a medical X-ray imaging system is actually determined by three factors: the pixel size of an image detector such as an FPD, the focal point size of an X-ray tube, and the magnification of a subject.

  The present invention has been made in view of the above problems, and by changing the gain of the MTF in the spatial frequency domain, an image having a low cost, excellent graininess, and a resolution that is more easily visible can be obtained. It is an object of the present invention to provide an image processing apparatus and an X-ray imaging apparatus that can be obtained.

In order to solve the above-described problems, an image processing apparatus according to the present invention is an image processing apparatus that performs predetermined image processing on an image obtained by photographing a subject, and is adapted to the coordinate position of each pixel in the image. By having a recording area for each pixel, the recording area for each pixel is horizontally divided into M rows (M is an integer of 2 or more) and vertically divided into N columns (N is an integer of 2 or more). A large-capacity memory storage means in which an expanded memory space corresponding to a coordinate space expanded by M × N times is set, and a signal value of a pixel in the input image is transferred to the large-capacity memory storage means, For each pixel, the same signal value is written in M × N memories corresponding to the recording area of the pixel set in the expanded memory space, and an image expanded in the expanded memory space is applied. Image processing means for performing frequency conversion filter processing, and the image processing means sets, as the frequency conversion filter processing, pixel signal values in the input image in the expanded memory space. In addition to recording in M × N memories corresponding to the recording area of the pixel, the input is performed at a position shifted from the recording area of the pixel in the input image vertically and horizontally within a range smaller than the recording area of the pixel. Recording the pixel signal value in the image that has been recorded in M × N memories in the expanded memory space, and performing the moving average process by adding the signal values recorded in each memory, respectively. Features.

In the image processing apparatus according to the present invention, after the image processing means performs the frequency conversion filter processing, an average of M × N memories for each recording area of the pixels in the expanded memory space. It is preferable to obtain the density and replace the signal value of each pixel in the input image .

  In addition, the X-ray imaging apparatus according to the present invention includes an X-ray image detection including the image processing apparatus, an X-ray source, and a pixel array in which pixels that receive X-rays that have passed through a predetermined point of a subject are two-dimensionally arranged. An image detector comprising a recording medium, a driving means for moving the image detector in a direction perpendicular to the X-ray incident direction, and a frame image for storing a frame image sent from the image detector at a predetermined frame rate An extended memory space corresponding to a coordinate space obtained by enlarging the frame image M × N times is set in the large-capacity memory storage unit. The frame image is transferred to the large-capacity memory storage means, and the frequency conversion filter process is performed on the frame image expanded in the expanded memory space.

  In addition, the X-ray imaging apparatus according to the present invention includes an X-ray image detection including the image processing apparatus, an X-ray source, and a pixel array in which pixels that receive X-rays that have passed through a predetermined point of a subject are two-dimensionally arranged. An image detector comprising a recording medium, a driving means for moving the image detector in a direction perpendicular to the X-ray incident direction, and a frame image for storing a frame image sent from the image detector at a predetermined frame rate Storage means, and arranged so that an arrangement direction of an arbitrary pixel group in a pixel array of the image detector is inclined with respect to a direction in which the image detector is moved by the driving means, In the large-capacity memory storage means, an expanded memory space corresponding to a coordinate space obtained by enlarging the frame image M × N times is set, and the image processing means stores the frame image in the large-capacity memory space. The output signal of the pixel of the image detector when the image detector is moved minutely by a distance smaller than the pixel width with respect to the direction in which the pixel group of the column is tilted is transferred to the memory storage means. Depending on the position, in the expanded memory space, the data is sequentially written in M × N memories corresponding to the position of the pixel, and the output signal of the pixel is written on the enlarged image. The frequency conversion filter processing is sequentially performed, and the signal value after the frequency conversion filter processing is added to each memory.

The image processing apparatus according to the present invention is excellent in graininess and can obtain an image having a resolution that is easier to visually recognize.
The X-ray imaging apparatus according to the present invention is excellent in graininess and can obtain an image having a resolution that is easier to visually recognize.

1 is a block diagram schematically showing an X-ray imaging apparatus according to an embodiment of the present invention. It is a figure which expands and shows typically a part of light-receiving surface of an image detector. It is a figure which shows typically the memory set in the memory storage means of the comparative example. It is a graph which shows typically MTF of an image recorded on memory storage means of a comparative example. It is a figure which expands and shows typically a part of light-receiving surface of an image detector. It is a figure which shows typically the memory set in the large capacity memory storage means. It is a graph which shows typically MTF of an image recorded on mass storage means. It is an enlarged view schematically showing a memory set in the large-capacity memory storage means. It is explanatory drawing of a moving average process. It is a schematic diagram which shows an example of FPD and the lead edge seen from the X-ray source side. It is a schematic diagram which shows an example of FPD and the lead edge seen from the side orthogonal to the irradiation direction of X-rays. It is a schematic diagram which shows an example of the signal value transferred from FPD to a mass memory storage means. It is a schematic diagram which shows an example of the algorithm of a DEMOT method. It is a graph which shows an example of ESF calculated | required from FIG. It is a graph which shows an example of LSF calculated | required from FIG. It is a graph which shows MTF calculated | required from FIG. 9B. The left is an example of an image when the DEMOT method is not used, and the right is an example of an image when the DEMOT method is used. The left is an example of an image when the DEMOT method is used, and the right is an example of an image when the DEMOT method and moving average processing are used in combination. It is a graph which shows MTF at the time of using the DEMOT method for image processing. It is a graph which shows MTF at the time of using a DEMOT method and a moving average process in combination for image processing. The left is another example of an image when the DEMOT method is used, and the right is another example of an image when the DEMOT method and a moving average process are used in combination. It is a graph which shows typically the improvement of the cutoff frequency of MTF. It is a graph which shows an example of a filter. It is a graph which shows improved MTF typically.

  DESCRIPTION OF EMBODIMENTS Embodiments for implementing an image processing apparatus and an X-ray imaging apparatus according to the present invention will be described in detail with reference to the drawings. Note that the size, positional relationship, and the like of the members shown in each drawing may be exaggerated for clarity of explanation.

(First embodiment)
The image processing device 20 performs predetermined image processing on an image obtained by photographing the subject 12. As shown in FIG. 1, the image processing apparatus 20 includes, for example, a large-capacity memory storage unit 8 and an image processing unit 10. Although the image processing apparatus 20 can be configured as a single apparatus, in the following description, for example, the X-ray imaging apparatus 1 will be described as including the image processing apparatus 20.

  The X-ray imaging apparatus 1 is, for example, a digital panoramic X-ray imaging apparatus, and as shown in FIG. 1, an X-ray source 2, an image detector 3, an arm 4, a turning drive unit 5, and A / D conversion. Means 6, large-capacity frame image storage means 7, large-capacity memory storage means 8, all-image display storage means 9, image processing means 10, and output means 11 are provided.

  The X-ray source 2 applies a slit-shaped X-ray beam generated by irradiating X-rays through a slit (lead slit) provided in a housing (not shown) made of lead, for example, to the subject 12 at a predetermined timing. Irradiation.

  The image detector 3 includes an X-ray image detection recording medium having a pixel array in which pixels that receive X-rays that have passed through a predetermined point of the subject 12 are two-dimensionally arranged. The image detector 3 images the portion of the subject 12 through which the X-rays have passed at a predetermined frame rate. The image detector 3 is an X-ray image sensor, an X-ray detector, or a combination thereof.

  Here, the image sensor is, for example, a CCD (Charge Coupled Device) image sensor, a CMOS image sensor, a TFT (Thin Film Transistor) sensor, a CdTe sensor, or the like. The X-ray detector is an X-ray image intensifier (I.I.), a flat panel detector (FPD), or the like.

  In the present embodiment, the image detector 3 will be described as an FPD. In this case, the image detector 3 can be configured as one pixel having a pixel size of, for example, 100 μm × 100 μm and arranged in, for example, horizontal direction 1024 pixels × vertical direction 1024 pixels.

  The arm 4 holds the X-ray source 2 and the image detector 3 at a predetermined interval. This interval is set to, for example, 30 cm to 1 m so that the subject 12 is accommodated between the X-ray source 2 and the image detector 3. In addition, the irradiation part in which the X-ray tube of X-ray source 2 is arrange | positioned, and the light-receiving surface of the image detector 3 are arrange | positioned facing. The arm 4 is configured to be rotatable and slidable around the rotation center O. The arm 4 fixes and holds the image detector 3 via a base (not shown) on the image detector 3 side.

  The turning drive means 5 moves the image detector 3 in a direction orthogonal to the X-ray incident direction by rotating and sliding the X-ray source 2 and the image detector 3 around the subject 12. The turning drive means 5 includes a motor, an actuator, and the like, and turns the arm 4 so as to rotate at a predetermined angular velocity. In the present embodiment, the turning drive unit 5 rotates and slides the image detector 3 around the rotation center O by rotating the arm.

  The turning drive unit 5, the X-ray source 2, and the image detector 3 are controlled by a controller (not shown), and the turning drive unit 5 rotates and slides the arm 4, while the X-ray source 2 is an X-ray source. The irradiation is repeated, and the image detector 3 captures a frame image of the subject 12 in synchronization with the X-ray irradiation timing and outputs it to the A / D conversion means 6.

  The A / D conversion means 6 acquires the output signal of the image detector 3, performs A / D conversion, and stores the output signal of the image detector 3 subjected to A / D conversion in the large-capacity frame image storage means 7. The output signal of the image detector 3 is an output signal of a plurality of pixels that are two-dimensionally arranged in the image detector 3, and is a frame image.

  The large-capacity frame image storage means 7, the large-capacity memory storage means 8, the all-image display storage means 9, and the image processing means 10 can be realized by a general computer (computer), for example, and a CPU ( A central processing unit (RAM), a random access memory (RAM), a read only memory (ROM), a hard disk drive (HDD), and an input / output interface are included.

  The large-capacity frame image storage means 7 is composed of a general image memory, a hard disk or the like, and stores output signals (frame images) sent from the image detector 3 at a predetermined frame rate. The large-capacity frame image storage means 7 stores the A / D converted frame image.

  The large-capacity memory storage unit 8 includes a general image memory, a hard disk, and the like. In the present embodiment, the image processing means 10 uses the large-capacity memory storage means 8 to perform frequency conversion filter processing and panoramic tomographic image construction processing.

  The large-capacity memory storage means 8 has a recording area for each pixel that is aligned with the coordinate position of each pixel in an image (here, a frame image), and M rows (M is 2 or more) across the recording area for each pixel. An expanded memory space corresponding to a coordinate space obtained by enlarging the image M × N times is set by dividing the image into N columns (N is an integer of 2 or more). In the following, as an example, it is assumed that M and N have the same value. This expanded memory space is used for frequency conversion filter processing.

  In the large-capacity memory storage unit 8, a memory space corresponding to each pixel in the frame image is set on a one-to-one basis. This one-to-one memory space is used for panoramic tomographic image construction processing. The storage means used for constructing the panoramic tomographic image may be a separate storage device.

  The all-image display storage unit 9 is composed of a general image memory or the like, and stores the panoramic tomographic image (image to be displayed) generated by the image processing unit 10. This panoramic tomographic image is represented by a luminance value, for example.

  The image processing means 10 reads out the frame image from the large-capacity frame image storage means 7 and transfers it to the large-capacity memory storage means 8, and performs predetermined image processing on the large-capacity memory storage means 8. That is, the image processing unit 10 acquires the frame image converted and output by the A / D conversion unit 6 and performs image processing. In the present embodiment, the image processing means 10 has a function of frequency conversion filter processing and a function of processing for constructing a panoramic tomographic image.

  As a function of the frequency conversion filter process, the image processing unit 10 transfers the pixel signal value in the input image (here, the frame image) to the large-capacity memory storage unit 8. Then, the image processing means 10 writes the same signal value to M × M memories corresponding to the recording area of the pixel set in the expanded memory space for each pixel. Then, the image processing means 10 performs frequency conversion filter processing on the image enlarged in the expanded memory space. Here, the frequency conversion filter process is a process of changing the frequency component of a signal included in an image when the image enlarged in the expanded memory space is viewed in the spatial frequency domain.

  Specifically, the frequency conversion filter processing includes, for example, a smoothing filter (smoothing), an edge extraction filter (edge detection), an unsharp mask (unsharp masking), a frequency filter (Frequency Filter), and multi-frequency processing (Multi-Objective Frequency). Processing: MFP) is preferable. Examples of the smoothing filter include a moving average process (Moving Average), a Gaussian filter (Gausian Filter), and a median filter (Median Filter).

  The moving average process is a process that averages the density values using the density values (luminance values) around the target pixel to obtain the density values of the processed image. The Gaussian filter is a process that averages the density values using a function of a Gaussian distribution in which the weight increases as the pixel is closer to the target pixel to obtain the density value of the processed image. The median filter is a process in which density values (luminance values) around the pixel of interest are arranged in ascending order and the median value is used as the density value of the processed image. These filters are smoothing processes performed in the real space region of the image.

The edge extraction filter is, for example, a differential filter. In the case of the first derivative, for example, a Sobel filter, a Prewitt filter, a Roberts filter, and the like can be given. In the case of the second derivative, for example, a Laplacian filter can be cited.
The unsharp mask is one of sharpening filters, and is a process for emphasizing the outline of the blurred portion of the image by masking the blurred portion of the image.

  The frequency filter is a process for adjusting the sharpness of an image using a high-pass filter (HPF) or the like. The frequency filter is a process performed in the frequency domain after the Fourier transform of the image.

  Multi-frequency processing is processing that emphasizes components in a plurality of spatial frequency bands.

  The image processing unit 10 can output an image having a size 100 times that of the input image after performing the frequency conversion filter process on the input image. However, in the following, as an example, an image having the same size as the input image is output. Will be described. In this case, after performing the frequency conversion filter process, the image processing means 10 obtains an average density of M × M memories for each pixel recording area in the expanded memory space, and each pixel in the input image By substituting this signal value, an image having the same size as the input image is output.

  As processing for constructing a panoramic tomographic image, the image processing means 10 reads a plurality of frame images whose signal values are replaced from the large-capacity frame image storage means 7. Then, the image processing means 10 has a predetermined space corresponding to a predetermined tomographic surface such as a dentition of the subject 12 in a memory space (memory space corresponding to one-to-one pixel) set in the large-capacity memory storage section 8. A plurality of frame images are overlapped with the shift width. Thereby, a panoramic tomographic image corresponding to a predetermined tomographic surface such as a dentition of the subject 12 is constructed. The process of constructing this panoramic tomographic image is known. Normally, the frame image detected by the image detector 3 and A / D converted is used as it is. However, in the X-ray imaging apparatus 1 of this embodiment, the frame image in which the signal value is replaced after the frequency conversion filter processing is used. Build a panoramic tomographic image.

  The function of the image processing means 10 is realized by the CPU developing and executing a predetermined program stored in the ROM or the like on the RAM. This program can be provided via a communication line, or can be provided by being written in a recording medium such as a CD-ROM.

  The output means 11 is comprised from CRT (Cathode Ray Tube), a liquid crystal display (LCD: Liquid Crystal Display), PDP (Plasma Display Panel), EL (Electronic Luminescence) etc., for example.

  Next, in the large-capacity memory storage unit 8, an expanded memory space set corresponding to a coordinate space obtained by enlarging an image (here, a frame image) M × M times, corresponding to a normal coordinate space. This will be described in comparison with the set memory space.

First, a memory space set corresponding to a normal coordinate space will be described with reference to FIGS. 2A to 2C. FIG. 2A is a diagram schematically showing an enlarged part of the light receiving surface of the image detector 3. Here, attention is focused on a region in which pixels P _{1 to} P ₉ having a pixel size of d [μm] × d [μm] are arranged in 3 rows and 3 columns. Hereinafter, the pixel size is simply expressed by the length of one side (pixel width) d. FIG. 2B is a diagram schematically showing, for example, a 5 × 5 memory (address) set in the memory storage unit (referred to as memory storage unit 18) of the comparative example. These 5 × 5 memories correspond to the 5 × 5 pixels of the image detector 3 in terms of position and size. If, for example, a pixel size of d = 100 [μm] is used as the pixel size of the image detector 3 shown in FIG. 2A, one memory of the memory storage means 18 also corresponds to a pixel of d = 100 [μm].

In FIG. 2B, the numerical value written in the 3 × 3 memory in the center portion shows an example of the signal level of the output signal of each pixel P _{1 to} P ₉ of the image detector 3. In this example, the level of the output signal of the pixel P ₅ is 100. Further, the level of the output signal of the pixels P ₂ , P ₄ , P ₆ and P ₈ is set to 80. Further, the level of the output signal of the pixels P ₁ , P ₃ , P ₇ , P ₉ is set to 60. In this case, the signal level 100 is written in the memory corresponding to the position of the pixel P _{5 in} the memory storage means 18. Similarly, the signal level 80 of the pixel P _{6 on} the right side of the pixel P ₅ is written in the memory corresponding to the position of the pixel P ₆ in the memory storage unit 18. In this way the memory storage means 18, for each pixel corresponding to the position of each pixel P ₁ to P ₉ of the normal coordinate space, the recording region for recording the signal level of the output signal of the pixel is set.

  FIG. 2C is a graph schematically showing an MTF (Modulation Transfer Function) when an image recorded in the memory storage unit 18 is viewed in the frequency domain. The horizontal axis of the graph is the spatial frequency ω [cycles / mm]. The vertical axis of the graph is MTF, and the value of MTF when ω = 0 is normalized to 1. Here, MTF is an absolute value of an optical transfer function (OTF), and is used as a numerical value representing the sharpness of an image. Note that OTF is a Fourier transform of a line spread function (LSF). Here, the resolving power of the imaging system depends on only the pixel size d [μm] of the image detector 3. In this case, the cutoff frequency of the MTF is 500 / d [cycles / mm] because it is expressed by the Nyquist frequency. When d = 100, the cutoff frequency of the MTF is 5 [cycles / mm].

  Next, the expanded memory space set in the large-capacity memory storage unit 8 will be described with reference to FIGS. 3A to 3C. FIG. 3A is a diagram schematically showing an enlarged part of the light receiving surface of the image detector 3, and is the same as FIG. 2A.

FIG. 3B is a diagram schematically showing, for example, a 50 × 50 memory (address) set in the large-capacity memory storage unit 8. These 50 × 50 memories correspond to the 5 × 5 pixels of the image detector 3 in alignment. That is, one pixel of the image detector 3 corresponds to the 10 × 10 memory set in the large-capacity memory storage unit 8 in alignment. This is an example in which the vertical and horizontal division number M of one pixel of the image detector 3 is set to 10.
Here, it is assumed that the signal levels of the output signals of the pixels P _{1 to} P ₉ of the image detector 3 shown in FIG. 3A are the same as those in the comparative example described above. That is, the level of the output signal of the pixel P ₅ is 100, the level of the output signal of the pixels P ₂ , P ₄ , P ₆ and P ₈ is 80, and the level of the output signal of the pixels P ₁ , P ₃ , P ₇ and P _9. Is 60.

In this case, for example, when the output signal of the pixel P ₅ is transferred to the large-capacity memory storage means 8, as shown in FIG. 4, the signal level 100 is written in the 10 × 10 memory corresponding to the position of the pixel P _5. It is. Similarly, the signal level 80 of the pixel P _{6 on} the right side of the pixel P ₅ is written in the 10 × 10 memory corresponding to the position of the pixel P _{6 in} the large-capacity memory storage unit 8. As described above, the large-capacity memory storage unit 8 records 10 × 10 for recording the signal level of the output signal of each pixel corresponding to the position of each pixel P _{1 to} P ₉ of the image detector 3. The area is set.

  When the pixel size of the image detector 3 shown in FIG. 3A is, for example, d = 100 [μm], one memory of the large-capacity memory storage unit 8 has a smaller d = 10 [μm]. It corresponds to the pixel. In this case, an expanded memory space corresponding to a coordinate space obtained by enlarging the frame image acquired by the image detector 3 by 10 × 10 times is set in the large-capacity memory storage unit 8.

  FIG. 3C is a graph similar to FIG. 2C, and is a graph schematically showing an MTF when an image recorded in the large-capacity memory storage unit 8 is viewed in the frequency domain. When the resolving power of the imaging system depends only on the pixel size d [μm] of the image detector 3, the MTF curve is represented by a solid line. For example, if the pixel size of the image detector 3 is actually used as d = 100 [μm], the cutoff frequency of the MTF is 5 [cycles / mm] in the frequency domain. As described above, when the pixel of d = 100 [μm] is actually used, and the pixel size is replaced with d = 10 [μm], the MTF curve is represented by a broken line in the frequency domain. The cutoff frequency is 50 [cycles / mm]. Therefore, when the large-capacity memory storage unit 8 set so as to correspond to the coordinate space obtained by enlarging the actual frame image acquired by the image detector 3 by 10 × 10 times is used, when the image is viewed in the frequency domain, the MTF This produces the same effect as increasing the value of the cut-off frequency of 10 times the actual value. Note that the slope of the dashed curve is exaggerated in the graph of FIG. 3C for easy viewing.

  Next, a specific example of processing by the image processing means 10 will be described with reference to FIG. Here, the case where a moving average process is performed as a frequency conversion filter process is demonstrated. If the moving average process is not performed, the signal level of the output signal of each pixel of the image detector 3 is only added once to the corresponding coordinate position of the large-capacity memory storage unit 8 as illustrated in FIG. On the other hand, when moving average processing is performed, for example, a value shifted from the corresponding coordinate position of the large-capacity memory storage unit 8 up, down, left, and right is added, for a total of five times.

FIG. 5 is a diagram schematically showing these addition processes. The large-capacity memory storage unit 8 shown in the upper left in FIG. 5 schematically shows a part of the large-capacity memory storage unit 8 shown in FIG. 4, and the pixel P ₅ of the image detector 3 shown in FIG. A 10 × 10 memory corresponding to the position and its surrounding memory are shown.

Here, it is assumed that the signal levels of the output signals of the pixels P _{1 to} P ₉ of the image detector 3 shown in FIG. 3A are the same as those in the comparative example described above. That is, the level of the output signal of the pixel P ₅ is 100, the level of the output signal of the pixels P ₂ , P ₄ , P ₆ and P ₈ is 80, and the level of the output signal of the pixels P ₁ , P ₃ , P ₇ and P _9. Is 60.

In FIG. 5, a diagram in which a thick rectangle indicated by reference numeral 40 is superimposed on a memory space indicated by reference numeral 51 in the upper right portion schematically illustrates the first addition process. A bold rectangle 40 is the signal level of the pixel P ₅ as shown in FIG. 4 and represents the same value of 10 × 10. As indicated by reference numeral 51, when the image processing means 10 performs the moving average process, it is initialized for each of the pixels P _{1 to} P ₉ in the expanded memory space set in the large-capacity memory storage means 8. The signal level of the output signal of each of the pixels P _{1 to} P ₉ is written in the 10 × 10 memory corresponding to the position. In FIG. 5, the numerical value of the signal level is omitted.

In FIG. 5, a diagram in which the thick rectangle indicated by reference numeral 40 is superimposed on the memory space indicated by reference numeral 52 in the middle left portion schematically illustrates the second addition process. For the second time, for example, a pixel shifted by 1 pixel (0.1 pixel in the real space) to the left in the memory space from the corresponding coordinate position of the large-capacity memory storage unit 8 is added. In this case, in the example shown in FIG. 4, the signal level 100 is written in the memory in which the signal level 80 in the second column, for example, is written from the left. As a result, the output signal of the pixel P ₅ affects the output signal of the pixel P ₄ of the image detector 3 in the memory space. At this time, in the example shown in FIG. 4, the signal level 80 is written into the memory in which the signal level 100 in the third column, for example, is written from the right. This is because the output signal of the pixel P ₆ affects the output signal of the pixel P ₅ of the image detector 3 in the memory space. Therefore, the densities of the pixels P ₄ , P ₅ and P ₆ are smoothed.

In FIG. 5, a diagram in which a thick rectangle indicated by reference numeral 40 is superimposed on the memory space indicated by reference numeral 53 in the middle right portion schematically illustrates the third addition process. In the third time, for example, a pixel shifted by one pixel (0.1 pixel in the real space) to the right in the memory space from the corresponding coordinate position of the large-capacity memory storage unit 8 is added. In this case, in the example shown in FIG. 4, the signal level 100 is written in the memory in which the signal level 80 in the second column from the right is written. This is because, in the memory space, the output signal of the pixel P ₆ of the image detector 3, influences of an output signal of the pixel P _5, also, the output signal of the pixel P ₅ of the image detector 3, the pixel P ₄ This contributes to the smoothing of the density of each pixel.

In FIG. 5, a diagram in which a thick-line rectangle indicated by reference numeral 40 is superimposed on a memory space indicated by reference numeral 54 in the lower left stage schematically illustrates the fourth addition process. In the fourth time, for example, a value shifted by one pixel (0.1 pixel in the real space) in the memory space from the corresponding coordinate position of the large-capacity memory storage unit 8 is added. In this case, in the example shown in FIG. 4, the signal level 100 is written in the memory in which the signal level 80 in the second column from the top is written. This is because, in the memory space, the output signal of the pixel P ₂ of the image detector 3, influences of an output signal of the pixel P _5, also, the output signal of the pixel P ₅ of the image detector 3, the pixel P ₈ This contributes to the smoothing of the density of each pixel.

In FIG. 5, the diagram in which the thick line rectangle indicated by reference numeral 40 is superimposed on the memory space indicated by reference numeral 55 in the lower right stage schematically shows the fifth addition process. For the fifth time, a value shifted by, for example, one pixel downward (0.1 pixel in the real space) in the memory space from the corresponding coordinate position of the large-capacity memory storage unit 8 is added. In this case, in the example shown in FIG. 4, the signal level 100 is written in the memory in which the signal level 80 in the second column from the bottom is written. This is because, in the memory space, the output signal of the image detector 3 pixel P _8, influences of an output signal of the pixel P _5, also, the output signal of the pixel P ₅ of the image detector 3, the pixel P ₂ This contributes to the smoothing of the density of each pixel.

Here, a process of obtaining an average density of M × M memories for each pixel recording area in the expanded memory space after the image processing means 10 performs frequency conversion filter processing on the input image will be described. If the above five addition processes are performed without shifting the position, the sum of the signal levels added to the 10 × 10 memory corresponding to the position of the pixel P _{5 in} the memory space is 100 × 100 × 5 = 50000. Therefore, the average value (average density) of the signal level per one addition process and one memory is 100. On the other hand, when moving average processing is performed by shifting one pixel up, down, left, and right in the memory space, the average value (average density) of the signal level is 98.4 as a result of calculation. The image processing means 10 replaces the average value 98.4 of the signal level with the signal level 100 at the position of the pixel P ₅ in the input frame image. Similarly, the image processing means 10 obtains the average density of the signal level for each pixel other than the pixel P ₅ and replaces it with the signal level of each pixel P _{1 to} P _{9 in} the input frame image.

  In the moving average process, the order of performing the five additions may be changed. Further, the number of pixels shifted vertically and horizontally in the memory space is not limited to one pixel. In the moving average process, addition is performed five times using the values of M × M memories themselves. That is, in the addition process, a thick line rectangle (in other words, a function that becomes a square wave) indicated by reference numeral 40 in FIG. 4 is used. By applying a function that becomes a sine wave, the value at the center of the thick rectangular area indicated by reference numeral 40 in FIG. 4 is increased, and then added vertically and horizontally. Good. Note that after performing frequency conversion filter processing on the input image, an image having a size 100 times that of the input image may be output without replacing the signal level of each pixel in the input image.

According to the X-ray imaging apparatus 1 of the present embodiment, the gain of the MTF can be changed by the processing on the memory set in the large-capacity memory storage unit 8. It is possible to obtain an image having a resolution that is easy to visually recognize with low cost and excellent graininess. According to the X-ray imaging apparatus 1, the image processing unit 10 can improve the graininess of the frame image by performing the moving average process as the frequency conversion filter process. Therefore, when the image processing means 10 of the X-ray imaging apparatus 1 constructs a panoramic tomographic image by superimposing a plurality of frame images with improved granularity with a predetermined shift width, this panoramic tomographic image is also grainy. Is improved.
Furthermore, even if the image processing means 10 performs frequency conversion filter processing other than moving average processing on a frame image, the granularity of the frame image can be improved and the image quality can be improved.

(Second Embodiment)
In the X-ray imaging apparatus 1 according to the second embodiment, the same components as those shown in FIG. 1 are denoted by the same reference numerals, description thereof will be omitted as appropriate, and description will be made with reference to FIG. In the X-ray imaging apparatus 1 according to the second embodiment, the image processing means 10 further performs image processing by the Detector Moving and Frame Additional Technique (hereinafter referred to as the DEMOT method) which is a super-resolution technique. is there. Image processing by the DEMOT method is described in Patent Document 1 and Patent Document 3, and an image reconstruction algorithm of the DEMOT method will be described below with reference to the drawings.

  Taking an edge image as an example will be described with reference to FIGS. 6A and 6B. Here, a flat panel detector (FPD) that is a recording medium is used as the image detector 3 and imaging is performed by an edge method in which the subject 12 is a metal plate (lead edge).

FIG. 6A shows the arrangement of the FPD 63 and the lead edge 64. The FPD 63 is placed on a stage (not shown) capable of minute linear movement. Reference numeral 65 in FIG. 6A indicates the moving direction of the FPD 63. A lead edge 64 is placed close to the FPD 63. FIG. 6B illustrates a part of a cross-sectional pixel column along the alternate long and short dash line illustrated in FIG. 6A. It is assumed that the edge portion of the lead edge 64 coincides with the boundary line between the pixels of the FPD 63 at the initial state before the FPD 63 moves slightly and at the start. Pixels hidden lead edge 64 among the pixels of the FPD63 and P _1, the pixel is irradiated in the X-ray 66 and P _2.

The pixels P ₁ , P ₂ ,... Are one-dimensionally arranged on the one-dot chain line shown in FIG. 6A of the FPD 63 having a large number of pixels. Hereinafter, the FPD 63 will be described focusing on this one-dimensional line. Assume a distance smaller than the pixel width. Specifically, assuming a minute distance in which the pixel width is divided into a plurality of parts, the operation of moving the FPD 63 by this minute distance and performing shooting is repeated as many times as the number of divisions, and finally the FPD 63 is made only by the pixel size. Shall be moved. Here, for example, assuming that the pixel size is d and the number of pixel divisions is 10, the operation of performing shooting by moving the FPD 63 slightly by d / 10 is repeated 10 times. As shown in FIG. 6B, the lead edge 64 is arranged slightly spaced from the surface of the FPD 63 and is fixed by a holder (not shown). Therefore, even if the FPD 63 moves, the lead edge 64 does not move in the initial arrangement.

And X-ray 66 is irradiated perpendicularly to FPD63, when shooting a lead edge image, an output signal corresponding to the incident X-ray amount from the pixel P ₂ is obtained. The two-dimensional image of the FPD 63 is transferred to the large-capacity memory storage unit 8. In the memory space of the large-capacity memory storage unit 8, the size and position of the pixels of the FPD 63 are matched in advance, and corresponding to the number of divisions of the pixels of the FPD 63, ten times that of one pixel of the one-dimensional line of the FPD 63 The number of pixels is set. Therefore, the memory space of the large-capacity memory storage unit 8 is set with a number of pixels that is 100 times the total number of pixels of the two-dimensional FPD 63. The output signal of one pixel on the one-dimensional line of the FPD 63 is similarly transferred to 10 pixels on the memory space of the large-capacity memory storage unit 8 corresponding to the pixel position.

Here, the level of the output signal of the pixel will be described. The shooting at the start of the minute movement of FPD63 as shown in FIG. 6B, X-ray 66 is irradiated to the entire pixel in the pixel P _2. The level of the output signal of the pixel P _{2 at} this time is set to 10 for convenience, corresponding to the number of divided pixels. On the other hand, since the pixel P ₁ is blocked by the lead edge 64 and is not irradiated with the X-ray 66, the level of the output signal of the pixel P _{1 at} this time is set to zero.

The output signal (level 10) transferred from the pixel P _{2 of the} FPD 63 is written to 10 pixels corresponding to the position of the P ₂ pixel in the memory space of the large-capacity memory storage unit 8.
The output signal (level 0) transferred from the pixel P _{1 of the} FPD 63 is written in 10 pixels corresponding to the position of the P ₁ pixel in the memory space of the large-capacity memory storage unit 8.

Next, as indicated by reference numeral 65 in FIG. 7, the FPD 63 is slightly moved to the left in FIG. 7 by 1/10 of the pixel size. Then, the pixel P ₂ is shielded by the lead edge 64 in 1/10 of the pixel area by this first movement. Therefore, the P ₂ pixel is irradiated with X-rays 66 on an area of 90% of the entire pixel. Accordingly, the level of the output signal of the pixel P _{2 at} this time is 9. The output signal (level 9) is transferred to the large-capacity memory storage means 8 and written is the numerical value shown in the middle of FIG. This output signal (level 9) is transferred to 10 pixels corresponding to the position after the P ₂ pixel has moved minutely by 1/10 of the pixel size in the memory space of the large-capacity memory storage means 8.

In FIG. 7, the position of the pixel P ₂ of the FPD 63 shown in the upper stage is the position at the start of the minute movement. In FIG. 7, the position of the pixel in the memory space of the large-capacity memory storage means 8 shown in the middle stage indicates the position at the time of the first movement. In the memory space of the large-capacity memory storage means 8, the leftmost pixel among the 10 pixels to which the level 9 output signal is written has the output signal (level 0) transferred from the pixel P ₁ at the start of the minute movement. This is the pixel that was written. That is, in the writing process associated with the first movement, level 9 is written to 10 pixels including pixels shifted to the left by one pixel from the pixels in the memory space of the large-capacity memory storage unit 8 at the start. Accordingly, in the first movement, for example, the output signal (level 0) transferred from the pixel P ₁ is obtained by shifting the pixel shifted to the left by one pixel from the pixel on the memory space of the large-capacity memory storage unit 8 at the start. It will be written in 10 pixels including.

After the first movement of the FPD 63 and the writing of the output signal are completed, the FPD 63 is further moved slightly to the left in FIG. 7 by 1/10 of the pixel size to capture an edge image. Due to the second movement, the movement amount from the first position becomes 2/10 of the pixel size. Then, by the second movement, 2/10 of the pixel area of the pixel P ₂ is shielded by the lead edge 64. Therefore, the P ₂ pixel is irradiated with X-rays 66 on an area of 80% of the entire pixel. Accordingly, the level of the output signal of the pixel P _{2 at} this time is 8. The output signal (level 8) is transferred to the large-capacity memory storage means 8 and written is the numerical value shown in the lower part of FIG.

  As described above, while the FPD 63 is moved and photographed, the output signal of each pixel is transferred to the large-capacity memory storage unit 8 each time. At this time, the pixel position on the memory space of the large-capacity memory storage unit 8 and the lead edge 64 do not move from the initial set position. FIG. 8 shows a list of the output signals of the respective pixels in accordance with the minute movement of the FPD 63 as a signal sequence transferred to the large-capacity memory storage unit 8.

  In this case, since the movement is performed by 1/10 of the pixel size, the movement is completed by 10 movements, and returns at the start. That is, when the circuit is completed, the edge of the lead edge 64 coincides with the boundary line between the next pixels of the FPD 63. Output signals written to each pixel in the memory space of the large-capacity memory storage unit 8 during the movement until returning to the start are sequentially added. This process is a process of adding the signal levels in the table of FIG. 8 in the column direction.

  In FIG. 8, the final addition result is indicated as “0, 0, 0, 1, 3, 6,..., 100” under the large-capacity memory storage unit 8. This numerical sequence is a photograph of the lead edge 64 and has a distance dimension. The curve represented by this numerical sequence is called an Edge Spread Function (ESF). The line spread function (LSF) calculated from the ESF distance derivative is shown as “0, 0, 1, 2, 3,..., 0” at the bottom of FIG. A theoretically obtained ESF graph is shown in FIG. 9A. A theoretically obtained LSF graph is shown in FIG. 9B. The shape of the LSF is a triangular wave as a whole, but a shape obtained by combining two symmetrical sawtooth waves.

  From FIG. 9B, the MTF obtained when the shape of the LSF is a sawtooth wave is shown by a solid line in FIG. For comparison, the MTF obtained when the shape of the LSF is a rectangular wave is shown by a broken line in FIG. When the shape of the LSF is a rectangular wave, the cutoff frequency of the MTF is represented by 1 / (2d), that is, the Nyquist frequency, where d is the pixel size. On the other hand, according to the DEMOT method, the cutoff frequency of the MTF is 1 / d when the pixel size is d, which is theoretically expressed by twice the Nyquist frequency, and the resolution is improved.

  In the above-described DEMOT image reconstruction algorithm, the FPD 63 has been described by focusing on the one-dot chain line shown in FIG. 6A in order to simplify the description. Actually, the FPD 63 has two-dimensionally arranged pixels. In the DEMOT method, Patent Document 3 discloses a technique for improving the resolving power in two directions of a horizontal direction (one axial direction) and a vertical direction (the other axial direction) of an image. Therefore, the X-ray imaging apparatus 1 according to the second embodiment has the image detector 3 in the direction in which the image detector 3 is moved by the turning drive unit 5 in order to improve the resolving power in the two directions of the image. The pixels are arranged so that the arrangement direction of an arbitrary group of pixels in the pixel array is inclined.

  This tilt angle is arbitrary, but it is preferable that the tilt angle is 45 degrees in order to improve the resolution in the vertical direction and the resolution in the horizontal direction equally in the obtained image. That is, assuming that the angle in the direction in which the image detector 3 is moved by the turning drive unit 5 is a reference azimuth (0 degree), the pixel group arranged in the vertical column in the pixel array of the image detector 3 has a direction of, for example, 45 degrees. It is preferable that the pixel groups arranged in the horizontal row are arranged in a direction of, for example, 135 degrees. Hereinafter, description will be made assuming that the inclination angle is 45 degrees.

  Specifically, assuming that the angle in the direction in which the image detector 3 is moved is a reference azimuth (0 degree), the pixel groups arranged in the vertical column in the pixel array of the image detector 3 are arranged in an azimuth of 45 degrees, for example. Then, the pixel groups arranged in the horizontal row are arranged, for example, in the direction of −45 degrees (that is, the direction of 135 degrees). In this case, when the turning drive unit 5 moves the image detector 3 by 1.41 times a minute distance of 1 / M of the pixel size with respect to the reference azimuth (0 degree), the image detector 3 is relatively moved. Can be moved minutely by 1 / M of the pixel size in a 45 degree azimuth. At the same time, the pixel groups arranged in the horizontal row of the image detector 3 can be relatively moved by 1 / M of the pixel size in the direction of −45 degrees.

  In the present embodiment, the image processing means 10 outputs the pixels of the image detector 3 when the image detector 3 is moved minutely by 1 / M of the pixel size in the direction of 45 degrees when viewed from the reference direction of the moving direction. In accordance with the position where the image detector 3 is moving, the signals are sequentially written in order in the M × M memories corresponding to the positions of the pixels in the expanded memory space. This is a part of the image processing by the DEMOT method during the minute movement of the pixel schematically shown in FIG.

  Next, the image processing unit 10 sequentially performs frequency conversion filter processing on the enlarged image corresponding to the signal level of the pixels allocated to the M × M memories for each pixel of the large-capacity memory storage unit 8. . This is part of the image processing described in the X-ray imaging apparatus 1 according to the first embodiment. Specifically, the image processing unit 10 moves the image detector 3 by 1 / M of the pixel size, and after the first movement and writing of the output signal are finished, the pixel of the large capacity memory storage unit 8 A first frequency conversion filter process is performed on the image expanded in each M × M memories. Then, the image processing means 10 slightly moves the image detector 3 by 2 / M of the pixel size, and after the second movement and writing of the output signal are finished, the M for each pixel of the large-capacity memory storage means 8. A second frequency conversion filter process is performed on the image enlarged by × M memories. Similarly, after the image detector 3 is slightly moved by the pixel size and the M-th movement and the writing of the output signal are finished, the image detector 3 is enlarged by M × M memories for each pixel of the large-capacity memory storage means 8. The Mth frequency conversion filter process is performed on the obtained image. Then, the image processing means 10 outputs the signal value after the frequency conversion filter process written in M × M memories for each pixel of the large capacity memory storage means 8 during the movement until the image detector 3 is slightly moved by the pixel size. Are added sequentially. This process is the same as the process of adding the signal levels in the table of FIG. 8 in the column direction.

  As described above, according to the X-ray imaging apparatus 1 according to the second embodiment, the gain of the MTF is changed by the process on the memory set in the large-capacity memory storage unit 8 as in the first embodiment. In the conventional case, where the resolution is greatly reduced, an image having a resolution that is excellent in graininess and more easily visible can be obtained at low cost. Furthermore, according to the X-ray imaging apparatus 1 according to the second embodiment, it is possible to raise the cutoff frequency of the MTF in the spatial frequency domain to twice the Nyquist frequency by image processing using the DEMOT method, and at low cost. While super-resolution, the image is excellent in graininess and has a resolution that is easier to visually recognize.

[Example]
In order to confirm the performance of the X-ray imaging apparatus 1 of the present invention, the following Experiments 1 to 3 were performed.
In each experiment, the effect of the X-ray imaging apparatus 1 according to the second embodiment was confirmed. This will be described with reference to FIG. In each experiment, a known Siemens star chart (hereinafter simply referred to as a star chart) was used as the common subject 12. The star chart is made of lead and formed in a disk shape, and a plurality of fan shapes that radiate from the center toward the periphery form a black and white striped pattern. Thus, it can be seen that the more the radial black and white striped pattern is seen near the center of the star chart, the higher the resolving power of the photographing apparatus is.

<Experiment 1>
Experiment 1 is a preliminary experiment, which is an experiment for photographing a star chart and confirming the improvement of the resolving power by the DEMOT method. For this reason, frequency conversion filter processing is not performed on the acquired image. It is simply an experiment comparing an image using the DEMOT method with an image not using it.

  The image shown on the left side of FIG. 11 is an image output when the DEMOT method is not used for image processing. The image shown on the right side of FIG. 11 is an image output when the DEMOT method is used for image processing. In each image, the surface of the star chart is darker than the background. Each image is an enlarged view near the center of the star chart. In the image shown on the right side in FIG. 11, the black and white striped pattern of the star chart is clearer than the image shown on the left side. Therefore, it was confirmed that the resolution was improved when the DEMOT method was used.

<Experiment 2>
Experiment 2 is an experiment for photographing a star chart and confirming that the X-ray imaging apparatus 1 according to the second embodiment reduces image noise without reducing the resolving power. Here, the case where the DEMOT method is used without performing the frequency conversion filter processing on the acquired image is taken as a comparative example. Moreover, the case where the moving average process was performed as a frequency conversion filter process and the DEMOT method was used was made into the Example.

<Experiment 2-1> Visual Evaluation The image shown on the left side of FIG. 12 is an image of a comparative example. The image shown on the right side of FIG. 12 is an image of the example. In FIG. 12, since the captured image is reversed in black and white, the surface of the star chart is brighter than the background in each image. Each image is an enlarged view near the center of the star chart. In visual evaluation, it can be read that there is no difference in resolving power between the left and right images in FIG. In addition, it can be seen that the image shown on the right side in FIG. 12 is more effective in reducing noise than the image shown on the left side.

<Experiment 2-2> Spatial Resolution Measurement For the images shown on the left and right of FIG. 12, the spatial resolution was measured by the edge method. The measurement results are shown in FIGS. 13A and 13B. FIG. 13A is a graph showing the MTF in the case of the comparative example, and FIG. 13B is a graph showing the MTF in the case of the example. The horizontal axis of each graph is the spatial frequency ω [cycles / mm]. The vertical axis of each graph is MTF, and the value of MTF when ω = 0 is normalized to 1. The pixel size of the image detector 3 used for photographing is d = 100 [μm].

  In the case of the comparative example, the moving average processing is not performed on the image, but since the DEMOT method is used, the cutoff frequency of the MTF is theoretically 10 [cycles / mm] as shown in FIG. 13A. . However, in consideration of error components and the like, the cut-off frequency is considered to be about 8.0 [cycles / mm] as shown by the broken line.

  In the case of the embodiment, the moving average processing is performed on the image, but since the DEMOT method is used, the cutoff frequency of the MTF is theoretically 10 [cycles / mm] as shown in FIG. 13B. . However, in consideration of error components and the like, the cut-off frequency is considered to be about 8.0 [cycles / mm] as shown by the broken line. Even if FIG. 13B and FIG. 13A are compared, it is considered that there is no decrease in the resolution.

<Experiment 3>
Experiment 3 is an experiment for photographing a star chart and confirming how much noise is reduced by the X-ray imaging apparatus 1 according to the second embodiment. Here, the case where the DEMOT method is used without performing the frequency conversion filter processing on the acquired image is taken as a comparative example. Moreover, the case where the moving average process was performed as a frequency conversion filter process and the DEMOT method was used was made into the Example.

  The image shown on the left side of FIG. 14 is an image of a comparative example. The image shown on the right side of FIG. 14 is an image of the example. In each image, the surface of the star chart is darker than the background. Each image is an enlarged view near the center of the star chart. In Experiment 3, a region within a rectangular frame in the left and right images in FIG. 14 was selected, the density of pixels in the selected region was measured, and the standard deviation value of the density was calculated. The measurement results are shown in Table 1.

In Table 1, No. 1 is the result of the comparative example, and No. 2 is the result of the example.
Area is the total number of pixels in the selected area of the image.
Mean is the average density of the pixels in the selected area of the image.
StdDev is the standard deviation value of the density of the pixels in the selected area of the image.
Min is the minimum density of pixels in the selected area of the image.
Max is the maximum value of the density of the pixels in the selected area of the image.

  The smaller the standard deviation value of the density, the smaller the noise. According to Table 1, the noise of the image of the example is clearly reduced as compared with the image of the comparative example. The standard deviation value of the concentration of the example was about 88% of the standard deviation value of the concentration of the comparative example. Therefore, according to the X-ray imaging apparatus according to the present invention, it is possible to improve the granularity of an image by suppressing a decrease in resolution and improve the image quality.

[Modification]
The present invention is not limited to the above-described embodiments. For example, although the frequency conversion filter process performed on the image in the memory space set in the large-capacity memory storage unit 8 has been described as a moving average process, a general filter function may be used. That is, it is possible to use a general filter function that is conventionally used for filter processing performed on the pixel coordinate space of the original image itself. In the case of the moving average process, the calculation process is performed in the real space area of the image. However, in the case of a frequency filter, for example, the calculation process may be performed in the frequency domain after the Fourier transform of the image.

  Here, the concept of filter processing in the frequency domain will be described with reference to FIGS. 15A to 15C. The horizontal axis of the graph of FIG. 15A is the spatial frequency ω [cycles / mm], the vertical axis is MTF, and the value of MTF when ω = 0 is normalized to 1. When a subject is photographed using the image detector 3 composed of pixels with a pixel size d = 100 [μm], the Nyquist frequency of the image detector 3 is 5 [cycles / mm]. Normally, in the case of the image detector 3 having a pixel size of d = 100 [μm], an image is detected with the characteristics indicated by the broken line in FIG. 15A. Here, the resolving power depends on the Nyquist frequency of the image detector 3.

  Even if the pixel size is d = 100 [μm], when the DEMOT method is used for image processing, the cutoff frequency of the MTF is theoretically 10 [cycles / mm] as shown by the solid line. In this case, the resolving power depends on a value twice the Nyquist frequency of the image detector 3. However, the characteristics of the MTF are curves that protrude downward in the range where the spatial frequency ω is 5 to 10 [cycles / mm], as shown by a solid line in FIG. 15A.

  Normally, when the pixel size is d = 100 [μm], there is no MTF value in the range where the spatial frequency ω is larger than 5 [cycles / mm] as shown by a broken line in FIG. 15A. Therefore, if an attempt is made to increase the MTF gain at a cutoff frequency (5 [cycles / mm]) corresponding to the resolving power, the granularity of the image deteriorates and the image cannot be used very much for diagnosis. Here, increasing the gain of the MTF means that the value of the MTF for a predetermined spatial frequency ω is made larger than that of the original image by image processing on the original image.

  On the other hand, in the X-ray imaging apparatus 1, for example, as shown in FIG. 3B, one pixel of the image detector 3 corresponds to the 10 × 10 memory set in the large-capacity memory storage unit 8. Yes. Then, as shown by the broken line in FIG. 3C, when the image is viewed in the frequency domain, the same effect as that obtained by increasing the value of the cutoff frequency of the MTF by 10 times the actual value is obtained. Therefore, in photographing with a pixel of 100 [μm], there is actually no MTF value in the range where the spatial frequency ω is larger than 5 [cycles / mm], but in the expanded memory space, the spatial frequency ω is It is possible to consider MTF information up to 50 [cycles / mm].

  A special filter such as a band-pass filter using the 50 [cycles / mm] band can be assumed. This filter may be a Laplacian filter. FIG. 15B is a diagram illustrating an example of a bandpass filter that emphasizes a range where the spatial frequency ω is 5 to 10 [cycles / mm]. The horizontal axis of the graph of FIG. 15B is the spatial frequency ω [cycles / mm], the vertical axis is the relative intensity, and the intensity when ω = 0 is normalized to 1.

  By convolving the bandpass filter shown in FIG. 15B with the normal MTF shown by the dashed curve in FIG. 15A, it is possible to improve the MTF and obtain an image with excellent graininess and more visibility. it can. Specifically, when the band pass filter shown in FIG. 15B is used, the image processing means 10 operates in the following procedure. First, the image processing means 10 copies the input image by enlarging it M × M times in the memory space (coordinate space) set in the large-capacity memory storage means 8, and then Fourier-transforms this enlarged image. To do. Next, the image processing means 10 applies a band pass filter to the Fourier-transformed image in the frequency space. Then, the image processing means 10 performs inverse Fourier transform on the image generated by applying the band-pass filter, so that the density distribution of the enlarged image is set in the memory space (coordinate space) set in the large-capacity memory storage means 8. Ask for.

  Further, the filtering process in the frequency domain and the DEMOT method can be combined. FIG. 15C shows the result of convolution (superimposition integration) of the MTF in the case of using the DEMOT method indicated by the solid curve in FIG. 15A and the bandpass filter shown in FIG. 15B. The cut-off frequency of the improved MTF shown in FIG. 15C is 10 [cycles / mm], which is the same as the MTF using the DEMOT method indicated by the solid line in FIG. 15A. Therefore, the resolution has not changed. However, paying attention to the area surrounded by the improved MTF curve, the vertical axis, and the horizontal axis shown in FIG. 15C, the spatial frequency ω in this area is 5 to 5 in the image after convolution as compared with the original image. The area of the region increases so as to expand toward the larger MTF in the range of 10 [cycles / mm]. That is, the gain of the MTF can be increased in a low frequency region lower than the cutoff frequency corresponding to the resolving power. Therefore, it is possible to obtain an image that is excellent in graininess and easier to visually recognize while super-resolution at low cost. Thereby, an image that can be used for diagnosis can be obtained without changing the resolution.

  The X-ray imaging apparatus 1 has an expanded memory space corresponding to a coordinate space obtained by enlarging an image M × M in the large-capacity memory storage unit 8 regardless of whether the DEMOT method is used or not. Since such a large memory has a very large band of high-frequency components, the original MTF can be changed to an improved MTF.

  In the second embodiment, the image processing performed in the memory space set in the large-capacity memory storage means 8 is used together with the image processing of the DEMOT method that improves the resolving power. Is not required. In the first and second embodiments, the frame image is subjected to the frequency conversion filter processing in the memory space set in the large-capacity memory storage unit 8. Instead, the panoramic tomographic image (image to be displayed) is applied. Frequency conversion filter processing may be performed, or both frame images and panoramic tomographic images may be subjected to frequency conversion filter processing. Furthermore, the present invention can be used not only for panoramic tomographic images but also for improving the granularity of X-ray CT images. For example, in the case of an X-ray CT image, frequency conversion filter processing may be performed on the CT image that is finally displayed in the memory space set in the large-capacity memory storage unit 8.

In addition, although the horizontal division number M and the vertical division number N of the recording area for each pixel set in the large-capacity memory storage unit 8 have been described as being equal, the values of M and N may be different. For example, the division number M is not limited to 10 and is arbitrary.
When the DEMOT method is used, the number of horizontal divisions and the number of vertical divisions of one pixel of the image detector 3 may be different, but if they are equal, the vertical resolution and the horizontal resolution are the same in the obtained image. Therefore, it is preferable. Further, the number of pixel divisions is not limited to 10 and is arbitrary. Considering the processing load and processing time of the apparatus, for example, it is preferably 3-20, and more preferably 5-10.

  Moreover, although each embodiment demonstrated the dental X-ray imaging apparatus 1, this invention is not limited to dental X-ray imaging, The granularity of a general medical image is improved. Can be used for For example, it may be applied to a chest X-ray imaging apparatus for internal medicine. In the present invention, the subject is not limited to the human body, and may be, for example, a naturally occurring object such as a mineral or a product of various industries. In this case, various types of analysis and inspection for damage can be performed.

  In each of the embodiments, the X-ray imaging apparatus 1 is described as including the image processing apparatus 20, but the present invention has the same effect even if the X-ray imaging apparatus 20 is a single image processing apparatus 20 separated from the X-ray imaging means. Play. The present invention can also be applied to images taken with a general camera. The present invention can be applied not only to an image captured using radiation, but also to an image captured using visible light, infrared light, or the like.

DESCRIPTION OF SYMBOLS 1 X-ray imaging apparatus 2 X-ray source 3 Image detector 4 Arm 5 Rotation drive means 7 Large-capacity frame image storage means 8 Large-capacity memory storage means 9 All image display storage means 10 Image processing means 11 Output means 12 Subject 18 Memory storage Means 20 Image processing device 63 FPD
64 Lead edge 65 Moving direction 66 X-ray d Pixel size O Rotation center

Claims (4)

An image processing apparatus that performs predetermined image processing on an image obtained by photographing a subject,
The image has a recording area for each pixel that is aligned with the coordinate position of each pixel in the image, and the recording area for each pixel is divided into M rows (M is an integer of 2 or more) horizontally, and N columns (N is Large-capacity memory storage means in which an expanded memory space corresponding to a coordinate space obtained by dividing the image by M × N times by dividing the image into two or more integers) is set;
M × N memories corresponding to the recording area of the pixel set in the expanded memory space for each pixel are transferred to the large-capacity memory storage unit. Image processing means for writing the same signal value to the image and performing frequency conversion filter processing on the image expanded in the expanded memory space;
Equipped with a,
As the frequency conversion filter processing, the image processing means converts the signal value of the pixel in the input image into M × N memories corresponding to the recording area of the pixel set in the expanded memory space. In addition to recording, the signal value of the pixel in the input image is shifted from the recording area of the pixel in the input image to the upper, lower, left, and right within a range smaller than the recording area of the pixel. An image processing apparatus , wherein the moving average process is performed by recording in the above M × N memories and adding the signal values recorded in the respective memories . The image processing means obtains an average density of M × N memories for each recording area of the pixels in the expanded memory space after performing the frequency conversion filter processing, and each image in the input image The image processing apparatus according to claim 1, wherein the image processing apparatus replaces the pixel signal value. The image processing apparatus according to claim 1 or 2,
An X-ray source;
An image detector comprising an X-ray image detection recording medium having a pixel array in which pixels that receive X-rays passing through a predetermined point of a subject are two-dimensionally arranged;
Driving means for moving the image detector in a direction orthogonal to the X-ray incident direction;
Frame image storage means for storing a frame image sent from the image detector at a predetermined frame rate,
In the large-capacity memory storage means, an expanded memory space corresponding to a coordinate space obtained by enlarging the frame image by M × N is set.
X-ray imaging characterized in that the image processing means transfers the frame image to the large-capacity memory storage means and performs the frequency conversion filter processing on the frame image enlarged in the expanded memory space apparatus. The image processing apparatus according to claim 1 or 2,
An X-ray source;
An image detector comprising an X-ray image detection recording medium having a pixel array in which pixels that receive X-rays passing through a predetermined point of a subject are two-dimensionally arranged;
Driving means for moving the image detector in a direction orthogonal to the X-ray incident direction;
Frame image storage means for storing a frame image sent from the image detector at a predetermined frame rate,
With respect to the direction in which the image detector is moved by the driving means, the arrangement direction of any one column of pixel groups in the pixel array of the image detector is arranged to be inclined,
In the large-capacity memory storage means, an expanded memory space corresponding to a coordinate space obtained by enlarging the frame image by M × N is set.
The image processing means transfers the frame image to the large-capacity memory storage means, and when the image detector is moved minutely by a distance smaller than the pixel width with respect to the direction in which the pixel group of the row is inclined. The output signal of the pixel of the image detector is proportionally written sequentially into M × N memories corresponding to the position of the pixel in the expanded memory space according to the moving position, and the output signal of the pixel The X-ray imaging apparatus is characterized in that the frequency conversion filter processing is sequentially performed on the enlarged image in which the frequency conversion filter is divided, and the signal value after the frequency conversion filter processing is added to each memory.

Images