JP4707471B2 - Image processing program, apparatus and method - Google Patents

Description

  The present invention relates to an image processing program, apparatus, and method, and more particularly, to an adaptive filtering processing technique for reducing image noise.

  There are various factors that determine the quality of a medical image, but in general, resolution, contrast, noise, false images, and the like are used as indicators.

  Basic characteristics such as spatial resolution greatly depend on the performance of a series of hardware for acquiring medical images. For example, in the case of a medical imaging apparatus for medical use, the fineness (aperture width) of the detector, the sampling interval, and the focus size in the case of a device using X-rays. Noise depends not only on hardware performance, but also on the imaging environment and the subject.

  Further, when imaging a large patient with a medical X-ray diagnostic apparatus, the influence of measurement system noise cannot be ignored because the attenuation of X-rays is large. On the other hand, increasing the X-ray intensity in a medical X-ray diagnostic apparatus has a problem of increasing the radiation exposure of the subject. Therefore, it is desired to reduce image noise without increasing the exposure dose of the subject.

  Conventionally, various methods for reducing noise by image processing have been proposed in response to such demands.

  In the patent document 1, the inventors change the matrix size according to the feature amount (standard deviation or the like) of the local region of the image, and further determine the image filtering coefficient based on the density distance from the density average value of the pixels. Disclosed. This method has a great effect on noise reduction, and can also be expected to reduce the exposure dose of the subject.

  Japanese Patent Application Laid-Open No. 2004-228561 discloses a technique for varying the coefficient of an image filter based on fuzzy inference.

  Patent Document 3 discloses a technique for improving (improving) the contrast of an image by smoothing the appearance frequency of pixels based on a histogram equalization method.

  Japanese Patent Application Laid-Open No. 2004-228561 discloses a method of identifying background density values using a histogram and removing unnecessary background information.

Non-Patent Document 1 describes a technique related to edge-preserving smoothing processing capable of preserving edges reliably. In this method, a plurality of masks are set in the local region, and when the edge is included in the mask, adjustments such as decreasing the degree of smoothing are performed.
JP 2003-225234 A JP-A-5-91532 JP-A-10-2933843 JP 2003-250046 A “Image Analysis Handbook” Mikio Takagi et al., University of Tokyo Press p542 Edge Preserving Smoothing

  However, the conventional adaptive filter has a problem that when the filter is strengthened, mosaic pattern noise appears and the image quality becomes unstable or the control becomes complicated due to the large number of filter parameters. Further, the technique related to the edge preserving smoothing process of Non-Patent Document 1 has a problem that an unnatural impression is given to the entire processed image when the strength of smoothing is increased.

  The present invention has been made in view of such circumstances, and it is possible to easily control the smoothing strength or the enhancement strength without determining complicated parameters in the adaptive filter processing, and to generate mosaic pattern noise. An object of the present invention is to provide an image processing program, an apparatus, and a method capable of reducing image noise while suppressing and obtaining a stable high-quality medical image.

  To achieve the above object, the present invention provides a reading step for reading a medical image, a local region setting step for setting a local region for each pixel of interest included in the medical image, and a local region for each local region. A determination step of determining a section width and the number of sections for generating a histogram indicating an appearance frequency of a pixel value included in the pixel value, and among the pixels constituting the local area, the pixel value section includes the pixel value section A generation step of measuring the number of pixels to obtain the appearance frequency and generating the histogram, and at least one pixel value section including the pixel value section to which the pixel value of the pixel of interest belongs in the histogram For each of the pixel value intervals included in the action interval and the action interval setting step to be set as, the appearance frequency of the pixel value interval is set as a relative weight. A filtering step of outputting a result pixel value that is a result of weighted addition to any one of a median value or an average value of the pixel value section, and replacing the pixel value of the pixel of interest with the result pixel value; It is made to perform.

  The section width and the number of sections determined in the above “determining step” may be determined based on the pixel value of the local area, or the average value of the pixel values of the local area is calculated and the average value and the local area are calculated. The section width and the number of sections may be obtained based on the difference value with the pixel value of each pixel included. Similarly, the histogram may be generated based on the pixel value of the local region, or may be generated based on a difference value from the average value of the pixel values of the local region.

  The “action interval” is a pixel value interval determined in accordance with the purpose of the pixel processing program according to the present invention, and is set as a smoothing interval when smoothing filtering is intended. When it is intended, it is set as an emphasis section.

  The “local region” may be a provisional local region, a standard deviation of pixel values of the local region may be calculated, and the size of the local region may be changed according to the standard deviation.

  In the determining step, a processing range indicating a range of pixel values for which the histogram is to be generated is set in the determining step, and when the pixel value of the pixel of interest belongs to the processing range, When the histogram is generated and the pixel value of the pixel of interest does not belong to the processing range, the pixel value corresponding to the pixel of interest is obtained in the filtering step without generating the histogram in the generation step. It outputs as a pixel value, It is characterized by the above-mentioned.

  The “processing target range” may be obtained by multiplying the section width by the number of sections, or may be determined based on the pixel value of the local area, or the average value of the pixel values of the local area is calculated. However, it may be based on a difference value between the average value and the pixel value of each pixel included in the local region.

  Further, the present invention is characterized in that, in the determining step, a standard deviation of pixel values of the local region is calculated, and a section width and a number of sections are determined according to the standard deviation.

  Further, according to the present invention, in the action interval setting step, a provisional action interval is set, a frequency of appearance of each pixel value interval included in the provisional action interval is added to calculate a total frequency, and according to the total frequency The actual action section is determined by changing the number of sections constituting the provisional action section.

  The “provisional action section” refers to an action section that is determined in order to calculate the total frequency. The “actual action section” means an action section that is a processing target in the filtering step. When the total frequency of the “provisional action interval” is appropriate, the provisional action interval and the actual action interval are the same, and when the total frequency is inappropriate, the provisional action interval and the actual action interval are Becomes a different section by changing the number of sections.

  Further, the present invention provides a reading unit that reads a medical image, a local region setting unit that sets a local region for each pixel of interest included in the medical image, and a pixel value included in the local region for each local region. Determining means for determining a section width and the number of sections of a pixel value section for generating a histogram indicating an appearance frequency, and measuring the number of pixels included in each pixel value section among the pixels constituting the local region. Generating means for determining the appearance frequency and generating the histogram, and action interval setting for setting at least one pixel value interval including the pixel value interval to which the pixel value of the pixel of interest belongs in the histogram as an action interval Means and, for each pixel value section included in the action section, either the median value or the average value of the pixel value section using the appearance frequency of the pixel value section as a relative weight. Outputs are the result of processing weighted addition in one result pixel values, and a filtering means for replacing the pixel value of the target pixel in the resultant pixel value.

  The present invention also includes a reading step for reading a medical image, a local region setting step for setting a local region for each pixel of interest included in the medical image, and a pixel value included in the local region for each local region. A determination step for determining a section width and a section number of a pixel value section for generating a histogram indicating an appearance frequency, and measuring a number of pixels included in each pixel value section among pixels constituting the local area. A generation step of obtaining the appearance frequency and generating the histogram, and an action interval setting for setting at least one of the pixel value intervals including the pixel value interval to which the pixel value of the pixel of interest belongs in the histogram as an action interval For each pixel value interval included in the step and the action interval, the center of the pixel value interval is expressed as a relative weight with the appearance frequency of the pixel value interval. Or it outputs the result in a result the pixel values of the processing weighted addition to any one of the average value, characterized in that it comprises a filtering step of replacing the pixel value of the target pixel in the resultant pixel value.

  According to the present invention, it is possible to easily control the smoothing strength or the enhancement strength without determining complicated parameters in the adaptive filtering process, and to reduce image noise while suppressing generation of mosaic pattern noise, It is possible to provide an image processing program, apparatus, and method capable of obtaining a stable high-quality medical image.

  Hereinafter, preferred embodiments of an image processing program, apparatus, and method according to the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a schematic hardware configuration diagram of an image processing system using an image processing program, apparatus, and method according to an embodiment of the present invention. The image processing system 1 includes a medical image capturing device 2, a medical image database 3 that stores medical images captured by the medical image capturing device 2, and an image processing device 10, and includes the medical image capturing device 2 and the medical image database. 3 and the image processing apparatus 10 are connected to each other via the network 4. Hereinafter, the X-ray CT apparatus 2 will be described as an example of the medical image photographing apparatus 2.

  The image processing apparatus 10 includes a central processing unit (CPU) 11 that mainly controls the operation of each component, a main memory 12 that reads a control program of the image processing apparatus 10 and medical image data and serves as a work area, and an image processing apparatus. A data storage device 13 for storing various application software such as 10 control programs and medical image display programs, medical image data, and the like; a display memory 14 for temporarily storing display data; and medical image data from the display memory 14 A display 15 for displaying a medical image based on the above, a pointing device such as a mouse 16, a trackball, and a touch panel for operating a soft switch on the display 15, a controller 16a of the pointing device, keys for setting various parameters, Keyboard with switch 17, and the image processing apparatus 10 a local area network, a telephone line, a network adapter 18 for connecting to a network 4 such as the Internet, from the data bus 19 which connects the above components. The data recording device 13 includes a storage device such as a memory or a magnetic disk built in or external to the image processing device 10, and a device that writes or reads data to or from a removable external medium.

  Next, a program executed by the CPU 11 of the image processing apparatus 10 will be described with reference to FIG. FIG. 2 is a block diagram showing the configuration of the program.

  The CPU 11 executes an image reading unit 11a, a local region setting unit 11b, a feature amount calculation unit 11c, a parameter determination unit 11d, a histogram generation unit 11e, an action interval setting unit 11f, a filtering unit 11g, and a display control unit 11h, which are programs. To do. These programs are stored in the data storage device 13, and the CPU 11 reads the program to the main memory 12 and executes it as appropriate.

  The image reading unit 11 a reads a medical image from the medical image database 3, the data storage device 13, or the X-ray CT apparatus 2. Hereinafter, “medical image” is referred to as “image”. In the present embodiment, the pixel value is a density value.

  The local area setting unit 11b sets a local area for each target pixel included in the image.

  The feature amount calculation unit 11c calculates a local feature amount (average value, variance, standard deviation, maximum value, minimum value, etc.) of the density value of the local region.

  The parameter determination unit 11d determines, based on the local feature amount calculated by the feature amount calculation unit 11c, the section width, the number of sections, the median value, and the like, which are histogram parameters indicating the frequency of appearance of the density value of the local region.

  The histogram generation unit 11e generates a histogram of the local region based on the parameter determined by the parameter determination unit 11d.

  The action interval setting unit 11f sets an action interval that is an object of adaptive filtering processing described later in the histogram.

  The filtering unit 11g performs an adaptive filtering process on the action section set by the action section setting unit 11f, and outputs the result pixel value indicating the processing result by replacing it with the target pixel value.

  The display control unit 11 h performs display control for displaying the processing result on the display 15.

  Next, the adaptive filtering process in the image processing apparatus 10 configured as described above will be described with reference to FIG. FIG. 3 is a flowchart showing the flow of the adaptive filtering process. The CPU 11 controls the image processing apparatus 10 according to this flowchart.

(Step S1)
In S1, the image reading unit 11a reads an image from the medical image database 3, the data storage device 13, or the X-ray CT apparatus 2 (S1).

(Step S2)
In S2, the local region setting unit 11b sets the matrix size of the local region for each pixel of interest included in the image (S2). The matrix size is arbitrary, and the matrix size may be constant in the image or may be changed for each pixel of interest.

(Step S3)
In S 3, the feature amount calculation unit 11c has an average value of pixel density in each local area, standard deviation, variance, differential value, and calculates the feature quantity indicating the feature of local areas such as (S3).

  The standard deviation of the density value indicates whether the density value distribution in the local region is wide or narrow. In the case of an X-ray CT medical image, when the density value distribution is large, it is in the vicinity of a high X-ray absorber such as a bone, that is, an edge part where the density change is large, and when the density value distribution is small, This is a region with subtle changes in density.

  The differential value of the density value indicates the amount of change in the density value of the local region. If the differential value is large, the presence of an edge in the local region is suggested. If the presence of an edge is suggested, the matrix size of the local region may be changed.

(Step S4)
In S4, the parameter determination unit 11d calculates a histogram parameter as preprocessing for generating a histogram indicating the appearance frequency of the density value included in the local region. Specifically, the average value of the pixel density obtained in S 3, determines the number of segments and the section width is a histogram parameters based on local features such as standard deviation (S4).

  The number of sections in the histogram is a class number indicating how many times the variation in the pixel density in the local region is divided to count the frequency. The section width indicates the density value width per section, that is, how much density value range one section of the histogram, that is, one class has.

(Step S5)
In S5, a processing range W indicating a range of density values to be generated in an S7 histogram to be described later is set (S5). The width of the density value in the processing range W is obtained by multiplying the section width (density width per section) by the number of sections in the processing range W.

  Note that the processing target range W may be determined based on the density value of the local region. When the density value of the pixel of interest is a specific density value, for example, a density value corresponding to the CT value of bone, the histogram generation process is not performed and the density value may be stored as it is.

(Step S6)
In S6, it is determined whether or not the density value of the pixel of interest is included in the processing range W set in S5 (S6). If the density value is not included in the processing range W, the process proceeds to S7. If the density value is included in the processing range W, the process proceeds to S8.

(Step S7)
In S7, if the density value of the target pixel is not included in the processing range W obtained in S 5, the density value is saved. The stored density value is output to the coordinate position of the pixel of interest as a result pixel value representing the result of image processing (S7).

(Step S8)
In S5, the histogram generator 11e generates a histogram of the local area (S8). The histogram generation unit 11e measures the number of pixels included in each section according to the number of sections and the section width of the histogram obtained in S4, and obtains the appearance frequency for each section.

  Note that the processes of S4 to S7 may be performed based on a difference value between an average value of pixel values in the local area and a pixel value of each pixel included in the local area. Hereinafter, the “difference value between the average value of the pixel values in the local area and the pixel value of each pixel included in the local area” is referred to as a “difference value”.

  That is, the section width and the number of sections in S4 may be set based on the difference value. Similarly, the processing range in S5 may be set based on the difference value. Similarly, the histogram in S7 may be generated based on the difference value.

  FIG. 4 is a diagram illustrating an example of a local area histogram. The horizontal axis of the histogram is a section (density section) based on the difference value. The vertical axis indicates the appearance frequency of pixels included in each section. The processing range W in the figure is the processing range set in S5, and includes at least one class in the histogram.

(Step S9)
In S9, the action section setting unit 11f sets at least one section including the section to which the pixel value of the pixel of interest belongs in the histogram as the action section (S9). The action interval is a pixel value interval determined according to the purpose. The action interval is set as a smoothing interval when the purpose is smoothing filtering, and is set as an enhancement interval when the purpose is enhancement filtering.

  The action section setting unit 11f obtains a section to which the density value of the target pixel belongs. Then, the action interval setting unit 11f sets an action interval for performing adaptive filtering processing in S10 described later. For example, when the density value of the target pixel belongs to the section m and the number of action sections is set to 3, the action sections are the three sections m-1, m, and m + 1, and these three sections are the processing targets ( (See FIG. 4).

(Step S10)
In S10, the filtering unit 11g outputs, for each section included in the action section, a result pixel value that is a result of weighted addition to the median value of the section using the appearance frequency of the section as a relative weight. A process of replacing the pixel density value with the result pixel value is performed (S10).

  If the median value of the section is M, the resulting pixel value R is obtained by the following equation. The median indicates the density value at the center of each section. For example, when the density value range of the section m is 10 to 15, it is 12.5.

  Multiplying each section median M (m−1), M (m), M (m + 1) by using the frequency C (m−1), C (m), C (m + 1) of each section as a relative weight, Find the sum. Then, the sum is divided by the sum of the frequencies of the three sections. Note that the average value of the section may be used instead of the median value M.

  In the above example, the action sections are m-1, m, m + 1, and the number of sections is 3, but can be arbitrarily set in advance in consideration of the presence of an edge.

    Then, the filtering unit 11g outputs the result pixel value R to the coordinate position of the target pixel.

(Step S11)
In S11, it is determined whether or not the processing from steps S2 to S10 has been performed up to the final pixel of the image (S11). If the process has not been performed up to the last pixel, the process returns to S2. If the process has been performed up to the last pixel, the process proceeds to S12.

(Step S12)
In S12, the display control unit 11h displays an image (S12), and the process ends.

  According to the present embodiment, since the action interval is set as shown in Equation 1, the smoothing strength or the enhancement strength can be easily controlled. Further, since the process is performed by weighting the appearance frequency, the noise can be reduced while suppressing the generation of pattern noise, and the medical image can be kept in a natural impression while preserving the edges.

<Second embodiment>
In 2nd embodiment, the action area in S9 is adjusted suitably. There are various methods for making the action interval adaptive. For example, in the case of a CT image, since the pixel value is defined as an absolute value for each tissue, a method of appropriately setting an action interval for each tissue is conceivable. In addition, there is a method based on a histogram frequency described as an example below.

  FIG. 5 is a flowchart showing a flow of processing according to the second embodiment. Hereinafter, the process of the second embodiment will be described with reference to FIG.

(Steps S501 to S508)
In S501 through S508, the same processing as in S1 through S8 is performed to calculate a histogram of the local region. If the density value of the target pixel is not included in the processing range W, the process proceeds to S513, and the density value of the target pixel is output as a result pixel value R to the coordinate position of the target pixel (S501 to S508).

(Step S509)
In S509, a provisional action section is set (S509). The “provisional action section” is an action section that is provisionally set for calculating the total frequency of the action sections.

(Step S510)
In S510, the total frequency within the provisional action section is calculated (S510).

(Step S511)
In S511, it is determined whether the total frequency of the provisional action section provisionally set in S510 is appropriate (S511). For example, if the total frequency of the provisional action interval is 9, an average effect of 9 pixels is obtained, that is, the same as the smoothing effect of the 3 × 3 pixel matrix filter. Therefore, the degree of smoothing can be made appropriate by changing the provisional action section according to the total frequency.

  The total frequency of the provisional action interval is, for example, 9 or more when the same effect as the smoothing effect of the 3 × 3 pixel matrix filter is expected, and the same effect as the smoothing effect of the 4 × 4 pixel matrix filter. For example, 16 or more is set.

  When the total frequency of the provisional action section is appropriate, the provisional action section and the actual action section are the same, and the process proceeds to S512. If the total frequency of the provisional action section is inappropriate, the process returns to S509.

(Step S512)
In S512, similarly to S10, the filtering unit 11g performs a filtering process using Equation 1, and outputs the resultant pixel value R to the coordinate position of the pixel of interest (S512).

(Steps S513 to S514)
In S513 to S514, as in S11 to S12, it is determined whether or not the processing from Step S502 to S512 has been performed on the final pixel of the image (S513). If the processing has not been performed on the final pixel, the process returns to S502. . If the process has been performed up to the final pixel, the display control unit 11h displays an image (S514), and the process ends.

  According to the present embodiment, the provisional action section is set according to the total frequency, and therefore the degree of smoothing strength or enhancement strength can be appropriately changed. Further, since the process is performed by weighting the appearance frequency, the noise can be reduced while suppressing the generation of pattern noise, and the medical image can be kept in a natural impression while preserving the edges.

Hardware configuration diagram of the present invention Program block diagram of the present invention Flow chart showing an example of filtering processing Example diagram explaining local area histogram Flow chart showing a processing example of the second embodiment

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Medical image processing system, 2 ... Medical image imaging device, 3 ... Medical image database, 4 ... Network, 10 ... Image processing device, 11 ... CPU, 12 ... Main memory, 13 ... Data storage device, 14 ... Display memory, 15 ... Display, 16 ... Mouse, 16a ... Controller, 17 ... Keyboard, 18 ... Network adapter, 19 ... Data bus

Claims (7)

A reading step for reading a medical image;
A local region setting step for setting a local region for each pixel of interest included in the medical image;
A determination step for determining a section width and a section number of a pixel value section for generating a histogram indicating an appearance frequency of a pixel value included in the local area for each local area;
Of the pixels constituting the local region, the number of pixels included in each pixel value section is measured to determine the appearance frequency, and the generation step of generating the histogram,
In the histogram, an action interval setting step for setting, as an action interval, at least one of the pixel value intervals including the pixel value interval to which the pixel value of the pixel of interest belongs,
Filtering the pixel of interest using only pixels that are included in the local region and whose pixel values are included in the action interval ; and
An image processing program for causing a computer to execute. In the determining step, a processing range indicating a range of pixel values for which the histogram is to be generated is set,
When the pixel value of the target pixel belongs to the processing range, the histogram is generated in the generation step,
When the pixel value of the target pixel does not belong to the processing range, the pixel value of the target pixel is stored without generating the histogram in the generation step.
The image processing program according to claim 1, wherein: In the determining step, the standard deviation of the pixel value of the local region is calculated, and the section width and the number of sections are determined according to the standard deviation.
The image processing program according to claim 1 or 2, characterized in that In the action section setting step, a provisional action section is set, the appearance frequency of each pixel value section included in the provisional action section is added to calculate a total frequency, and the provisional action section is configured according to the total frequency Change the number of intervals to determine the actual action interval,
The image processing program according to any one of claims 1 to 3. In the filtering step, for each pixel value section included in the action section, the appearance frequency of the pixel value section is used as a relative weight, and the weighted addition is performed on either the median value or the average value of the pixel value section. Output a result pixel value that is a result of the processing, and replace the pixel value of the target pixel with the result pixel value.
The image processing program according to any one of claims 1 to 4, wherein Reading means for reading medical images;
Local region setting means for setting a local region for each pixel of interest included in the medical image;
Determining means for determining a section width and a section number of a pixel value section for generating a histogram indicating an appearance frequency of a pixel value included in the local area for each local area;
Generation means for measuring the number of pixels included in each pixel value section among the pixels constituting the local region to obtain the appearance frequency, and generating the histogram;
In the histogram, action section setting means for setting, as an action section, at least one pixel value section including the pixel value section to which the pixel value of the pixel of interest belongs,
Image processing comprising: a pixel included in the local region, and a filtering unit that performs filtering on the pixel of interest using only a pixel whose pixel value is included in the action interval. apparatus. A reading step for reading a medical image;
A local region setting step for setting a local region for each pixel of interest included in the medical image;
A determination step for determining a section width and a section number of a pixel value section for generating a histogram indicating an appearance frequency of a pixel value included in the local area for each local area;
Of the pixels constituting the local region, the number of pixels included in each pixel value section is measured to determine the appearance frequency, and the generation step of generating the histogram,
In the histogram, an action interval setting step for setting, as an action interval, at least one of the pixel value intervals including the pixel value interval to which the pixel value of the pixel of interest belongs,
Filtering the pixel of interest using only pixels that are included in the local region and whose pixel values are included in the action interval; and
An image processing method comprising:

Images