JP5726482B2 - Image processing apparatus, X-ray diagnostic apparatus using the same, and operation method of image processing apparatus - Google Patents

Description

  The embodiment relates to an image processing technique.

  An X-ray image taken by an apparatus such as an X-ray fluoroscopic image apparatus or an X-ray TV is used for operating a medical instrument such as a catheter, a guide wire, a stent, or a balloon, and confirming the position of the instrument in the body. That is, the X-ray image is used for image diagnosis and navigation during treatment. For this reason, the visibility of an X-ray image is very important for a doctor who operates a medical instrument.

  When the X-ray image to be irradiated is increased, the signal-to-noise ratio is increased and the visibility is improved. However, increasing the X-ray dose also increases patient exposure. For this reason, an attempt is made to improve the visibility without changing the X-ray dose by applying image processing to the acquired image.

  A noise reduction method has been proposed as a method for improving visibility by image processing. However, the noise reduction method sometimes weakens the signal component together with the noise component in the image. On the other hand, for the purpose of improving visibility, a sharpness improving method by image processing has been proposed, but noise components may be emphasized together with signal components.

  In order to solve such a problem, a method for simultaneously suppressing noise and enhancing a signal in an image has been proposed. This method is designed to selectively perform noise suppression and signal enhancement on each pixel of an image (see, for example, Patent Document 1).

  An image processing filter that simultaneously suppresses noise in an image and emphasizes sharpness has been proposed (see, for example, Non-Patent Document 1). This image processing filter has parameters capable of controlling the noise suppression degree and sharpness in the processed image. A response of a Laplacian of Gaussian filter (hereinafter referred to as “LoG filter”) for each pixel of the image is used as a feature amount, and parameters of the image processing filter are determined according to the feature amount.

  However, in the method of Patent Document 1, the processing for each pixel of the image is determined according to the pixel value of the target image. For this reason, it has not been possible to accurately determine which pixels should be emphasized and which pixels should be subjected to noise suppression processing. In addition, the LoG filter is an isotropic filter, and is insufficient for selectively processing a medical device in the body that is noticed in an X-ray image from the background.

  That is, in the prior art, there is a problem that it is difficult to improve the visibility of an object having a locally linear structure that exists in an image such as an X-ray image.

JP 2002-374418 A

Buyue Zhang, Jan P. Allebach "Adaptive Bilateral Filater for Sharpness Enhancement and Noise Removal", IEEE Transactions on image processing, vol. 17 No. 5, 2008.

  In view of this, an object of the embodiment is to improve the visibility of a linear pattern in a target image.

According to an embodiment, the image processing apparatus obtains a line value indicating a linear pattern like dark lines or bright lines having continuity in a specific direction for each pixel in the target image, and brightness of the detection target. And a linear pattern detection unit that sets one of a negative value and a positive value to zero in the line value, and (1) a pixel value of a filter center pixel existing in the filter range; Regarding an image processing filter configured with a first weight function in which the first weight is reduced as the difference between pixel values of surrounding pixels is larger, (2) the first parameter and the second parameter in the first weight function (3) The first parameter defines the spread of the first weight function, and the larger the first parameter, the wider the spread, 4) The second parameter The center position of the spread of the first weight function is determined. The larger the second parameter is, the more the center position is shifted from the filter center pixel. (5) The positive line value is set to zero. Sometimes the larger the line value is in the negative direction, or when the negative line value is zero, the larger the line value is in the positive direction, the smaller the first parameter and the second parameter are. (6) a parameter determining unit that determines the first parameter and the second parameter for each pixel of the target image, the first parameter determined for each pixel of the target image, and the A filter processing unit configured to apply an image processing filter including the first weight function including a second parameter to each pixel of the target image to calculate a pixel value of each pixel of the output image; Including.

1 is a block diagram of an X-ray diagnostic apparatus having an image processing apparatus according to Embodiment 1. FIG. The figure explaining a X-ray fluoroscopic image and a linear pattern. The figure of a linear pattern detection filter. Explanatory drawing of an image processing filter. The figure showing the effect of the 1st weight function and parameter of an image processing filter. FIG. 4 is a block diagram of an image processing apparatus according to a second embodiment. FIG. 9 is a block diagram of an image processing apparatus according to a third embodiment. FIG. 10 is an explanatory diagram of an image processing filter according to a fourth embodiment. FIG. 10 is a diagram illustrating an effect of a first weighting function and parameters of an image processing filter in Embodiment 5.

  An image processing apparatus according to an embodiment of the present invention will be described below with reference to the drawings.

  The image processing apparatus according to the present embodiment is intended to improve the visibility of an object having a locally linear structure existing in an image. In Examples 1 to 4, the image processing apparatus has a linear structure. As an example of the object, a guide wire 22 used for catheter surgery is given, and in Example 6, a crack formed on a building wall and a defect on a semiconductor are given.

  An image processing apparatus 16 according to the first embodiment will be described with reference to FIGS.

  The guide wire 22 inserted into the patient's body is imaged by the X-ray diagnostic apparatus 1 during the catheter operation. The guide wire 22 is an object such as a thin wire and travels in the blood vessel of the patient. The doctor operates the guide wire 22 while observing an X-ray fluoroscopic image (hereinafter simply referred to as “X-ray image”) 20. For this reason, the visibility of the guide wire 22 in the image greatly affects the safety and efficiency of the operation. Naturally, the higher the visibility, the easier it is to grasp the position of the guide wire 22. Therefore, the image processing apparatus 16 of the present embodiment aims to improve the visibility of the guide wire 22 in the X-ray image.

  As shown in FIG. 1, the image processing apparatus 16 in the present embodiment is installed in the X-ray diagnostic apparatus 1. The X-ray diagnostic apparatus 1 includes an imaging apparatus 10 mainly composed of an X-ray generator 12 and an X-ray sensor 14 and can acquire an X-ray image G of a patient A. Further, an image display device 18 such as a CRT or a liquid crystal display is provided, and an X-ray image G taken by a doctor can be seen. The image processing device 16 of this embodiment is connected between the imaging device and the image display device 18. The image processing device 16 performs image processing on the X-ray image G obtained by the imaging device 10, outputs the processed image to the image display device 18, and presents the processed output image F.

  The configuration of the image processing device 16 will be described with reference to the block diagram of FIG. As illustrated in FIG. 1, the image processing apparatus 16 includes a linear pattern detection unit 100, a parameter determination unit 200, and a filter processing unit 300.

  The linear pattern detection unit 110 will be described with reference to FIGS. The linear pattern detection unit 110 detects the guide wire 22 in the X-ray image G. The guide wire 22 has a linear pattern of dark lines on the X-ray image G. The “dark line” refers to a linear pattern that is darker than other image areas and has continuity in a specific direction.

  The linear pattern will be described with reference to FIG. The “linear pattern” is a pattern of pixel values that appear linearly when a local region including the guide wire 22 is cut out. Since the guide wire 22 is thin enough to run in blood vessels throughout the body, it is drawn as a thin and dark line on the X-ray image G.

  FIG. 2A is a diagram of an X-ray image G displayed on the image display device 18 by imaging the chest of the patient A with the imaging device 10.

  FIG. 2B shows the distribution of the X-ray transmittance (pixel value) of the region 26 in the X-ray image G shown in FIG. 2A, and the X-ray transmittance is lower than the vicinity by the guide wire 22. The local pattern including the guide wire 22 has a profile like a valley that is darker than the periphery on the X-ray image G.

  FIG. 2C shows the distribution of the X-ray transmittance (pixel value) of the region 28 in the vicinity of the boundary of the object 24 in the X-ray image G shown in FIG. become. For example, the local pattern including the boundary between the heart and lung object 24 has a cliff-like profile.

  FIG. 2D shows the distribution of the X-ray transmittance (pixel value) of the region 30 that does not include the boundary of the object 24 and the guide wire 22 shown in FIG. 2A, and this cut-out local pattern is flat. Profile.

  In order to realize the visibility improving method of this embodiment, it is necessary to detect such a linear pattern separately from other patterns.

  Therefore, the linear pattern detection unit 100 uses a linear pattern detection filter that detects the profile of the linear pattern as described above.

  As a filter for detecting a linear pattern, for example, there is a LoG filter described in Non-Patent Document 1. The LoG filter is a second-order differential filter, and determines whether it is a linear pattern based on the response.

  However, since the LoG filter is an isotropic filter, the ability to selectively detect a linear pattern is not sufficient. An anisotropic filter is superior for detecting a linear pattern having continuity of pixel values in a specific direction.

  Therefore, in this embodiment, a linear pattern detection filter H (φ) as shown in FIGS. 3A to 3F is used, and the linear pattern has a width of one pixel in various directions. The rotation angle φ becomes a linear pattern detection filter in an arbitrary direction by entering an appropriate value. 3A to 3F, the filter size (x-axis, y-axis) is 9 × 9 pixels, the z-axis is the filter coefficient, and φ is a linear pattern detection filter created in increments of 30 degrees from 0 to 180 degrees. Indicates. Note that the linear pattern detection filter sets the sum of the filter coefficients to zero so that the response at the flat portion becomes zero.

  Next, a case where the linear pattern detection unit 100 detects a linear pattern using the linear pattern detection filter H (φ) described above will be described.

In order to detect this linear pattern, the concept of a value (hereinafter referred to as “line value”) α representing the linear pattern-likeness is adopted. And the linear pattern detection part 100 uses the following formula | equation (1) which has the linear pattern detection filter H ((phi)), and X-ray image G (henceforth "target image G") used as the object of filter processing. The line value α at each pixel is calculated. Here, since the linear pattern by the guide wire 22 in the target image G is a dark line pattern, the degree of the linear pattern increases as the line value α decreases. For example, the line value α is a negative value.

  However, g represents a pixel value (luminance value) of each pixel [m, n] of the target image G, and * represents a convolution operation. When the target image G is an image of M × N pixels, M−1> m> 0 and N−1> n> 0.

  When the filter processing is performed by the linear pattern detection filter H (φ), the pixel value of the pixel on the linear pattern darker than the surroundings such as the guide wire 22 becomes a large negative value. Since the traveling direction θ of the linear pattern varies depending on the position and orientation of the target object and the shooting direction, it cannot be determined in advance.

  Therefore, the linear pattern detection unit 100 obtains respective response values by applying the linear pattern detection filters H (φ) in various directions to the respective pixels [m, n] of the target image G, respectively. Let the minimum value be the line value α of each pixel [m, n]. The line value α is almost zero on the pixels of the flat portion, and the response increases in the negative direction on the pixels of the guide wire 22. Therefore, the pixels on the guide wire 22 are determined by the magnitude and the sign of the absolute value of the line value α. Can be determined. For example, if the line value α of a certain pixel is a negative value, it can be determined that the pixel is on a linear pattern. For the following description, in this embodiment, when the line value α is a positive value, it is determined that the line pattern α is not a dark linear pattern, and the value of the line value α is set to zero.

  Note that the filter size (number of pixels) of the linear pattern detection filter, the step size of the rotation angle φ, and the like are determined according to the target.

  In addition to the one-pixel-wide linear pattern detection filter shown in FIG. 3, a linear pattern detection filter having a plurality of line widths may be prepared so that linear patterns with various widths can be detected. it can.

  Next, the parameter determination unit 200 will be described. The parameter determination unit 200 uses the linear pattern detection unit 100 to set parameters (first parameter σ, second parameter ζ, etc.) of the filter coefficient (first weight) w of the image processing filter used in the filter processing unit 300. The detected line value α is used for each pixel of the target image G.

Before describing the method for determining the parameter of the first weight w, an image processing filter used in the filter processing unit 300 will be described. The image processing filter used in the filter processing unit 300 can be described by the following equation (2) as a weighted average of surrounding pixels.

  However, g is a pixel value (for example, luminance value) of each pixel [m0, n0] of the target image G to be filtered, and f is a pixel of each pixel [m0, n0] of the output image F after the filter processing. The value (luminance value), w is a first weight, z is a normalization term for weighted averaging, and is calculated by the sum of the first weight w.

  The filter range Ω is a range to which the image processing filter is applied with the filter center pixel [m0, n0] as the center.

  The peripheral pixels [m, n] of the filter center pixel [m0, n0] mean all the pixels existing in the filter range Ω and pixels other than the filter center pixel [m0, n0].

  With reference to FIG. 4, the filtering process represented by Expression (2) will be described. The pixel value f [m0, n0] of each pixel of the output image F is obtained by using the first weight w [m0, n0] as the pixel value g [m, n] of each peripheral pixel [m, n] in the filter range Ω. , M, n]. The first weight w [m0, n0, m, n] is calculated from the pixel value of the filter center pixel [m0, n0] and the surrounding pixels [m, n] in the filter range Ω as shown in the equation (3). Determined from the pixel value.

Here, the first weight w is obtained by a first weight function expressed by the following equation (3).

  The first weight function is a pixel value g [m0, n0] of the filter center pixel [m0, n0], each pixel value g [m, n] of the peripheral pixel [m, n], and the first weight spread. And a second parameter ζ [m0, n0] for determining the center of the spread of the first weight as variables.

  Here, the effects of the first parameter σ that determines the spread of the first weight and the second parameter ζ that determines the center of the spread of the first weight will be described with reference to FIG.

  FIG. 5A shows a 9 × 9 pixel size in which a linear shadow (pixel value 10) having a width of 1 pixel is placed on a flat portion having a pixel value 20 and Gaussian noise (standard deviation 1.0) is given to the entire image. This is a target image G. The pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G is 9.65.

  FIG. 5B is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G of FIG. 5A based on the equation (2), and the first parameter σ = 1. In this case, the second parameter ζ = −1. In FIG. 5B, since σ = 1 and ζ = −1, the peripheral pixel [1] lower than the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G Since the first weight w of m, n] is the largest, the pixel value f [m0, n0] of the output image F after the filtering process is a lower value than when ζ = 0. That is, the pixel value f [m0, n0] of the filter center pixel [m0, n0] of the output image F is 8.93. And the contrast with a flat part becomes large and visibility improves.

  FIG. 5C is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G of FIG. 5A based on the equation (2), and the first parameter σ = 1. In this case, the second parameter ζ = 0. Since σ = 1 and ζ = 0, the smaller the difference between the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G and the pixel value g [m, n] of the peripheral pixels is, the smaller the difference is. Since the weight w of 1 is increased, the first weight w applied to the filter center pixel [m0, n0] is larger than the peripheral pixels, and a value close to the pixel value g [m0, n0] of the target image G is Obtained as output after filtering. That is, the pixel value f [m0, n0] of the filter center pixel of the output image F after the filter processing is 9.55.

  FIG. 5D is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G in FIG. 5A based on the equation (2), and the first parameter σ = 3. In this case, the second parameter ζ = 0. Since σ = 3 and ζ = 0, the difference between the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G and the pixel value g [m, n] of the peripheral pixel [m, n] Since the change in the first weight w due to becomes dull, the weight of the pixel on the linear pattern becomes gentle and the smoothing effect increases, and the pixel value f [m0, n0] of the output image F is linear. It approaches the true value (= 10) of the pixel on the shadow. That is, the pixel value f [m0, n0] of the filter center pixel of the output image F after the filter processing is 9.89.

  Therefore, the first parameter σ defines the level of smoothing. For pixels that are not on the linear pattern, the noise reduction effect can be enhanced by increasing σ. That is, σ is reduced as the line value α increases in the negative direction.

  The second parameter ζ can shift the pixel value after the weighted average from the original pixel value. As shown in FIG. 5B, in the pixel on the linear pattern, the contrast can be enhanced by giving ζ in a lower direction than the peripheral pixel by comparing the pixel value on the linear pattern with the peripheral pixel value. . That is, ζ is increased as the line value α increases.

  In the parameter determination unit 200 of FIG. 1, in order to determine the first parameter σ and the second parameter ζ described above from the line value α, functions of α shown in Expressions (4) and (5) are used.

The first parameter σ is obtained for each pixel of the target image G from Expression (4) and the line value α.

The second parameter ζ is obtained for each pixel of the target image G from the equation (5) and the line value α.

  The constant t, the constant u, and the constant c in the expressions (4) and (5) are positive values, and are determined in advance according to the X-ray dose, the body thickness of the patient A, and the like. It is also possible to configure the apparatus so that the user can make settings while viewing the output image F after processing.

  Further, in addition to the method using Expression (4) and Expression (5), as a method for the parameter determination unit 200 to determine the first parameter σ and the second parameter ζ, a reference table using the line value α as an argument is used. It can be realized by preparing.

  The filter processing unit 300 uses the parameters determined by the parameter determination unit 200 and applies the image processing filter represented by the first weight function having the first weight represented by Expression (3) to the target image G. Thus, the filter process indicated by the first weight function is applied to Expression (2) to obtain an output image F.

  In other words, the filter processing unit 300 performs an image processing filter represented by the first weighting function with each pixel of all the pixels of the target image G as a filter center pixel. In this case, with respect to all the pixels of the target image G, all of the output images F are obtained using the pixel value g [m0, n0] of the filter center pixel [m0, n0], the first weight w, and Expression (2). The pixel value f [m0, n0] of the pixel [m0, n0] is obtained.

  Thereby, in addition to the first parameter σ and the second parameter ζ in the first weight w, it is determined by the pixel value of the filter center pixel of the target image G and the pixel value of the surrounding pixels. Adaptive filtering that cannot be realized by combination is possible.

  According to the present embodiment, an image processing filter having a large smoothing effect is applied to the pixel that is not considered to be the guide wire 22 due to the effect of the first parameter σ. On the other hand, for the pixel that seems to be the guide wire 22, the first weight is determined so that the contrast with the surrounding pixels is increased by the effect of the second parameter ζ. Therefore, since the contrast of the guide wire 22 can be improved while reducing the noise of the entire target image G, the visibility of the guide wire 22 is improved.

  Next, the image processing apparatus 16 according to the second embodiment will be described with reference to FIG.

  The configuration of the image processing apparatus 16 of the present embodiment will be described with reference to FIG. FIG. 6 shows a block diagram of the image processing apparatus 16 in the present embodiment.

  As illustrated in FIG. 6, the image processing apparatus 16 includes a linear pattern detection unit 100, a connection evaluation unit 110, a parameter determination unit 200, and a filter processing unit 300. Note that the linear pattern detection unit 100 and the filter processing unit 300 have the same configurations as those in the first embodiment, and thus description thereof is omitted.

The connection evaluation unit 110 will be described. The line value α of each pixel of the target image G is detected by the linear pattern detection filter of Expression (1) shown in the first embodiment. When a linear pattern is detected using an anisotropic filter in various directions such as Equation (1), the traveling direction θ of the linear pattern can be detected as shown in Equation (6).

  The connection evaluation unit 110 increases the accuracy of the linear pattern detection by evaluating the connection of the linear patterns using the traveling direction θ of the linear patterns. For example, since the guide wire 22 in the target line image G is a continuous line, the connection evaluation unit 110 determines whether or not the local linear pattern is continuous in the traveling direction, and noise that is mixed randomly. The direction of the linear pattern above is excluded.

  The connection evaluation unit 110 can determine whether the linear pattern is a continuous linear pattern or a linear pattern due to noise by evaluating the traveling direction θ of the linear pattern. Specific processing is as follows.

  First, the connection evaluation unit 110 recognizes the line value α of each pixel of the entire target image G from Expression (1), and detects the traveling direction θ from Expression (6).

  Next, the connection evaluation unit 110 sets one of the pixels of the entire target image G as a target pixel. The connection evaluation unit 110 detects the linear direction θx of the linear pattern at a pixel several pixels away from the linear pattern in the traveling direction θ at the target pixel.

  Next, if the linear direction θx matches the linear direction θ of the target pixel itself, or if it is within a specific angle B (that is, | θ−θx | <B). ), It is considered that the line segments of the pixel that is several pixels away from the target pixel are connected, and the line value α is decreased to correct the linear pattern. In Example 5 described below, in the case of a linear pattern of bright lines, correction is performed so that the degree of the linear pattern decreases by increasing the line value α.

  Next, the connection evaluation unit 110 moves the target pixel in order, and performs the above determination for all the pixels on the linear pattern of the target image G.

  Conversely, the connection evaluation unit 110 can increase the line value α if the traveling directions θ and θx do not match.

Next, the parameter determination unit 200 will be described. The parameter determination unit 200 creates an image processing filter having a second weight w2 for each pixel of the target image G based on the traveling direction θ of the linear pattern. Here, the second weight w2 can be described by a second weight function expressed by the following equation (7).

  In the second weight function of the equation (7), the second weight w2 of the peripheral pixels in the traveling direction θ of the linear pattern is high, and the second weight w2 of the peripheral pixels in the direction orthogonal to the traveling direction θ. Is determined so as to decrease with increasing distance from the filter center pixel.

  The parameter determination unit 200 determines the filter coefficient of the image processing filter of the filter processing unit 300 by multiplying the first weight w and the second weight w2. For example, this is effective when the contrast with the surroundings is not sufficient, or when the contrast cannot be sufficiently enhanced only with the first weight w.

  The second weight w2 can be determined using the line value α as a parameter in addition to the traveling direction θ of the linear pattern. For example, the parameter σd that determines the spread of the second weight function is determined according to the line value α. When the line value α is high (when the degree of likelihood of the linear pattern is low), σd is decreased to minimize the influence of pixels other than the linear pattern traveling direction θ. On the contrary, when the line value α is low (when the degree of the linear pattern is high), σd can be increased to enhance the smoothing effect.

  For the determination of σd, functions such as those shown in the equations (4) and (5) can be used, or can be realized by a table using the line value α as an argument.

  Next, the image processing apparatus 16 according to the third embodiment will be described with reference to FIG.

  When trying to detect the line value α, the line value α may be obtained from a pattern other than the linear pattern. For example, when Expression (1) is used, the line value α may increase in the negative direction even on the edge pattern (in some cases, the degree of the linear pattern is high). Further, the line value α may increase in the negative direction even on an isolated point where the change in the pixel value is large. Therefore, the image processing apparatus 16 of this embodiment solves this problem.

  The configuration of the image processing apparatus 16 of the present embodiment will be described with reference to FIG. FIG. 7 is a block diagram of the image processing apparatus 16 in the present embodiment.

  As shown in FIG. 7, the image processing apparatus 16 includes a linear pattern detection unit 100, a connection evaluation unit 110, an edge pattern detection unit 120, an isolated point pattern detection unit 130, a parameter determination unit 200, and a filter processing unit 300. In addition, since the linear pattern detection part 100, the connection evaluation part 110, the parameter determination part 200, and the filter process part 300 are the structures similar to Example 2, the description is abbreviate | omitted.

  The edge pattern detection unit 120 positively detects the edge pattern from the target image G in advance, and increases the line value α of the linear pattern obtained in the pixel in which the edge pattern appears to increase the degree of linear pattern likeness. Correct so that it falls. The edge pattern detection unit 120 can use an edge detection filter such as a Sobel filter that is common in the image processing field. Thereby, the line value α obtained from the edge pattern can be increased.

  Further, the isolated point pattern detection unit 130 positively detects the isolated point pattern from the target image G in advance, and increases the line value α obtained in the pixel in which the isolated point pattern appears to increase the degree of linear pattern likeness. Correct so that is lowered.

  In this embodiment, edge patterns and isolated point patterns are shown as patterns that are easily mistaken for linear patterns, but if there are other patterns that are easily mistaken, a pattern detection unit specialized for those patterns should be provided. May be.

  Further, the value obtained by the edge pattern detection unit 120 is defined as an edge likelihood value β, the line value α and the edge likelihood value β are handled independently, and the parameter determination unit 200 determines two parameters σ, ζ may be determined by a reference table having two values α and β as arguments.

  Next, an image processing apparatus 16 according to the fourth embodiment will be described with reference to FIG.

When the target image G is a time-series moving image, the image processing apparatus 16 according to the present embodiment extends the filter range Ω in the time direction in addition to the two-dimensional pixel position of the target image. As shown in FIG. 8, the number of pixels to be weighted averaged can be increased by extending the filter range Ω to an appropriate past time. This expansion method is particularly effective when the object being imaged does not move significantly between temporally adjacent images. The first weight w, which is a filter coefficient when the input is a moving image, is determined by expanding the first weight function of Expression (3) as Expression (8).

  Here, g [m, n, t] represents the luminance value at the pixel position [m, n] photographed at time t, and g [m0, n0, t0] represents the filter center pixel. τ represents a time constant and has a value of 1 or less. That is, the first weight w becomes smaller as the image is more distant from time t0. This assumes that the position of the object changes as the shooting time between images increases, but τ may be set to 1 if the shooting target object does not move. The filter range Ω can be defined as t = [t0-1, t0, t0 + 1] as an image at one time before and after, not only when only past images are used.

  Modification 1 of the present embodiment will be described. In the above embodiment, the first weight function and the filter range are expanded by taking the time axis as an example. However, in this modified example, the expansion is performed using other information. For example, when the image is three-dimensional volume data, the image can be expanded by adding another variable l in addition to the position information m and n.

  Modification 2 of the present embodiment will be described. In the second modification, other information that is not the pixel position or the image capturing time is used. For example, in the second modification, it is possible to expand a plurality of images (for example, stereo images) captured at the same time by a large number of cameras as targets. When the same object is photographed with a plurality of cameras that are photographed at different positions, the closer the cameras are, the less the change in appearance. Therefore, the camera position can be used instead of the time t.

  Next, an image processing apparatus 16 according to the fifth embodiment will be described with reference to FIG.

  In the first to fourth embodiments, the image processing device 16 is used to make it easy to visually recognize the linear pattern formed by the guide wire 22 in the target image G. Therefore, since the linear pattern by the guide wire 22 is a dark line pattern, the degree of the linear pattern increases as the line value α decreases. For example, the line value α is a negative value.

  However, when the linear pattern is a bright line, the degree of the linear pattern increases as the line value α increases. For example, the line value α is a positive value. Note that the “bright line” refers to a linear pattern that is brighter than other image areas and has continuity in a specific direction.

  Therefore, the case where the image processing apparatus 16 according to the present embodiment visually recognizes a bright line will be described. Similar to the first embodiment, the image processing apparatus 16 according to the present embodiment includes a linear pattern detection unit 100, a parameter determination unit 200, and a filter processing unit 300.

  Hereinafter, the linear pattern detection unit 100, the parameter determination unit 200, and the filter processing unit 300 will be described.

  First, the linear pattern detection unit 100 will be described.

  The linear pattern detection unit 10 obtains response values by applying the linear pattern detection filters H (φ) in various directions to the respective pixels [m, n] of the target image G, and then obtains the maximum values thereof. The line value α of each pixel [m, n] is assumed to be. In the present embodiment, if the line value α is a negative value, it is determined that the line pattern α is not a bright line pattern, and the value of the line value α is set to zero.

  Next, the parameter determination unit 200 will be described with reference to FIG.

  FIG. 9A shows a size of 9 × 9 pixels in which a linear shadow (pixel value 20) having a width of 1 pixel is placed on a flat portion of the pixel value 20 and Gaussian noise (standard deviation 1.0) is given to the entire image. This is a target image G. The pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G is 20.35.

  FIG. 9B is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G of FIG. 9A based on the equation (2), and the first parameter σ = 1. In this case, the second parameter ζ = 1. In FIG. 9B, since σ = 1 and ζ = 1, the peripheral pixel [m having a pixel value one higher than the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G , N] has the largest first weight w, and thus the pixel value f [m0, n0] of the output image F after the filtering process is higher than when ζ = 0. That is, the pixel value f [m0, n0] of the filter center pixel [m0, n0] of the output image F is 21.07. And the contrast with a flat part becomes large and visibility improves.

  FIG. 9C is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G in FIG. 9A based on the equation (2), and the first parameter σ = 1. In this case, the second parameter ζ = 0. Since σ = 1 and ζ = 0, the smaller the difference between the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G and the pixel value g [m, n] of the peripheral pixels is, the smaller the difference is. Since the weight w of 1 is increased, the first weight w applied to the filter center pixel [m0, n0] is larger than the peripheral pixels, and a value close to the pixel value g [m0, n0] of the target image G is Obtained as output after filtering. That is, the pixel value f [m0, n0] of the filter center pixel of the output image F after the filter processing is 20.45.

  FIG. 9D is a diagram showing a three-dimensional distribution of the first weight w obtained from the target image G of FIG. 9A based on the equation (2), and the first parameter σ = 3. In this case, the second parameter ζ = 0. Since σ = 3 and ζ = 0, the difference between the pixel value g [m0, n0] of the filter center pixel [m0, n0] of the target image G and the pixel value g [m, n] of the peripheral pixel [m, n] Since the change in the first weight w due to becomes dull, the weight of the pixel on the linear pattern becomes gentle and the smoothing effect increases, and the pixel value f [m0, n0] of the output image F is linear. It approaches the true value (= 20) of the pixel on the shadow. That is, the pixel value f [m0, n0] of the filter center pixel of the output image F after the filter processing is 20.10.

  Therefore, even in the present embodiment, the first parameter σ defines the level of smoothing, and for pixels that are not on the linear pattern, the noise reduction effect is enhanced by increasing σ. be able to. That is, σ is reduced as the line value α increases in the positive direction.

  The second parameter ζ can shift the pixel value after the weighted average from the original pixel value. As shown in FIG. 9B, in the pixel on the linear pattern, the contrast can be enhanced by giving ζ in a higher direction than the peripheral pixel by comparing the pixel value on the linear pattern with the peripheral pixel value. . That is, ζ is increased as the line value α increases.

  The operation of the filter processing unit 300 is the same as that of the first embodiment.

  Next, an image processing apparatus 16 according to the sixth embodiment will be described.

  In the image processing device 16 of each of the embodiments described above, the improvement in the visibility of the guide wire 22 in the X-ray image G has been described as an example. However, the image processing device 16 can also be applied to an image taken with a general camera.

  For example, when it is desired to detect cracks (dark lines) in an image obtained by photographing a wall surface of a building, the cracks have a linear structure locally, so that the image processing device 16 described so far distinguishes cracks from others. This makes it easy to create an image with high visibility.

  In addition, in order to detect defects in a semiconductor inspection apparatus, it has been described so far to make it easy to find defects (dark lines or bright lines) having a local linear pattern such as scratches in an inspection image. Visibility can be improved by the method.

Example of change

  The image processing apparatus 16 in the first to fourth embodiments is provided in the X-ray diagnostic apparatus 1, but is not limited thereto, and can be used alone.

  Although one embodiment of the present invention has been described above, this embodiment is presented as an example and is not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 16 ... Image processing apparatus, 100 ... Linear pattern detection part, 200 ... Parameter determination part, 300 ... Filter processing part

Claims (10)

For each pixel in the target image, a line value indicating the likelihood of a linear pattern of dark lines or bright lines having continuity in a specific direction is obtained, and a negative value among the line values according to the brightness of the detection target Or a linear pattern detection unit that sets one of positive values to zero, and
(1) Regarding an image processing filter configured with a first weight function in which the first weight decreases as the difference between the pixel value of the filter center pixel existing in the filter range and the pixel value of the surrounding pixels increases (2) The first parameter and the second parameter in the first weight function are determined from the line values. (3) The first parameter defines the spread of the first weight function. The larger the first parameter, the wider the spread. (4) The second parameter defines the center position of the spread of the first weight function. The larger the second parameter, the larger the first parameter. The center position is shifted from the filter center pixel. (5) When the positive line value is zero, the larger the line value is in the negative direction, or the negative line value is zero. (6) For each pixel of the target image, the first parameter and the second parameter are set such that as the line value increases in the positive direction, the first parameter decreases and the second parameter increases. A parameter determination unit for determining parameters;
An image processing filter composed of the first weight function including the first parameter and the second parameter determined for each pixel of the target image is applied to each pixel of the target image, and an output image A filter processing unit for calculating a pixel value of each pixel;
An image processing apparatus comprising: The linear pattern detection unit is
Detecting the traveling direction of the linear pattern of each pixel of the target image,
If the traveling direction of the pixel and the traveling direction of the peripheral pixels around the pixel match or are within a specific angle, the line value is corrected so that the degree of the linear pattern is increased. ,
The image processing apparatus according to claim 1. The parameter determination unit
The second weight of the surrounding pixels in the traveling direction is high, and the second weight of the surrounding pixels in the direction orthogonal to the traveling direction has a second weight function that decreases as the distance from the filter center pixel increases. And
The image processing filter is configured by a product of the first weight function and the second weight function;
The image processing apparatus according to claim 2. The parameter determination unit determines a parameter for determining a spread of the second weighting function based on the line value;
The image processing apparatus according to claim 3. An edge pattern detection unit for detecting an edge pattern in the target image;
The parameter determination unit corrects the line value of the pixel in which the edge pattern is detected so that the degree of the linear pattern is reduced.
The image processing apparatus according to claim 1. An isolated point pattern detecting unit for detecting an isolated point pattern in the target image;
The parameter determination unit corrects the line value of the pixel in which the isolated point pattern is detected so that the degree of the linear pattern is reduced.
The image processing apparatus according to claim 1. The target image is a time-series moving image;
The filter range of the image processing filter used in the filter processing unit is expanded in the time direction,
The image processing apparatus according to claim 1. The target image is an X-ray image taken by an X-ray diagnostic apparatus.
The image processing apparatus according to claim 1. Means for acquiring an X-ray image using an X-ray detector;
The image processing apparatus according to claim 8, wherein the X-ray image is input;
Image display means for displaying the output image from the image processing device;
An X-ray diagnostic apparatus comprising: For each pixel in the target image, a line value indicating the likelihood of a linear pattern of dark lines or bright lines having continuity in a specific direction is obtained, and a negative value among the line values according to the brightness of the detection target Or one of the positive values is zero,
(1) Regarding an image processing filter configured with a first weight function in which the first weight decreases as the difference between the pixel value of the filter center pixel existing in the filter range and the pixel value of the surrounding pixels increases (2) The first parameter and the second parameter in the first weight function are determined from the line values, and (3) the first parameter defines the spread of the first weight function. And the larger the first parameter, the wider the spread. (4) The second parameter determines the center position of the spread of the first weight function, and the second parameter is larger. The center position deviates from the filter center pixel. (5) When the positive line value is zero, the larger the line value is in the negative direction, or the negative line value is zero. (6) For each pixel of the target image, the first parameter and the first parameter are set such that as the line value increases in the positive direction, the first parameter decreases and the second parameter increases. Determining the second parameter;
An image processing filter composed of the first weight function including the first parameter and the second parameter determined for each pixel of the target image is applied to each pixel of the target image, and an output image Calculate the pixel value of each pixel,
An operation method for an image processing apparatus .