JP6121013B2 - Image processing apparatus and image processing method - Google Patents

Description

  The present invention relates to an image processing apparatus and an image processing method for emphasizing or suppressing a predetermined frequency component in an image signal.

Conventionally, a number of image processing techniques for improving the diagnostic performance of radiographic images by performing frequency enhancement have been proposed. The frequency enhancement or suppression method is performed by creating a plurality of band limited images representing frequency components in a limited frequency band based on the original image and enhancing or suppressing each band limited image. As a method of creating a plurality of band limited images, there are a method using Laplacian pyramid decomposition, a method using wavelet transform, a method using an unsharp mask, and the like. For example, when an unsharp mask is used and the blurred image is _SusLv , the band limited image H _Lv is as shown in (Expression 1).

Here, the blurred image S _us0 when Lv = 0 represents the original image S _org . Lv is an index of the band limited image. By creating blurred images with different frequency response characteristics, various band limited images can be obtained. When the lowest frequency image of the band limited images is L, the relationship with the original image can be expressed by (Equation 2).

  That is, it is reconstructed into the original image by adding the decomposed band-limited images. For the low frequency image L, a band limited image other than the low frequency image L is a high frequency image. By adjusting the gain with respect to the high-frequency image by the gain coefficient α as in (Equation 3), it is possible to create an image with various frequency enhancements or suppressions.

  The gain coefficient α is set for each frequency band. The frequency band component can be emphasized by increasing the gain coefficient α. By reducing the gain coefficient α, the frequency band component can be suppressed. However, in the enhancement process using the constant α, all components are subjected to the same enhancement process. That is, there is a problem that not only the edge component to be emphasized is emphasized, but also the noise component is equally enhanced, and a desired enhancement effect cannot be obtained. In response to this problem, a technique has been proposed in which only an edge component is detected from a high-frequency image and enhanced to obtain an edge enhancement effect (see, for example, Patent Documents 1 and 2). In particular, Patent Document 2 proposes a method of performing an enhancement process by decomposing into a plurality of band-limited high-frequency images, and performing an enhancement process by detecting an edge for each frequency band.

JP-A-9-248291 JP 2005-296331 A

  The user can adjust the gain coefficient α for each frequency band and create an image that has been subjected to various frequency processes. In the radiographic image, there are regions where noise is dominant and regions where this is not the case for each frequency band depending on the characteristics of the detector and the imaging region. Therefore, the adjustment of the gain coefficient α requires fine adjustment for each imaging region and for each frequency band. A desired frequency response characteristic can be obtained by adjusting the gain coefficient α. On the other hand, the adjustment work of the gain coefficient α is a very difficult work for a user with little experience, and a time-consuming work for a skilled user. By applying the gain coefficient set in this way only to the edge region and performing processing, an image in which only the edge is emphasized can be created.

  However, in the techniques as described in Patent Documents 1 and 2, the gain coefficient is adjusted to an optimal balance, but the result of edge detection fails, and as a result, noise is emphasized as well as the edge. Had occurred. This is because the edge detection method is common to all frequency bands, so that edge detection becomes difficult in the frequency band where noise is dominant, and noise is detected as an edge.

  An object of the present invention is to provide an image processing apparatus and an image processing method capable of enhancing edges without enhancing noise components even in a frequency band in which noise is dominant.

An image processing apparatus according to the present invention includes a frequency component generation unit that generates an image of a plurality of frequency components from an original image, and a coefficient that acquires a gain coefficient for performing at least one gain correction of the images of the plurality of frequency components. an acquiring unit, and select the edge detection method based on the above gain coefficient, a detecting means for detecting at least one of edge information of the image of the plurality of frequency components, based on the edge information, adjusts the gain factor and adjustment means, based on at least one of the front Sulfur butterfly integer unit by the plurality of frequency components gain coefficients are adjusted image, characterized by comprising an image producing means for producing an image.

  According to the present invention, even in a frequency band in which noise is dominant, an edge can be emphasized without enhancing a noise component, and an enhancement effect intended by a user can be obtained.

It is a figure which shows the structural example of the image processing apparatus in embodiment of this invention. It is a figure which shows the computer function which can implement | achieve the image processing apparatus in this embodiment. It is a flowchart which shows the flow of the image processing apparatus in this embodiment. It is a figure which shows the example of the gain coefficient correction value calculation method in this embodiment. It is a figure which shows the example of the edge detection by a moment operator. It is a figure for demonstrating a reference distance.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a configuration example of an image processing apparatus according to an embodiment of the present invention, and illustrates a configuration of an X-ray image processing apparatus as an example. As shown in FIG. 1, the X-ray image processing apparatus includes a frequency component creation unit 101, a gain coefficient acquisition unit 102, a gain coefficient correction value calculation unit 103, a gain adjustment unit 104, and a processed image creation unit 105.

  The frequency component creating unit 101 receives an image obtained by performing predetermined preprocessing on an X-ray image acquired by an X-ray sensor, and based on the image, a plurality of high frequency components representing frequency components in a limited frequency band. Create and output an image and one low-frequency image. The gain coefficient acquisition unit 102 sets a frequency response characteristic related to frequency enhancement by a user and outputs a gain coefficient necessary for frequency enhancement processing. The gain coefficient correction value calculation means 103 receives the gain coefficient output from the gain coefficient acquisition means 102, calculates and outputs a gain coefficient correction value for correcting the gain coefficient. The calculation of the gain coefficient correction value is performed by a gain coefficient correction value calculation method corresponding to the gain coefficient.

  The gain adjustment unit 104 receives the high frequency image output from the frequency component generation unit 101, the gain coefficient output from the gain coefficient acquisition unit 102, and the gain coefficient correction value output from the gain coefficient correction value calculation unit 103. To do. The gain adjusting unit 104 adjusts the high frequency image with the gain coefficient corrected using the gain coefficient correction value, and outputs the adjustment result. The processed image creation means 105 receives the result of adjusting the high-frequency image by the gain adjustment means 104 and the low-frequency image output from the frequency component creation means 101 as input, reconstructs the image, and outputs the reconstruction result To do.

  FIG. 2 is a diagram illustrating an example in which the image processing apparatus illustrated in FIG. 1 is realized using a computer (PC). The control PC 201, the X-ray sensor 202, and the X-ray generator 210 are connected by an optical fiber 222. A signal line for connecting the control PC 201, the X-ray sensor 202, and the X-ray generator 210 may not be an optical fiber. In addition, the control PC 201, the X-ray sensor 202, and the X-ray generator 210 may be connected to be communicable via a network such as CAN (Controller Area Network) or Gigabit Ethernet (registered trademark). A display unit 209, a storage unit 211, a network interface unit 212, and the like are connected to the optical fiber 222.

  The control PC 201 includes a CPU (Central Processing Unit) 203, a RAM (Random Access Memory) 204, a ROM (Read Only Memory) 205, an input unit 206, a display unit 207, and a storage unit 208. The CPU 203, the RAM 204, the ROM 205, the input unit 206, the display unit 207, and the storage unit 208 are connected to be communicable via a bus 221, for example. A command is sent to the X-ray sensor 202 and the display unit 209 via the control PC 201. In the control PC 201, the processing content for each shooting mode is stored as a software module in the storage unit 208, and is read into the RAM 204 and executed by an instruction unit (not shown). The means 101 to 105 shown in FIG. 1 are stored in the storage unit 208 as software modules. Note that the present embodiment is not limited to this, and each of the units 101 to 105 shown in FIG. 1 may be mounted as, for example, a dedicated image processing board, or may be optimally mounted depending on the purpose. Good.

Hereinafter, details of the X-ray image processing apparatus will be described according to four embodiments.
(First embodiment)
A first embodiment will be described.
Image processing in the first embodiment will be described with reference to the configuration diagram shown in FIG. 1 and the flowchart shown in FIG. First, an X-ray image is acquired by an X-ray sensor (step 301). Next, preprocessing is performed on the acquired X-ray image (step 302). The preprocessing includes, for example, processing for correcting sensor characteristics such as offset correction, log conversion, gain correction, and defect correction, and grid stripe suppression processing for suppressing grid moire. Further, if necessary, processing for improving the signal-to-noise ratio (S / N) such as processing for reducing random noise may be performed as preprocessing.

  Next, the frequency component creation unit 101 creates a plurality of high frequency images and one low frequency image from the preprocessed original image (step 303). As a method of creating the band limited image, a method using Laplacian pyramid decomposition, a method using wavelet transform, or the like is used. Although high-frequency images that can be effectively obtained by downsampling are limited, in the present embodiment, at least one or more high-frequency images are required.

  Next, the gain coefficient acquisition unit 102 sets a gain coefficient (step 304). Here, the gain coefficient indicates the value of α shown in (Equation 3), and is designated by the user for each frequency band. The user may directly specify the value of the gain coefficient α. Alternatively, a frequency response characteristic creation tool is prepared as another means, and the characteristic created without considering the gain coefficient α is changed to the gain coefficient α. It may be specified by the automatic conversion method.

  Next, the gain coefficient correction value calculation means 103 selects a gain coefficient correction value calculation method using the gain coefficient α value designated by the gain coefficient acquisition means 102 (step 305). To correct the gain coefficient is to weaken the gain coefficient of pixels other than the edge in order to selectively perform enhancement processing only on the edge portion of the high-frequency image. Therefore, the gain coefficient correction value calculation method is a method of detecting an edge component of a high-frequency image and outputting a value for correcting the gain coefficient based on the detection result. In the present embodiment, a plurality of edge detection methods are prepared and selected for each frequency band based on the gain coefficient.

  FIG. 4A shows an example of a table for selecting an edge detection method. As shown in FIG. 4A, a plurality of edge detection methods having different noise immunity are prepared. The edge detection method with higher noise resistance is selected as the gain coefficient is smaller, thereby preventing erroneous detection of noise as an edge. In the example shown in FIG. 4A, a detection method with high noise resistance such as a moment operator that uses an integrated value as a feature quantity for edge detection or a percentile filter that sets a threshold value for the absolute value of a signal component is used as a gain coefficient. Allocation is made in case of small.

Next, the gain coefficient correction value calculation means 103 outputs the result of calculating the gain coefficient correction value by the edge detection method selected in the gain coefficient correction value calculation method selection process (step 306). The gain coefficient correction value reflects the result of edge detection performed on the target high-frequency image, that is, edge information as the correction value. Edge detection process may be performed direct detection processing for high-frequency image, it performs processing on the original image S _org or the unsharp image S _us shown in (Equation 1), indirectly by utilizing the results May be.

  An example in which a moment operator is selected as an edge detection method will be described with reference to FIG. The method using a moment operator indirectly detects the edge of a high-frequency image using the original image as a processing target image. First, the gain coefficient correction value calculation unit 103 sets a local region centered on the processing target pixel, and calculates a zeroth-order moment and a first-order moment in the local region (steps 501 and 502). Each moment is obtained by (Equation 4).

In (Equation 4), m ₀₀ represents a zero-order moment, m ₁₀ represents a first moment in the x direction, m ₀₁ represents a first-order moment of the y-direction. Next, the gain coefficient correction value calculation means 103 calculates the center of gravity from the primary moment (step 503). Since the center of gravity is a position where the force is balanced, the coordinates (gx, gy) are obtained as in (Equation 5).

  Next, the gain coefficient correction value calculation means 103 calculates the distance between the center-of-gravity coordinates and the center coordinates of the local area using (Equation 6) (step 504).

If an edge is present in the local area, the center of gravity deviates from the center of the local area. Therefore, the value of Δ calculated by (Equation 6) increases. On the other hand, when there is no edge in the local area, the center of gravity is very close to the center of the local area, and the value of Δ calculated by (Equation 6) is small. Therefore, the value of Δ calculated by (Expression 6) is a feature amount for detecting an edge. The value of delta is calculated by (Equation 6), since hard to use as a value remains absolute value, in the present embodiment by normalizing at last reference distance delta _base, as the delta _norm (Equation 7) Calculate (step 505).

Note that the reference distance delta _base, is a value calculated by (Equation 6) If the edge of the predetermined step difference d is present in the local area as shown in Figure 6. Using the edge feature amount at the predetermined step d as a reference value means that the predetermined step d is an edge to be detected. If the value of Δ in each local area calculated by (Equation 6) is larger than this edge feature amount, it is recognized that an edge exists, and if it is smaller than this edge feature amount, it is considered that the probability that an edge exists is small. Accordingly, Δ _norm normalized by the reference distance Δ _base represents a maximum value as an edge detection feature amount when the value of Δ _norm is 1, indicating that an edge exists. On the other hand, the case where the value of Δ _norm is 0 represents the minimum value as the edge detection feature amount, and indicates that no edge exists. Value between 0 and 1 in the value of delta _norm represents the high probability that an edge closer to 1 is present, the probability that more edge exists closer to 0 low. In the present embodiment, the value of _Δnorm is set as a gain coefficient correction value.

  Next, the gain adjustment unit 104 uses the gain coefficient α designated by the gain coefficient acquisition unit 102 and the gain coefficient correction value C calculated by the gain coefficient correction value calculation unit 103 to obtain a high frequency image ( An adjustment is made based on equation (8) (step 307).

  The gain coefficient correction value C is a value close to 1 for a region that is highly likely to be an edge, and thus is adjusted to a value close to the gain coefficient value. On the other hand, the gain coefficient correction value C is a value close to 0 for a region that has a high possibility of not being an edge, so that the gain coefficient is adjusted to be small. The adjustment of the high-frequency image using the gain coefficient α and the gain coefficient correction value C by the gain adjusting unit 104 is performed for each pixel.

  Next, the processed image creation means 105 reconstructs an image from the high-frequency image H ′ whose gain has been adjusted. By replacing the high-frequency image H shown in (Expression 2) with the high-frequency image H ′ whose gain has been adjusted, a reconfigured image with gain adjusted is created (step 308).

  The image reconstructed by the processed image creation means 105 is subjected to post-processing such as geometric transformation, WW (window width) / WL (window level), and the image is output to a monitor or memory (steps 309 and 310).

  By repeating the processing from step 304 to step 307 described above for each frequency band, edge detection optimized for each frequency band is performed, and a gain correction coefficient value corresponding to the detection result is obtained. By using these, selective frequency emphasis processing that is not easily affected by noise is performed, and an image that can obtain an emphasis effect while the frequency response selected by the user is maintained.

(Second Embodiment)
Next, a second embodiment will be described.
In the first embodiment, a plurality of edge detection methods are shown as a method of calculating a plurality of gain coefficient correction values. In the second embodiment, a method of using a plurality of output results with the same detection method will be described. Image processing in the second embodiment will be described with reference to the configuration diagram shown in FIG. 1 and the flowchart shown in FIG. Since the processing from step 301 to step 304 is the same as that in the first embodiment, description thereof is omitted.

  In step 305, the gain coefficient correction value calculation unit 103 selects a method for calculating the gain coefficient correction value using the gain coefficient α value designated by the gain coefficient acquisition unit 102. The gain coefficient correction value calculation method is a method of detecting an edge component of a high-frequency image as described above and outputting a value for correcting the gain coefficient based on the detection result. In the present embodiment, a plurality of methods having different detection sensitivities for one edge detection method are prepared, and are selected for each frequency band based on a gain coefficient.

FIG. 4B shows an example of the relationship between different edge detection sensitivities and gain coefficients. The edge detection sensitivity indicates the ease of reaction with respect to the edge. If the edge detection sensitivity is high, the possibility of recognition as an edge increases. On the other hand, when the edge detection sensitivity is high, there is a high possibility that an object other than an edge such as noise is erroneously recognized as an edge. Therefore, as shown in FIG. 4B, the edge detection sensitivity is reduced when the gain coefficient is a small value. When the gain coefficient is set small, erroneous recognition is prevented because noise is a dominant frequency band. When the gain coefficient is a large value, processing is performed with a predetermined detection sensitivity _SD . Alternatively, more detailed edge detection may be performed by setting a value larger than the detection sensitivity _SD .

An example in which the edge detection sensitivity is changed when a moment operator is used as the edge detection method will be described. The edge feature amount when the moment operator is used is a value of Δ _{norm obtained} by normalizing the value of Δ shown in (Expression 7) shown in the first embodiment with the reference distance Δ _base .

The reference distance is a value calculated by (Equation 6) when the edge of the predetermined step d exists in the local region, as shown in the first embodiment. The predetermined step d is an edge that is a detection target, and the ratio of the step to the distance indicates the possibility of an edge. Therefore, when the predetermined level difference d is reduced, the target edge is set to be small, and the detection sensitivity is increased. The value of _Δnorm is also the same as Δ and increases as the value of the predetermined step d is decreased. That is, the value is highly likely to be an edge. It can be said that the value of the predetermined step d is a parameter that determines the detection sensitivity. As shown in FIG. 4B, in order to decrease the detection sensitivity as the gain coefficient decreases, the value can be adjusted by decreasing the value of the predetermined step d. Also in this embodiment, the value of _Δnorm is set as the gain coefficient correction value C (step 306).

  Next, the gain adjustment unit 104 uses the gain coefficient α designated by the gain coefficient acquisition unit 102 and the gain coefficient correction value C calculated by the gain coefficient correction value calculation unit 103 to obtain a high frequency image ( An adjustment is made based on equation (8) (step 307). Since the subsequent processing from step 308 to step 310 is the same processing as in the first embodiment, the description thereof will be omitted.

  According to the second embodiment, it is possible to perform frequency emphasis processing for emphasizing an edge without emphasizing a noise component, and it is possible to obtain more emphasis effect while maintaining the frequency response selected by the user.

(Third embodiment)
Next, a third embodiment will be described.
In the third embodiment, a second method of using a plurality of output results by the same detection method as a method of calculating a plurality of gain coefficient correction values will be described. Image processing in the third embodiment will be described with reference to the configuration diagram shown in FIG. 1 and the flowchart shown in FIG. Since the processing from step 301 to step 304 is the same as that in the first embodiment, description thereof is omitted.

  In step 305, the gain coefficient correction value calculation unit 103 selects a method for calculating the gain coefficient correction value using the gain coefficient α value designated by the gain coefficient acquisition unit 102. To correct the gain coefficient is to weaken the gain coefficient of pixels other than the edge in order to selectively perform enhancement processing only on the edge portion of the high-frequency image. In addition, in order to selectively suppress only the noise component of the high-frequency image, the gain coefficient of pixels other than noise is also weakened. Therefore, the gain coefficient correction value calculation method is a method of detecting an edge component or a noise component of a high-frequency image and outputting a value for correcting the gain coefficient based on the detection result. In the present embodiment, different output value correction coefficients are prepared for one detection method, and are selected for each frequency band based on the gain coefficient.

An example in which the output value correction coefficient is changed when a moment operator is used as the detection method will be described. The edge feature amount when the moment operator is used is a value of Δ _{norm obtained} by normalizing the value of Δ shown in (Expression 7) shown in the first embodiment with the reference distance Δ _base . The closer the value of _Δnorm is to 1, the higher the possibility of being an edge, and the closer the value of _Δnorm is to 0, the more likely it is not an edge. That is, the possibility of noise increases. Thus, when the detection correction coefficient is K, the gain correction coefficient value C is as shown in (Equation 9).

Here, the detection correction coefficient K indicates a step function as shown in FIG. When the gain coefficient value α is 1 or more, K = 1, and the value of _Δnorm is used as it is as the gain correction coefficient value C. On the other hand, when the gain coefficient value α is less than 1, K = 0, and the value of (1−Δ _norm ) is used as the gain correction coefficient value C. When the gain coefficient value α is less than 1, the component of the high frequency image is suppressed, and the noise is suppressed. In this case, if the gain coefficient of a pixel detected as an edge is held in the same manner as when the gain coefficient value α is 1 or more, the edge is suppressed and the entire image is blurred. Therefore, in that case, it is necessary to increase the gain coefficient of noise, not the edge gain coefficient.

  Therefore, in order to increase the noise detection sensitivity over the edge detection sensitivity, the result detected as the edge is inverted and used so that the noise gain coefficient is increased. When the gain coefficient α is less than 1, it operates as a noise detection method. In the present embodiment, since the normalized value is used as the feature amount, the value that is the first edge information output as the edge detection method and the value that is the second edge information output as the noise detection method Is a one's complement relationship. In the example shown in FIG. 4C, the gain function α is a step function with a threshold value of 1. However, the step function is not limited to a step function, and may be a function that switches continuously (step 306).

  Next, the gain adjustment unit 104 uses the gain coefficient α designated by the gain coefficient acquisition unit 102 and the gain coefficient correction value C calculated by the gain coefficient correction value calculation unit 103 to obtain a high frequency image ( An adjustment is made based on equation (8) (step 307). Since the subsequent processing from step 308 to step 310 is the same processing as in the first embodiment, the description thereof will be omitted.

  According to the third embodiment, it is possible to perform frequency emphasis processing for emphasizing an edge without emphasizing a noise component, and a more emphasis effect can be obtained while the frequency response selected by the user is maintained.

(Fourth embodiment)
Next, a fourth embodiment will be described.
In the fourth embodiment, a third method of using a plurality of output results by the same detection method as a method of calculating a plurality of gain coefficient correction values will be described. Image processing according to the fourth embodiment will be described with reference to the configuration diagram shown in FIG. 1 and the flowchart shown in FIG. Since the processing from step 301 to step 304 is the same as that in the first embodiment, description thereof is omitted.

  In step 305, the gain coefficient correction value calculation unit 103 selects a method for calculating the gain coefficient correction value using the gain coefficient α value designated by the gain coefficient acquisition unit 102. To correct the gain coefficient is to weaken the gain coefficient of pixels other than noise in order to selectively perform suppression processing only on the noise portion of the high-frequency image. Therefore, the gain coefficient correction value calculation method is a method of detecting a noise component of a high-frequency image and outputting a value for correcting the gain coefficient based on the detection result, that is, noise information. In the present embodiment, a plurality of methods having different detection sensitivities for one noise detection method are prepared, and are selected for each frequency band based on a gain coefficient.

FIG. 4D shows an example of the relationship between different noise detection sensitivities and gain coefficients. An example in which the detection sensitivity is changed in the case where the method of calculating the variance value is used as the noise detection method will be described. With respect to the noise feature amount when the method for calculating the variance value is used, a local region is set on the image, and is used as the variance value in the region. Then, the amount of noise is theoretically calculated from the average value in the region as follows. Random noise on the X-ray image is represented by a pixel value fluctuation amount in a specified region when irradiation is performed with an X-ray dose having a constant intensity “X”. Then, random quantum noise that fluctuates the pixel value with the standard deviation σ _q (X) caused by “X” that is the X-ray dose, and electrical that fluctuates the pixel value with the standard deviation σ _s received from the sensor and the peripheral electric circuit. Random system noise, which is noise, is classified into two types. Since these two types of random noise are added as random noise on the X-ray image, the theoretical noise amount σ (X) can be expressed by (Expression 10) and (Expression 11). Here, the variable X is an average value of the local region.

Here, the standard deviation σ _q (X) changes according to (Equation 10) depending on the average value X, but the standard deviation σ _s is an electrical thermal noise and is constant regardless of the X-ray intensity. Value. In (Expression 11), K _q is a conversion coefficient when calculating the noise amount from the X-ray intensity. As shown in (Equation 12), a pixel whose variance value σ _img in the local region is less than or equal to the theoretical noise amount σ (X) is detected as noise, and a gain coefficient is used to perform noise suppression processing. The correction value C is 1. On the other hand, a pixel whose variance value σ _img in the local region is equal to or greater than the theoretical noise amount σ (X) is detected as an edge, and the gain coefficient correction value C is set to 0 in order to prevent noise suppression processing. . When making this comparison, the detection sensitivity S _D is multiplied by the theoretical noise amount.

The detection sensitivity S _D is a function of the gain coefficient α and has a relationship as shown in FIG. When the gain coefficient α is a small value, the detection sensitivity is increased. When the gain coefficient α is set to a small value, noise is a dominant frequency band, and thus leakage of noise extraction is prevented. When the gain coefficient α is a large value, processing is performed with a predetermined detection sensitivity _SD . Alternatively, a value smaller than the detection sensitivity _SD may be used to prevent excessive noise extraction (step 306).

  Next, the gain adjustment unit 104 uses the gain coefficient α designated by the gain coefficient acquisition unit 102 and the gain coefficient correction value C calculated by the gain coefficient correction value calculation unit 103 to obtain a high frequency image ( An adjustment is made based on equation (8) (step 307). Since the subsequent processing from step 308 to step 310 is the same processing as in the first embodiment, the description thereof will be omitted.

  According to the fourth embodiment, it is possible to perform frequency emphasis processing for emphasizing an edge without emphasizing a noise component, and a more emphasis effect can be obtained while the frequency response selected by the user is maintained.

(Other embodiments of the present invention)
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

  The above-described embodiments are merely examples of implementation in carrying out the present invention, and the technical scope of the present invention should not be construed as being limited thereto. That is, the present invention can be implemented in various forms without departing from the technical idea or the main features thereof.

101 frequency component creation means, 102 gain coefficient acquisition means, 103 gain coefficient correction value calculation means, 104 gain adjustment means, 105 processed image creation means

Claims (11)

Frequency component generation means for generating an image of a plurality of frequency components from the original image;
Coefficient acquisition means for acquiring a gain coefficient for performing at least one gain correction of the images of the plurality of frequency components;
Detecting means for selecting an edge detection method based on the gain coefficient and detecting at least one edge information of the images of the plurality of frequency components;
Adjusting means for adjusting the gain coefficient based on the edge information;
An image processing apparatus comprising: an image generation unit configured to generate an image based on at least one of the plurality of frequency component images whose gain coefficients have been adjusted by the adjustment unit.   The image processing apparatus according to claim 1, wherein the adjustment unit corrects the gain coefficient based on the edge information and adjusts the gain coefficient.   The image processing apparatus according to claim 1, wherein the detection unit detects at least one edge information of the images of the plurality of frequency components with edge detection sensitivity based on the gain coefficient.   When the gain coefficient value is less than 1, the detection means detects edge information whose edge detection sensitivity is lower than noise detection sensitivity from at least one of the plurality of frequency component images. The image processing apparatus according to claim 1.   3. The image according to claim 2, wherein the adjustment means has a one's complement relationship between the gain coefficient corrected by the first edge information and the gain coefficient corrected by the second edge information. Processing equipment.   The image processing apparatus according to claim 1, wherein the detection unit detects at least one edge information of the images of the plurality of frequency components with a reference value of an edge feature amount based on the gain coefficient.   The image processing apparatus according to claim 1, wherein the detection unit selects an edge detection method based on the gain coefficient for each of the plurality of frequency component images from among the plurality of edge detection methods. . The image processing apparatus according to claim 7, wherein the plurality of edge detection methods include a plurality of edge detection methods having different noise tolerances. The image processing apparatus according to claim 8 , wherein the detection unit selects an edge detection method having higher noise resistance as the gain coefficient value is smaller. A frequency component generation step of generating an image of a plurality of frequency components from the original image;
A coefficient acquisition step of acquiring a gain coefficient for performing at least one gain correction of the image of the frequency component;
A detection step of selecting an edge detection method based on the gain coefficient and detecting at least one edge information of the images of the plurality of frequency components;
An adjustment step of adjusting the gain coefficient based on the edge information;
And an image generation step of generating an image based on at least one of the images of the plurality of frequency components, the gain coefficient of which has been adjusted in the adjustment step. A frequency component generation step for generating an image of a plurality of frequency components from the original image;
A coefficient acquisition step of acquiring a gain coefficient for performing at least one gain correction of the images of the plurality of frequency components;
A detection step of selecting an edge detection method based on the gain coefficient and detecting at least one edge information of the images of the plurality of frequency components;
An adjustment step of adjusting the gain coefficient based on the edge information;
A computer program for causing a computer to execute an image generation step of generating an image based on at least one of the images of the plurality of frequency components in which the gain coefficient is adjusted in the adjustment step.

Images