Noise reduction processing method and noise reduction processing device

Abstract

A noise reduction processing method and a noise reduction processing device are provided. The noise reduction processing method comprises the following steps: applying a first filter having a first filter window to an object signal to obtain a low-frequency signal, applying a second filter having a second filter window independent of the first filter window to a high-frequency signal obtained based on the object signal and the low-frequency signal to estimate a noise intensity signal; and obtaining a low-noise signal.

Description

Technical Field

[0001] The present invention relates to a noise reduction processing method and a noise reduction processing device, and in particular to a noise reduction processing method and a noise reduction processing device for reducing noise while preserving high-frequency components of a target signal. Background Art

[0002] Conventionally, there is a known noise reduction processing method for reducing noise while preserving the high frequency components of the target signal. Such a noise reduction processing method is disclosed in, for example, Dutta, Joyita, Richard M. Leahy, and Quanzheng Li., "Non-local means denoising of dynamic PET images.", PLOS ONE, 8.12 (2013), e81390. (hereinafter referred to as "non-patent document 1").

[0003] In the above-mentioned non-patent document 1, a noise reduction process based on edge-preserving smoothing filtering is disclosed, which suppresses the decrease in noise reduction effect caused by the position dependence of the noise contained in the image. Specifically, in the above-mentioned non-patent document 1, a technology for performing non-local means filtering processing is disclosed, and the non-local means filtering processing is the following edge-preserving smoothing filtering: for a specified pixel, a template matching is performed between a pixel in a support window (Support Window) set for the neighboring area of ​​the specified pixel and a pixel in a template window (Template Window), thereby obtaining the similarity between the pixel in the support window and the pixel in the template window, and performing weighted filtering processing corresponding to the similarity. In addition, in the structure disclosed in the above-mentioned non-patent document 1, the noise contained in the image is estimated, and the filter coefficient used in the filtering process is obtained. In addition, in the above-mentioned non-patent document 1, the following noise reduction processing is disclosed: the standard deviation in the neighboring area is calculated for each pixel, and it is set as the estimated value of the noise of each pixel and used in the calculation of the filter coefficient, thereby suppressing the decrease in noise reduction effect based on the position dependence of the noise.

[0004] The inventors of the present application have conducted intensive studies and have found that, when the standard deviation in the neighborhood is calculated for each pixel to estimate the noise of each pixel as disclosed in the above-mentioned Non-Patent Document 1, the target high-frequency component is lost. Summary of the invention

[0005] The present invention has been made to solve the above-mentioned problems, and one object of the present invention is to provide a noise reduction processing method capable of reducing noise while suppressing damage to target high-frequency components.

[0006] It has been found that by making the size of the filter window used when extracting the low-frequency component and the size of the filter window used when estimating the noise component from the high-frequency component mixed with the original signal and the noise component (differentiating the original signal from the extracted low-frequency component) independently different from each other, it is possible to suppress the damage of the high-frequency component that is the object of the filtering process. In addition, in order to improve the estimation accuracy of the noise, it is considered to set the size of the filter window used when estimating the noise component from the high-frequency component to be large.

[0007] The noise reduction processing method in the first aspect of the present invention is a noise reduction processing method for an object signal, and the noise reduction processing method includes the following steps: estimating a noise intensity signal based on an object signal, the noise intensity signal representing the intensity distribution of the noise of the object signal; calculating a similarity index between a prescribed signal obtained at a first signal point and a signal obtained at a second signal point adjacent to the first signal point based on the object signal and the noise intensity signal; calculating a filter coefficient of a noise reduction filter based on the similarity index; and applying a noise reduction filter to the object signal based on the filter coefficient to obtain a low-noise signal, and in the step of estimating the noise intensity signal, performing the following processing: applying a first filter having a first filter window to the object signal to obtain a low-frequency signal; obtaining a high-frequency signal based on the object signal and the low-frequency signal; and applying a second filter having a second filter window independent of the first filter window to the high-frequency signal to estimate the noise intensity signal. In addition, the second filter window independent of the first filter window means that the size of the second filter window can be set independently of the size of the first filter window.

[0008] The noise reduction processing device in the second aspect of the present invention is a noise reduction processing device for an object signal, and the noise reduction processing device comprises: an object signal acquisition unit, which acquires an object signal; and a noise reduction unit, which reduces the noise of the object signal, wherein the noise reduction unit performs the following processing: applying a first filter having a first filter window to the object signal to acquire a low-frequency signal, acquiring a high-frequency signal based on the object signal and the low-frequency signal, and applying a second filter having a second filter window independent of the first filter window to the high-frequency signal to estimate a noise intensity signal; calculating a similarity index between a specified signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point based on the object signal and the noise intensity signal; calculating a filter coefficient of the noise reduction filter based on the similarity index; and applying the noise reduction filter to the object signal based on the filter coefficient to acquire a low-noise signal.

[0009] In addition, the noise reduction processing method in the third aspect of the present invention is a noise reduction processing method for an object signal, and the noise reduction processing method includes the following steps: based on the object signal, estimating a noise intensity signal representing the intensity distribution of noise of the object signal; based on the object signal and the noise intensity signal, calculating a similarity index between a specified signal obtained at a first signal point and a signal obtained at a second signal point adjacent to the first signal point; based on the similarity index, calculating a filter coefficient of a noise reduction filter; based on the filter coefficient, applying a noise reduction filter to the object signal to obtain a low-noise signal; and synthesizing the object signal and the low-noise signal to obtain a low-noise synthesized signal to adjust the noise reduction effect.

[0010] The noise reduction processing device in the fourth aspect of the present invention is a noise reduction processing device for an object signal, and the noise reduction processing device comprises: an object signal acquisition unit, which acquires an object signal; and a noise reduction unit, which reduces the noise of the object signal, wherein the noise reduction unit is configured to perform the following processing: based on the object signal, estimating a noise intensity signal representing the intensity distribution of the noise of the object signal; based on the object signal and the noise intensity signal, calculating a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; based on the similarity index, calculating a filter coefficient of a noise reduction filter; based on the filter coefficient, applying a noise reduction filter to the object signal to acquire a low-noise signal; and synthesizing the object signal and the low-noise signal to acquire a low-noise synthesized signal to adjust the noise reduction effect.

[0011] In the noise reduction processing method in the first aspect of the present invention and the noise reduction processing device in the second aspect, a second filter having a second filter window is applied to a high-frequency signal to estimate a noise intensity signal, and the second filter window is independent of the first filter window of the first filter when acquiring a low-frequency signal. As a result, the size of the first filter window and the size of the second filter window can be set independently of each other, so that the damage to the high-frequency component as the object can be suppressed. In addition, a second filter having a second filter window is applied to the high-frequency signal to estimate the noise intensity signal, and the second filter window is independent of the first filter window when acquiring the low-frequency signal. Therefore, the size of the first filter window and the size of the second filter window can be set according to the object signal to estimate the noise intensity signal. Therefore, by changing the size of the second filter window according to the object signal, the statistical variation of the noise level in the estimated noise intensity signal can be suppressed, so that the estimation accuracy of the noise intensity signal can be improved. Therefore, the accuracy of the similarity index acquired based on the noise intensity signal can be improved. In addition, since the accuracy of the similarity index can be improved, the filter coefficient can be calculated with high accuracy. As a result, noise can be reduced while suppressing the damage to the high-frequency component as the object.

[0012] In addition, in the noise reduction processing method in the third aspect of the present invention and the noise reduction processing device in the fourth aspect, the object signal and the low noise signal are synthesized to obtain a low noise synthesis signal to adjust the noise reduction effect. Thus, by synthesizing the object signal and the low noise signal, the object signal can be used to suppress the damage of the high frequency component of the object, and the low noise signal can be used to reduce the noise. As a result, a noise reduction processing method and a noise reduction processing device that can reduce noise while suppressing the damage of the high frequency component of the object can be provided. In addition, in the noise reduction processing method in the third aspect of the present invention and the noise reduction processing device in the fourth aspect, the object signal and the low noise signal are synthesized to obtain a low noise synthesis signal to adjust the noise reduction effect. Thus, by adjusting the ratio of synthesizing the object signal and the low noise signal, the noise reduction effect can be easily adjusted. As a result, compared with a structure in which a series of processes are performed again according to the object signal to obtain the low noise signal after changing the parameters, the time required for the adjustment process of the noise reduction effect can be shortened. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a block diagram showing the structure of a noise reduction processing device according to one embodiment.

[0014] Figure 2 This is a table for explaining examples of target signals.

[0015] Figure 3 This is a table for explaining examples of auxiliary signals.

[0016] Figure 4 This is a block diagram for explaining a configuration of a noise reduction processing device according to an embodiment for reducing noise of a target signal.

[0017] Figure 5 This is a block diagram for explaining a configuration in which a noise intensity signal estimation unit estimates a noise intensity signal according to one embodiment.

[0018] Figure 6 are schematic diagrams and graphs used to illustrate the model of noise distribution.

[0019] Figure 7 This is a block diagram for explaining a structure in which a synthesis processing unit according to one embodiment acquires a low-noise synthesis signal.

[0020] Figure 8 This is a flowchart for explaining a noise reduction process performed by a noise reduction process method according to one embodiment.

[0021] Fig. 9 1 is an image example for explaining an original image used in one embodiment.

[0022] Fig.10 This is an image example for explaining an image obtained by adding noise to an original image in one embodiment.

[0023] Fig.11 1 is an image example for explaining a low-noise image according to the first comparative example.

[0024] Fig.12 This is an image example for explaining an edge comparison image which is an image obtained by comparing the edge of the low-noise image of the first comparative example with the edge of the original image.

[0025] Fig.13 1 is an image example for explaining a low-noise image according to the second comparative example.

[0026] Fig.14 This is an image example for explaining an edge comparison image which is an image obtained by comparing the edge of the low-noise image of the second comparative example with the edge of the original image.

[0027] Fig.15 This is an example of an image of a low-noise image for explaining the first embodiment.

[0028] Fig.16 This is an image example for explaining an edge comparison image which is an image obtained by comparing the edge of the low-noise image of the first embodiment with the edge of the original image.

[0029] Fig.17 FIG. 4 is an image example of a low-noise image for explaining the second embodiment.

[0030] Fig.18 This is an image example for explaining an edge comparison image which is an image obtained by comparing the edge of the low-noise image of the third embodiment with the edge of the original image.

[0031] Fig.19 This is a flowchart for explaining the noise reduction process performed by the noise reduction process method in the modification example. DETAILED DESCRIPTION

[0032] Hereinafter, embodiments for implementing the present invention will be described based on the drawings.

[0033] First, refer to Figure 1 to Figure 7 The configuration of a noise reduction processing apparatus 100 and a noise reduction processing method according to an embodiment will be described. The noise reduction processing apparatus 100 reduces noise of a target signal 30 .

[0034] like Figure 1As shown, the noise reduction processing apparatus 100 includes a target signal acquisition unit 1, a noise reduction unit 2, a storage unit 3, an input reception unit 4, and a display unit 5. The noise reduction processing apparatus 100 is implemented by a computer.

[0035] The object signal acquisition unit 1 acquires the object signal 30. In the present embodiment, the object signal acquisition unit 1 can acquire the object signal 30 acquired by various medical imaging modalities such as an X-ray imaging device, a CT (Computed Tomography) device, and a PET (Positron Emission Tomography) device. The object signal acquisition unit 1 is, for example, an input / output interface. In addition, the object signal 30 is an X-ray image, a CT image, a PET image, etc. The details of the object signal 30 are described later.

[0036] The noise reduction unit 2 is configured to reduce the noise of the target signal 30. The noise reduction unit 2 includes a processor or circuit such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit) and an FPGA (field-programmable gate array), and a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory).

[0037] In addition, the noise reduction unit 2 includes a noise intensity signal estimation unit 20, a similarity index calculation unit 21, a filter coefficient calculation unit 22, a filter processing unit 23, and a synthesis processing unit 24 as functional blocks. The functional units 20 to 24 are configured as functional blocks in software that are implemented by executing various programs stored in the storage unit 3 by the noise reduction unit 2. The functional units 20 to 24 may also be provided with dedicated processors (processing circuits) and configured by hardware independent of each other. The details of the functions of the functional units 20 to 24 will be described later.

[0038] The storage unit 3 stores various programs executed by the noise reduction unit 2. In addition, the storage unit 3 stores the object signal 30, the low-noise signal 34 described later, the auxiliary signal 35 described later, the low-noise composite signal 36 described later, the first filter 40, the second filter 41, the noise model 50 described later, the reference data set 51 described later, the first weight 60 described later, and the second weight 61 described later. The first filter 40 includes a smoothing filter. In addition, the second filter 41 includes a standard deviation filter. The storage unit 3 includes, for example, a non-volatile storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0039] The input reception unit 4 receives the operation input of the operator. The input reception unit 4 includes input devices such as a keyboard and a mouse.

[0040] The display unit 5 displays at least one of the low-noise signal 34 and the low-noise composite signal 36. The display unit 5 is a display device such as a liquid crystal monitor or an organic EL (Electro Luminescence) monitor.

[0041] (Object signal)

[0042] The object signal 30 (reference Figure 1 ) includes signals of one dimension or more. Figure 2 The table 52 shown represents an example of the object signal 30 and the maximum dimension of the object signal 30. As shown in the table 52, the object signal 30 includes an optical image (still image), an optical image (moving image), a chromatogram, an MS (Mass Spectrometry) spectrum, an MS image, a cell image, an X-ray image (still image), an X-ray image (moving image), an absorption CT image, an energy discrimination CT image, a phase CT image, a dark field CT image, an MR (Magnetic Resonance) image, a PET image, a dynamic parameter PET image, a SPECT (Single Photon Emission Computed Tomography) image, a dynamic parameter SPECT image, an absorption coefficient image, and an ultrasonic image. In this specification, the dimension of the object signal refers to the number of coordinate axes required to determine a specified signal element included in the object signal.

[0043] A cell image is a still image or a moving image of a cell used to observe the shape, size, internal structure, etc. of a cell. An energy-discriminating CT image is a CT image obtained by imaging using the energy information of X-rays. A phase CT image and a dark-field CT image are CT images obtained by imaging using the phase information of X-rays. A dynamic parameter PET image is a PET image obtained by imaging based on physiological parameters (such as the velocity constant of the compartment model) obtained from the time series data of radiation energy concentration (time radiation energy curve). A dynamic parameter SPECT image is a SPECT image obtained by imaging based on physiological parameters obtained from the time series data of radiation energy concentration. An absorption coefficient image is an image obtained by imaging based on the absorption coefficient and is used in PET and SPECT.

[0044] Among the object signals 30 shown in Table 52, one-dimensional signals are chromatograms and MS spectra, two-dimensional signals are optical images of still images, X-ray images of still images, and cell images, three-dimensional signals are optical images of moving images, MS graphs, X-ray images of moving images, cell images of moving images, dynamic parameter PET images, and dynamic parameter SPECT images, four-dimensional signals are absorption CT images, phase CT images, dark field CT images, PET images, SPECT images, absorption coefficient images, and ultrasonic images, and five-dimensional signals are energy discrimination CT images and MR images.

[0045] Optical images (still images), optical images (moving images), and cell images may contain noises such as high sensitivity noise due to increased sensitivity, block noise generated when compressing images, amplifier noise due to heat, and dark current noise due to dark current. In chromatograms, noises such as white noise and tailing may be contained. In MS spectra, noises such as white noise and Poisson noise may be contained. In MS graphs, noises such as white noise and Poisson noise may be contained. X-ray images (still images) and X-ray images (moving images) may contain noises such as quantum noise, noise caused by scattered X-rays, noise caused by defects in the detector, and dark current noise due to dark current. Absorption CT images, energy-discriminating CT images, phase CT images, and dark-field CT images may contain noises such as quantum noise, noise caused by scattered X-rays, noise caused by defects in the detector, and dark current noise due to dark current. MR images, PET images, dynamic parameter PET images, SPECT images, dynamic parameter SPECT images, and absorption coefficient images may contain noises such as quantum noise. Ultrasonic images may contain noise caused by noise generated from components such as a power supply and an amplifier.

[0046] When the target signal 30 contains the above-mentioned noise, the accuracy of analysis, analysis, diagnosis, etc. performed using the target signal 30 decreases. Figure 1 ) reduces the noise of the object signal 30 to suppress the decrease in accuracy of analysis, analysis, diagnosis, etc. performed using the object signal 30. In this embodiment, the structure of reducing the noise of the object signal 30 in an image signal of two dimensions or more is mainly described.

[0047] (Auxiliary signal)

[0048] Next, refer to Figure 3 The auxiliary signal 35 used when reducing the noise of the object signal 30 is described. The auxiliary signal 35 is acquired for the same object as the object from which the object signal 30 is acquired, using a device different from the device from which the object signal 30 is acquired, under acquisition conditions different from the conditions when the object signal 30 is acquired, or under processing conditions different from those of the object signal 30.

[0049] Figure 3 The table 53 shown is a table showing the object signal 30 and the corresponding auxiliary signal 35 .

[0050] When the object signal 30 is an optical image (still image) and an optical image (moving image), the auxiliary signal 35 includes an image obtained by photographing the same scene with or without a flash (lighting) and an image obtained by photographing the same scene with a changed ISO sensitivity.

[0051] In the case where the object signal 30 is a chromatogram, the auxiliary signal 35 includes a signal after baseline correction. In the case where the object signal 30 is an MS spectrum, the auxiliary signal 35 includes a signal after scale conversion. In the case where the object signal 30 is an MS map and a cell image, the auxiliary signal 35 includes a microscope image of the same sample, an X-ray image of the same sample, a CT image of the same sample, and an MR image of the same sample. In the case where the object signal 30 is a signal of an X-ray image (still image), the auxiliary signal 35 includes a DRR (Digital Reconstructed Radiograph: digital reconstruction image) image of the same subject. The DRR image is a two-dimensional image of a specified focal position reconstructed based on the CT image. In the case where the object signal 30 is a signal of an X-ray image (moving image), the auxiliary signal 35 includes a dynamic DRR image of the same subject. The dynamic DRR image is a dynamic DRR image composed of multiple DRR images. In the case where the object signal 30 is an absorption CT image, the auxiliary signal 35 includes an energy discrimination CT image of the same subject, a phase CT image of the same subject, and a dark field CT image of the same subject. When the object signal 30 is an energy-discriminating CT image, a phase CT image, and a dark-field CT image, the auxiliary signal 35 includes an absorption CT image of the same subject. When the object signal 30 is an MR image, the auxiliary signal 35 includes a CT image of the same subject, a PET image of the same subject, and a SPECT image of the same subject. When the object signal 30 is a PET image and a dynamic parameter PET image, the auxiliary signal 35 includes a CT image, an MR image, a SPECT image, and a tissue composition ratio image. The tissue composition ratio image is an image obtained by imaging based on the composition ratio of the substances contained in each pixel of the object. When the object signal 30 is a SPECT image, a dynamic parameter SPECT image, and an absorption coefficient image, the auxiliary signal 35 includes a CT image, an MR image, a PET image, and a tissue composition ratio image. When the object signal 30 is an ultrasonic image, the auxiliary signal 35 includes a CT image, an MR image, a PET image, and a SPECT image.

[0052] If it is a multimodality imaging device such as a PET-CT device, a PET-MR device, an MS imaging device capable of simultaneously acquiring an MS image and an optical image, or a CT device capable of simultaneously acquiring an absorption CT image, a phase CT image, and a dark field CT image, it is possible to simultaneously acquire the object signal 30 and the auxiliary signal 35. Even if it is not a multimodality imaging device, it is possible to acquire the auxiliary signal 35 equivalent to the multimodality imaging device by photographing the same imaging object using a plurality of different medical imaging devices and performing position alignment processing.

[0053] The auxiliary signal 35 may also be obtained by performing signal processing on the object signal 30. For example, the auxiliary signal 35 may also be obtained by performing noise reduction processing, contrast emphasis processing, histogram flattening processing, edge emphasis processing, hard segmentation processing, soft segmentation processing, multi-resolution decomposition processing, monochrome conversion processing, etc. on the object signal 30. The hard segmentation processing is a processing for performing region segmentation. The soft segmentation processing includes tissue composition ratio estimation processing. The multi-resolution decomposition processing includes wavelet transformation.

[0054] (Noise reduction processing and generation of low-noise synthetic signals)

[0055] Next, refer to Figures 4 to 7 The noise reduction unit 2 (see Figure 4 ) Reduce the target signal by 30 (refer to Figure 4 ) of the noise structure.

[0056] like Figure 4 As shown, the noise reduction unit 2 estimates a noise intensity signal 31, calculates a similarity index 32, calculates a filter coefficient 33, obtains a low noise signal 34, and obtains a low noise synthesis signal 36, thereby reducing the noise of the target signal 30. Each process will be described in detail below.

[0057] (Estimation of noise intensity signal)

[0058] like Figure 4 and Figure 5 As shown, the noise intensity signal estimation unit 20 acquires the object signal 30 from the storage unit 3, and estimates the noise intensity signal 31 representing the intensity distribution of the noise of the object signal 30 based on the object signal 30. Specifically, the noise intensity signal estimation unit 20 applies the first filter 40 having the first filter window to the object signal 30 to acquire the low-frequency signal 30a. The noise intensity signal estimation unit 20 acquires the high-frequency signal 30b based on the object signal 30 and the low-frequency signal 30a. Specifically, the noise intensity signal estimation unit 20 acquires the high-frequency signal 30b by differentiating the object signal 30 with the low-frequency signal 30a.

[0059] Then, the noise intensity signal estimation unit 20 applies a second filter 41 having a second filter window independent of the first filter window to the high frequency signal 30b to estimate the noise intensity signal 31. The noise intensity signal estimation unit 20 applies a standard deviation filter different from the smoothing filter to the high frequency signal 30b to estimate the noise intensity signal 31. In the present embodiment, the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 using a standard deviation filter having a second filter window larger than the first filter window of the smoothing filter. The noise intensity signal estimation unit 20 uses a smoothing filter such as a moving average filter, a Gaussian filter, a median filter, a weighted median filter, a bilateral filter, a non-local mean filter, a guided filter, etc. as a smoothing filter. The guided filter is a filter that can reduce noise while preserving the edge (high frequency signal 30b) by using an input image and a guide image.

[0060] (Model of noise distribution)

[0061] The object signal 30 includes noise caused by the acquisition principle when acquiring the image signal (object signal 30) and noise caused by the device that acquires the image signal. These noises have a predetermined distribution based on the acquisition principle and the device. For example, Figure 6 As shown, sometimes the intensity of noise is distributed in a downward convex shape.

[0062] Figure 6 The graph 80 shown is a graph showing the distribution of noise generated by the acquisition principle when acquiring a PET image. In the graph 80, the vertical axis is the noise intensity and the horizontal axis is the body axis coordinate.

[0063] When acquiring a PET image, the imaging object 90 is arranged inside the cylindrical detector 6. In this case, the detection sensitivity of the radiation in the central portion of the cylindrical detector 6 becomes high, and the detection sensitivity becomes low as it approaches the end of the detector 6. That is, when acquiring a PET image, the noise intensity becomes low at a position close to the central portion of the detector 6, and the noise intensity tends to become high as it approaches the end of the detector 6.

[0064] A curve 80a shown in the graph 80 shows the distribution of noise. A dotted line 80b is the body axis coordinate corresponding to the end of one side of the detector 6, and a dotted line 80c is the body axis coordinate corresponding to the end of the other side of the detector 6. As shown in the curve 80a, the noise intensity becomes lower as the body axis coordinate approaches 0 (central part).

[0065] Therefore, in this embodiment, the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 based on the noise distribution of any one of the noise caused by the acquisition principle when acquiring the image signal and the noise caused by the device for acquiring the image signal, using the noise model 50 obtained by modeling the noise distribution.

[0066] The noise model 50 can be expressed by, for example, the following equation (1).

[0067] [Number 1]

[0068] σ prior (Z) = a*z p +b ... (1) a, b and p are parameters. The noise model 50 can be obtained by finding the parameters corresponding to the distribution of the noise. The noise model 50 is obtained in advance and stored in the storage unit 3 (see Figure 4 ). In addition, as long as the noise model 50 can be used when performing noise reduction processing on the object signal 30, it can be acquired before or after acquiring the object signal 30.

[0069] Refer again Figure 5 The noise intensity signal estimation unit 20 estimates the noise intensity signal 31 by multiplying the high frequency signal 30b by the noise model 50 after applying the second filter 41. Specifically, the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 based on the following equation (2). The noise model 50 includes a uniform noise model (a noise model that does not substantially change the value of the high frequency signal 30b even if multiplied).

[0070] [Number 2]

[0071]

[0072] σ m is the noise intensity signal 31 after multiplying the noise model 50. In addition, σ prior is the noise model 50. Also, σ is the noise intensity signal 31 before being multiplied by the noise model 50. Also, the cross symbol in the circle means the product per pixel.

[0073] Refer again Figure 4The noise intensity signal estimation unit 20 may also be configured to estimate the noise intensity signal 31 using the auxiliary signal 35. In this case, the noise intensity signal estimation unit 20 obtains the auxiliary signal 35 from the storage unit 3. Then, the noise intensity signal estimation unit 20 uses an edge-preserving smoothing filter such as a joint bilateral filter or a joint non-local means filter using the auxiliary signal 35 to obtain the high-frequency signal 30b (see Figure 5 ).

[0074] The noise intensity signal estimation unit 20 outputs the estimated noise intensity signal 31 to the similarity index calculation unit 21 .

[0075] (Calculation of similarity index)

[0076] like Figure 4 As shown, the similarity index calculation unit 21 obtains the object signal 30 from the storage unit 3 and obtains the noise intensity signal 31 from the noise intensity signal estimation unit 20. The similarity index calculation unit 21 calculates a similarity index 32 between a prescribed signal obtained at a first signal point and a signal obtained at a second signal point adjacent to the first signal point based on the object signal 30 and the noise intensity signal 31.

[0077] The similarity index 32 is an index based on the similarity between the signal acquired at the first signal point and the signal acquired at the second signal point. The similarity index 32 includes, for example, a signal deviation index defined for any two pixels using the following equation (3).

[0078] [Number 3]

[0079]

[0080] Δ 1jk is the similarity index 32. σ j is the value of pixel j of the noise intensity signal 31 (image σ). j is a circular template window centered at pixel j. j |It is T j The number of pixels contained in t j By T j The template window may also be a rectangular window.

[0081] The similarity index calculation unit 21 may be configured to calculate the similarity index 32 using the following formula (4) instead of the above formula (3).

[0082] [Number 4]

[0083]

[0084] For example, when the object signal 30 is a multi-channel image having a plurality of pixel values ​​for one pixel such as an RGB image, the similarity index calculation unit 21 may calculate the similarity index 32 using the following equation (5) or (6).

[0085] [Number 5]

[0086]

[0087] x j =(x j1 , x j2 ,···,x jNc ) is the channel of pixel j (channel number N c ) is a vector obtained by summarizing the pixel values ​​of the two vectors. d(·,·) is an arbitrary distance function with two vectors as independent variables. The distance function includes, for example, any one of the square distance, the first power distance, the Lp distance, the Huber distance, the Kullback-Leibler distance, and the cross entropy.

[0088] The similarity index calculation unit 21 is configured to use the auxiliary signal 35 when calculating the similarity index 32. Specifically, the similarity index calculation unit 21 obtains the auxiliary signal 35 from the storage unit 3. Then, the similarity index calculation unit 21 calculates the similarity index 32 using the auxiliary signal 35 based on the following equation (7).

[0089] [Number 6]

[0090]

[0091] Δ 2jk is the similarity index 32 when the auxiliary signal 35 is used. j is the template vector of the auxiliary signal 35, ρ j is the noise intensity signal of the auxiliary signal 35 .

[0092] Alternatively, the similarity index 32 (Δ 1jk ) and the similarity index 32 (Δ 2jk ) is set as the final similarity index 32. In other words, Δ 1jk and Δ 2jk to combine.

[0093] The similarity index calculation unit 21 outputs the calculated similarity index 32 to the filter coefficient calculation unit 22 .

[0094] (Calculation of filter coefficients)

[0095] The filter coefficient calculation section 22 acquires the similarity index 32 from the similarity index calculation section 21 , and calculates a filter coefficient 33 for the noise reduction filter based on the similarity index 32 .

[0096] Specifically, the filter coefficient calculation unit 22 calculates the filter coefficient 33 based on the following equation (8).

[0097] [Number 7]

[0098]

[0099] W 1jk is the filter coefficient 33. In addition, S j is a circular filter window centered at pixel j. In addition, p ≥ 0. The filter window can also be a rectangular window.

[0100] In the present embodiment, the filter coefficient calculation unit 22 may be configured to calculate the filter coefficient 33 by using the object signal 30 while using the similarity index 32. Specifically, the filter coefficient calculation unit 22 acquires the object signal 30 from the storage unit 3. Then, the filter coefficient calculation unit 22 calculates the filter coefficient 33 by using the object signal 30 while using the similarity index 32 based on the following formula (9).

[0101] [Number 8]

[0102]

[0103] W 2jk is the filter coefficient 33 calculated using the object signal 30 while using the similarity index 32. 3jk These are coefficients of the bilateral filter acquired based on the object signal 30. The coefficients of the bilateral filter can be calculated by the product of the "coefficient based on the difference in pixel values" and the "coefficient based on the distance between pixels".

[0104] In the present embodiment, the filter coefficient calculation unit 22 may be configured to calculate the filter coefficient 33 using the auxiliary signal 35. Specifically, the auxiliary signal 35 is acquired from the storage unit 3. Then, the filter coefficient calculation unit 22 calculates the filter coefficient 33 using the auxiliary signal 35 based on the following equation (10).

[0105] [Number 9]

[0106]

[0107] W 4jkIt is the filter coefficient 33 calculated using the auxiliary signal 35. The filter coefficient calculation unit 22 can also be configured to calculate the filter coefficient 33 based on the product of two exponential functions in the above formula (10), but the filter coefficient 33 can also be calculated based on any one of the weighted arithmetic mean, weighted geometric mean, and weighted harmonic mean of the two exponential functions.

[0108] The filter coefficient calculation unit 22 outputs the calculated filter coefficient 33 to the filter processing unit 23 .

[0109] (Filtering)

[0110] The filter processing unit 23 acquires the object signal 30 from the storage unit 3. In addition, the filter processing unit 23 acquires the filter coefficient 33 from the filter coefficient calculation unit 22. Then, the filter processing unit 23 applies a noise reduction filter to the object signal 30 based on the filter coefficient 33 to acquire a low noise signal 34. The noise reduction filter is a filter for reducing the noise of the object signal 30. In addition, the low noise signal 34 is a signal with reduced noise compared to the object signal 30.

[0111] In the present embodiment, the filter processing unit 23 obtains the low-noise signal 34 using the following equations (11) and (12).

[0112] [Number 10]

[0113]

[0114] m j is a low noise signal 34. In addition, W jk It is the filter coefficient 33.

[0115] The filter processing unit 23 may also obtain the low-noise signal 34 using the following equation (13).

[0116] [Number 11]

[0117]

[0118] m is the low noise signal 34. Φ(·) is an arbitrary potential function. β is a hyperparameter. The potential function includes, for example, any one of the Lp norm function and the Huber function.

[0119] In this embodiment, the filter processing unit 23 may obtain the low noise signal 34 by using the noise intensity signal 31 together with the filter coefficient 33. Specifically, the filter processing unit 23 may obtain the low noise signal 34 based on the following equation (14).

[0120] [Number 12]

[0121]

[0122] σ x is a representative statistic of σ. The representative statistic includes, for example, any one of a mean value, a trimmed mean value, a median value, a maximum value, and (maximum value+minimum value) / 2.

[0123] The filter processing unit 23 may also obtain the low noise signal 34 based on the auxiliary signal 35. Specifically, the filter processing unit 23 obtains the auxiliary signal 35 from the storage unit 3. Then, the filter processing unit 23 may also obtain a signal obtained by synthesizing the auxiliary signal 35 and the low noise signal 34 as the low noise signal 34. The synthesis of the auxiliary signal 35 and the low noise signal 34 can be performed as long as it can be performed through image fusion processing. Image fusion processing is an image synthesis processing including synthesis based on space, synthesis in the frequency domain, and synthesis using deep learning (learning completed model).

[0124] The filter processing unit 23 outputs the acquired low noise signal 34 to the noise intensity signal estimation unit 20. Thereafter, the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 based on the low noise synthesis signal 36. In this way, the noise reduction process performed by the noise reduction unit 2 may be repeated a predetermined number of times. Then, when the noise reduction process is repeated a predetermined number of times, the filter processing unit 23 outputs the low noise signal 34 to the synthesis processing unit 24. The filter processing unit 23 may also be configured to store the acquired low noise signal 34 in the storage unit 3.

[0125] (Synthesis Processing)

[0126] The synthesis processing unit 24 obtains the object signal 30 from the storage unit 3 and obtains the low noise signal 34 from the filter processing unit 23. Then, the synthesis processing unit 24 synthesizes the object signal 30 and the low noise signal 34 to obtain a low noise synthesis signal 36 to adjust the noise reduction effect. The low noise synthesis signal 36 is a low noise signal after the noise reduction effect is adjusted. In this embodiment, the synthesis processing unit 24 synthesizes the object signal 30 and the low noise signal 34 in a weighted state to obtain the low noise synthesis signal 36. Specifically, Figure 7 As shown, the synthesis processing unit 24 (see Figure 4 ) The object signal 30 is multiplied by the first weight 60, and the low noise signal 34 is multiplied by the second weight 61. Then, the synthesis processing unit 24 adds the object signal 30 multiplied by the first weight 60 and the low noise signal 34 multiplied by the second weight 61 to obtain the low noise synthesis signal 36.

[0127] The acquisition of the low-noise composite signal 36 by the composite processing unit 24 can be expressed by the following equation (15).

[0128] [Number 13]

[0129] x * =(1-r)x+rm…(15)

[0130] x * is the low noise synthesis signal 36. x is the object signal 30. In addition, (1-r) is the first weight 60. r is the second weight 61. m is the low noise signal 34.

[0131] The synthesis processing unit 24 may be configured to obtain the low-noise synthesis signal 36 based on the following equation (16) or equation (17).

[0132] [Number 14]

[0133]

[0134] N j is the set of neighboring pixels of pixel j, d jk It is a value proportional to the distance between pixel j and pixel k.

[0135] Refer again Figure 4 In the present embodiment, the synthesis processing unit 24 may be configured to obtain the low-noise synthesis signal 36 by using at least one of the similarity index 32 and the filter coefficient 33 while using the target signal 30 and the low-noise signal 34 .

[0136] Specifically, when using the similarity index 32, the synthesis processing unit 24 acquires the similarity index 32 from the similarity index calculation unit 21. Then, the synthesis processing unit 24 acquires the low-noise synthesis signal 36 based on the following equation (18).

[0137] [Number 15]

[0138]

[0139] When using the filter coefficient 33, the synthesis processing unit 24 obtains a low-noise synthesis signal 36 based on the following equation (19).

[0140] [Number 16]

[0141]

[0142] The synthesis processing unit 24 may also obtain a low-noise synthesis signal 36 based on the auxiliary signal 35. Specifically, the synthesis processing unit 24 obtains the auxiliary signal 35 from the storage unit 3. Then, the synthesis processing unit 24 obtains a signal obtained by synthesizing the auxiliary signal 35 and the low-noise synthesis signal 36 as the low-noise synthesis signal 36. The synthesis of the auxiliary signal 35 and the low-noise synthesis signal 36 can be performed as long as it can be performed by image fusion processing.

[0143] The synthesis processing unit 24 stores the acquired low noise synthesis signal 36 in the storage unit 3. The synthesis processing unit 24 may also be configured to display the low noise synthesis signal 36 on the display unit 5 (see Figure 1 ) shows the low noise synthetic signal 36.

[0144] The processing by the noise intensity signal estimation unit 20, the processing by the similarity index calculation unit 21, the processing by the filter coefficient calculation unit 22, the processing by the filter processing unit 23, and the processing by the synthesis processing unit 24 may also be based on the reference data set 51 (refer to Figure 1 ). The reference data set 51 includes at least one of a signal of a sample different from the object sample of the acquired object signal 30 and a learned model generated by machine learning based on the training data. The processing performed by the functional units 20 to 24 can also be performed using the learned model. The learned model can be generated by learning the input and output in each processing as training data.

[0145] (Noise reduction processing)

[0146] Next, refer to Figure 8 The noise reduction unit 2 (see Figure 4 ) Reduce the target signal by 30 (refer to Figure 4 ) noise processing.

[0147] In step 110, the noise reduction unit 2 acquires the target signal 30. Specifically, the noise reduction unit 2 obtains the target signal 30 from the target signal acquisition unit 1 (see Figure 1 ) to obtain the object signal 30. The object signal 30 is stored in the storage unit 3 (refer to Figure 4 ), the noise reduction unit 2 may also obtain the object signal 30 from the storage unit 3.

[0148] In step 111, the noise reduction unit 2 stores the auxiliary signal 35 (see Figure 4 )'s storage unit 3 is accessed to obtain the auxiliary signal 35.

[0149] In step 112, the noise reduction unit 2 obtains a noise model 50 (see FIG. 5 ) of noise distribution based on the noise distribution of any one of the noise caused by the acquisition principle when acquiring the image signal and the noise caused by the device that acquires the image signal. Figure 4 ). The noise model 50 is acquired in advance and stored in the storage unit 3. Therefore, the noise reduction unit 2 stores the noise model 50 (refer to Figure 4 )'s storage unit 3 is accessed to obtain the noise model 50.

[0150] In step 113, the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 (see FIG. 1 ) indicating the intensity distribution of the noise of the object signal 30 based on the object signal 30. Figure 4 Specifically, the noise intensity signal estimation unit 20 applies a first filter 40 having a first filter window to the target signal 30 (see Figure 5 ) to obtain the low frequency signal 30a (refer to Figure 5 ). In addition, the noise intensity signal estimation unit 20 obtains the high-frequency signal 30b based on the object signal 30 and the low-frequency signal 30a (see Figure 5 ). Then, the noise intensity signal estimator 20 applies a second filter 41 having a second filter window independent of the first filter window to the high frequency signal 30b to estimate the noise intensity signal 31. In the present embodiment, the noise intensity signal estimator 20 applies a standard deviation filter having a filter window larger than the filter window of the smoothing filter to the high frequency signal 30b to estimate the noise intensity signal 31. In the present embodiment, the noise intensity signal estimator 20 estimates the noise intensity signal 31 by multiplying the high frequency signal 30b by the noise model 50 after applying the second filter 41 to the high frequency signal 30b in step 113.

[0151] In step 114, the similarity index calculation unit 21 (refer to Figure 4 ) Based on the object signal 30 and the noise intensity signal 31, a similarity index 32 (refer to Figure 4 ).

[0152] In step 115, the filter coefficient calculation unit 22 (see Figure 4 ) Based on the similarity index 32, the filter coefficient 33 of the noise reduction filter is calculated (refer to Figure 4 ). In the present embodiment, the filter coefficient calculation unit 22 calculates the filter coefficient 33 using the object signal 30 while using the similarity index 32 in step 115 .

[0153] In step 116, the filter processing unit 23 (see Figure 4 ) applies a noise reduction filter to the object signal 30 based on the filter coefficient 33 to obtain a low noise signal 34. In the present embodiment, the filter processing unit 23 uses the noise intensity signal 31 while using the filter coefficient 33 in step 116 to obtain the low noise signal 34.

[0154] In step 117, the noise reduction unit 2 determines whether the noise reduction process has been performed a predetermined number of times. Specifically, the noise reduction unit 2 sets the process of steps 113 to 116 as one set, and determines whether the process of steps 113 to 116 has been performed a predetermined number of times. If the noise reduction process has been performed a predetermined number of times, the process proceeds to step 118. If the noise reduction process has not been performed a predetermined number of times, the process proceeds to step 113. In this case, the process of step 113 is performed based on the low noise signal 34 acquired in step 116. That is, a series of processes of step 113 of estimating the noise intensity signal 31, step 114 of calculating the similarity index 32, step 115 of calculating the filter coefficient 33, and step 116 of acquiring the low noise signal 34 are repeated using the low noise signal 34 as the object signal 30. In addition, the predetermined number of sets may be set in advance by the user, or may be automatically determined, for example, by repeating until the amount of change in noise reduction becomes less than a specified value.

[0155] In step 118, the synthesis processing unit 24 (see Figure 4 ) synthesizes the object signal 30 and the low noise signal 34 to obtain a low noise synthesized signal 36 to adjust the noise reduction effect. In the present embodiment, the synthesis processing unit 24 synthesizes the object signal 30 and the low noise signal 34 in a weighted state to obtain the low noise synthesized signal 36 in step 118. The synthesis processing unit 24 uses the noise intensity signal 31 while using the object signal 30 and the low noise signal 34 in step 118 to obtain the low noise synthesized signal 36. In addition, the synthesis processing unit 24 may also use one or both of the similarity index 32 and the filter coefficient 33 while using the object signal 30 and the low noise signal 34 in step 118 to obtain the low noise synthesized signal 36.

[0156] Thereafter, the noise reduction unit 2 stores the low noise synthesis signal 36 in the storage unit 3 or displays it on the display unit 5 (see Figure 1 ) and ends the processing.

[0157] In the present embodiment, the auxiliary signal 35 is used for processing in at least any one of the following cases: when estimating the noise intensity signal 31 (step 113), calculating the similarity index 32 (step 114), calculating the filter coefficient 33 (step 115), acquiring the low noise signal 34 (step 116), and acquiring the low noise synthesis signal 36 (step 118). In addition, the auxiliary signal 35 includes the low noise signal 34. Therefore, in the present embodiment, the low noise signal 34 is used as the auxiliary signal 35 for processing in at least any one of the following cases: when estimating the noise intensity signal 31 (step 113), calculating the similarity index 32 (step 114), calculating the filter coefficient 33 (step 115), and acquiring the low noise signal 34 (step 116). The auxiliary signal 35 may be used in all the processes of step 113 to step 116, may be used in only one of the steps, or may be used in multiple processes of step 113 to step 116.

[0158] In addition, in the present embodiment, in at least any one of estimating the noise intensity signal 31 (step 113), calculating the similarity index 32 (step 114), calculating the filter coefficient 33 (step 115), acquiring the low noise signal 34 (step 116), and acquiring the low noise synthetic signal 36 (step 118), a reference data set 51 (reference data set 51) is used. Figure 1 ) processing. The reference data set 51 may be used in the processing of each step in steps 113 to 116 and step 118, or may be used in only one step, or may be used in multiple processing in steps 113 to 116 and step 118.

[0159] The processing of step 110 to step 112 is irrelevant in the order in which the processing is executed.

[0160] (Example)

[0161] Next, refer to Figures 9 to 18 The following is an experiment to confirm that noise can be reduced while suppressing loss of high-frequency components.

[0162] Fig. 9 The original image 131 shown is a slice image of a predetermined cross section in a Shepp-Logan phantom image that is frequently used in computer simulation experiments such as X-ray CT and PET. Fig.10 The noise added image 132 shown is an image obtained by adding Gaussian noise having a noise standard deviation of 0.05 to the original image 131 .

[0163] Fig.11The low noise image 132a in the first comparative example shown is obtained by using the original image 131 (refer to Fig. 9 ) is an image obtained by applying a 3×3 standard deviation filter to estimate the noise. In addition, Fig.12 The edge comparison image 140a of the first comparison example shown is a Fig.11 The low noise image 132a of the first comparative example shown is obtained by subtracting the original image 131. The mean absolute error of the edge as the object in the edge comparison image is obtained as the evaluation result that the high frequency signal 30b is not damaged. The smaller the value of the mean absolute error, the smaller the error with the original image 131 and the less damaged the high frequency signal 30b is. In the first comparative example, the mean absolute error 141a of the edge comparison image 140a is 0.0274.

[0164] Fig.13 The low noise image 132b in the second comparative example shown is obtained by using the original image 131 (refer to Fig. 9 ) is an image obtained by applying a 5×5 standard deviation filter to estimate the noise. Fig.14 The edge comparison image 140b of the second comparative example shown is a Fig.13 The low noise image 132b of the second comparative example shown is obtained by subtracting the original image 131. The mean absolute error 141b obtained in the edge comparison image 140b is 0.0363. It can be confirmed that when the filter window of the standard deviation filter is increased to improve the estimation accuracy of the noise, the high frequency component is damaged.

[0165] Fig.15 The low noise image 133 in the first embodiment shown is an image obtained by estimating the noise intensity signal 31 by making the size of the smoothing filter different from the size of the standard deviation filter. Fig. 9 ) Apply 3×3 average filtering to obtain the low-frequency signal 30a (refer to Figure 5 ). The original image 131 is differentiated by the acquired low-frequency signal 30a to acquire the high-frequency signal 30b (refer to Figure 5 ). A 5×5 standard deviation filter is applied to the acquired high frequency signal 30b to estimate the noise intensity signal 31. Then, based on the structure of the above embodiment, a low noise signal 34 is acquired (see Figure 4 ). Fig.15 The illustrated low-noise image 133 is an image based on the low-noise signal 34 .

[0166] Fig.16 The edge comparison image 134 of the first embodiment shown is a Fig.15 The low noise image 133 of the first embodiment shown in FIG. 1 is different from the original image 131 (see Fig. 9 ) is obtained by performing a difference operation on the edge comparison image 134. The mean absolute error 135 obtained in the edge comparison image 134 is 0.0134. This value is smaller than the value of the mean absolute error 141a of the first comparison example (0.0274) and the value of the mean absolute error 141b of the second comparison example (0.0363). It has been confirmed that by making the size of the filter window of the smoothing filter different from the size of the filter window of the standard deviation filter, it is possible to suppress the loss of high-frequency components.

[0167] Fig.17 The low noise image 136 of the second embodiment shown in the figure is obtained by converting the low noise image 133 (see Fig.15 ) and the original image 131 (refer to Fig. 9 ) are weighted and added. Specifically, the low-noise image 136 of the second embodiment is obtained by multiplying the low-noise image 133 by a weight of 0.7 and the original image 131 by a weight of 0.3 and then adding them. When the edge comparison image is obtained based on the low-noise image 136 of the second embodiment to obtain the mean absolute error 137, it is 0.0191. This value is smaller than the value of the mean absolute error 141a of the first comparison example (0.0274) and the value of the mean absolute error 141b of the second comparison example (0.0363). Therefore, it is confirmed that in the structure of the second embodiment in which the low-noise image 136 and the original image 131 are weighted and then added, it is also possible to suppress the loss of high-frequency components.

[0168] Fig.18 The edge comparison image 138 of the third embodiment shown is an edge comparison image obtained based on a low noise image obtained by repeatedly performing noise reduction processing. Fig. 9 ) and the low noise image 133 of the first embodiment (refer to Fig.15 ) is filtered using a 5×5 standard deviation filter to estimate the noise intensity signal 31 (refer to Figure 4 ), based on the structure of the above embodiment, a low noise signal 34 is obtained (refer to Figure 4 ) and set it as a low noise image. Then, the edge comparison image 138 of the third embodiment is obtained by subtracting the low noise image from the original image 131. That is, the low noise image of the third embodiment is an image after the noise reduction process is repeated twice.

[0169] When the mean absolute error 139 is obtained based on the edge comparison image 138 of the third embodiment, it is 0.0104. This result is a value smaller than the value (0.0274) of the mean absolute error 141a of the first comparative example and the value (0.0363) of the mean absolute error 141b of the second comparative example. In addition, it is a value smaller than the value (0.0134) of the mean absolute error 135 of the first embodiment. Therefore, it is confirmed that by repeating the noise reduction process, the damage to the high-frequency component can be further suppressed.

[0170] (Effects of this embodiment)

[0171] In this embodiment, the following effects can be obtained.

[0172] In the present embodiment, as described above, the noise reduction processing method is a noise reduction processing method for an object signal 30, and the noise reduction processing method includes the following steps: step 113, estimating a noise intensity signal 31 representing the intensity distribution of noise of the object signal 30 based on the object signal 30; step 114, calculating a similarity index 32 between a specified signal obtained at a first signal point and a signal obtained at a second signal point adjacent to the first signal point based on the object signal 30 and the noise intensity signal 31; step 115, calculating a filter coefficient 33 of a noise reduction filter based on the similarity index 32; and step 116, applying a noise reduction filter to the object signal 30 based on the filter coefficient 33 to obtain a low noise signal 34, and in step 113 of estimating the noise intensity signal 31, performing the following processing: applying a first filter 40 having a first filter window to the object signal 30 to obtain a low-frequency signal 30a; obtaining a high-frequency signal 30b based on the object signal 30 and the low-frequency signal 30a; and applying a second filter 41 having a second filter window independent of the first filter window to the high-frequency signal 30b to estimate the noise intensity signal 31.

[0173] Thus, the size of the first filter window and the size of the second filter window can be set independently of each other, so that the high-frequency component as the target can be suppressed from being damaged. In addition, the second filter 41 having the second filter window independent of the first filter window when the low-frequency signal 30a is acquired is applied to the high-frequency signal 30b to estimate the noise intensity signal 31. Therefore, the size of the first filter window and the size of the second filter window can be set according to the target signal 30, so as to estimate the noise intensity signal 31. Therefore, by changing the size of the second filter window according to the target signal 30, the statistical variation of the noise level in the estimated noise intensity signal 31 can be suppressed, so that the estimation accuracy of the noise intensity signal 31 can be improved. Therefore, the accuracy of the similarity index 32 acquired based on the noise intensity signal 31 can be improved. In addition, since the accuracy of the similarity index 32 can be improved, the filter coefficient 33 can be calculated with high accuracy. As a result, the noise can be reduced while suppressing the damage of the high-frequency component as the target.

[0174] In addition, in the present embodiment, as described above, the noise reduction processing device 100 is a noise reduction processing device for an object signal 30, and the noise reduction processing device 100 includes: an object signal acquisition unit 1, which acquires the object signal 30; and a noise reduction unit 2, which reduces the noise of the object signal 30, and the noise reduction unit 2 is configured to perform the following processing: applying a first filter 40 having a first filter window to the object signal 30 to acquire a low-frequency signal 30a, acquiring a high-frequency signal 30b based on the object signal 30 and the low-frequency signal 30a, and applying a second filter 41 having a second filter window independent of the first filter window to the high-frequency signal 30b to estimate a noise intensity signal 31; calculating a similarity index 32 between a specified signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point based on the object signal 30 and the noise intensity signal 31; calculating a filter coefficient 33 of a noise reduction filter based on the similarity index 32; and applying a noise reduction filter to the object signal 30 based on the filter coefficient 33 to acquire a low-noise signal 34.

[0175] Thus, it is possible to provide the noise reduction processing apparatus 100 that can reduce noise while suppressing damage to target high-frequency components, similarly to the above-mentioned noise reduction processing method.

[0176] In addition, in the present embodiment, as described above, the noise reduction processing method is a noise reduction processing method for the object signal 30, and the noise reduction processing method includes the following steps: step 113, estimating a noise intensity signal 31 representing the intensity distribution of the noise of the object signal 30 based on the object signal 30; step 114, calculating a similarity index 32 between a specified signal obtained at a first signal point and a signal obtained at a second signal point adjacent to the first signal point based on the object signal 30 and the noise intensity signal 31; step 115, calculating a filter coefficient 33 of a noise reduction filter based on the similarity index 32; step 116, applying a noise reduction filter to the object signal 30 based on the filter coefficient 33 to obtain a low noise signal 34; and step 118, synthesizing the object signal 30 and the low noise signal 34 to obtain a low noise synthesized signal 36 to adjust the noise reduction effect.

[0177] Thus, by synthesizing the object signal 30 and the low noise signal 34, it is possible to use the object signal 30 to suppress the damage to the high frequency component of the object, and to use the low noise signal 34 to reduce the noise. As a result, it is possible to provide a noise reduction processing method that can reduce the noise while suppressing the damage to the high frequency component of the object. In addition, the object signal 30 and the low noise signal 34 are synthesized to obtain the low noise synthesis signal 36 to adjust the noise reduction effect. Thus, by adjusting the ratio of synthesizing the object signal 30 and the low noise signal 34, the noise reduction effect can be easily adjusted. As a result, compared with a structure in which a series of processes are performed again according to the object signal 30 to obtain the low noise signal 34 after changing the parameters, the time required for the adjustment process of the noise reduction effect can be shortened.

[0178] In addition, in the present embodiment, as described above, the noise reduction processing device 100 is a noise reduction processing device for an object signal 30, and the noise reduction processing device 100 comprises: an object signal acquisition unit 1, which acquires the object signal 30; and a noise reduction unit 2, which reduces the noise of the object signal 30, wherein the noise reduction unit 2 is configured to perform the following processing: based on the object signal 30, estimating a noise intensity signal 31 representing the intensity distribution of the noise of the object signal 30; based on the object signal 30 and the noise intensity signal 31, calculating a similarity index 32 between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; based on the similarity index 32, calculating a filter coefficient 33 of a noise reduction filter; based on the filter coefficient 33, applying a noise reduction filter to the object signal 30 to acquire a low noise signal 34, and synthesizing the object signal 30 and the low noise signal 34 to acquire a low noise synthesized signal 36 to adjust the noise reduction effect.

[0179] Thus, it is possible to provide a noise reduction processing device 100 that can reduce noise while suppressing damage to the high-frequency components of the object, as in the above-mentioned noise reduction processing method. In addition, it is possible to provide a noise reduction processing device 100 that can shorten the time required for the adjustment process of the noise reduction effect compared to a structure in which a series of processes are repeated to obtain the low-noise signal 34 based on the object signal 30 after changing the parameters.

[0180] In addition, in the above-mentioned embodiment, by configuring as follows, further effects as described below can be obtained.

[0181] That is, in the present embodiment, as described above, the first filter 40 includes a smoothing filter, and the second filter 41 includes a standard deviation filter. In step 113 of estimating the noise intensity signal 31, a standard deviation filter different from the smoothing filter is applied to the high-frequency signal 30b to estimate the noise intensity signal 31. Thus, for example, by using a standard deviation filter having a filter window larger than that of the smoothing filter, the noise intensity signal 31 is estimated from the high-frequency signal 30b, thereby making it easy to improve the estimation accuracy of the noise intensity signal 31. As a result, noise can be reduced with higher accuracy.

[0182] In addition, in the present embodiment, as described above, the object signal 30 is an image signal of two dimensions or more, and further includes step 112, in which a noise model 50 of noise distribution is acquired based on the noise distribution of any one of the noise caused by the acquisition principle when acquiring the image signal and the noise caused by the device that acquires the image signal, and in step 113 of estimating the noise intensity signal 31, the noise intensity signal 31 is estimated by multiplying the high-frequency signal 30b by the noise model 50 after applying the second filter 41. Thus, the noise intensity signal 31 reflecting the noise distribution of any one of the noise caused by the acquisition principle of the image signal and the noise caused by the device can be estimated using the noise model 50. As a result, the estimation accuracy of the noise intensity signal 31 can be further improved.

[0183] In addition, in the present embodiment, as described above, the low noise signal 34 is used as the object signal 30, and a series of processes of step 113 of estimating the noise intensity signal 31, step 114 of calculating the similarity index 32, step 115 of calculating the filter coefficient 33, and step 116 of acquiring the low noise signal 34 are repeatedly performed. Thus, the noise reduction process is performed based on the low noise signal 34 with reduced noise compared to the object signal 30, so that the noise can be further reduced compared to the configuration in which the noise reduction process is performed only once on the object signal 30.

[0184] In addition, in the present embodiment, as described above, in step 116 of acquiring the low noise signal 34, the noise intensity signal 31 is used while using the filter coefficient 33 to acquire the low noise signal 34. Thus, the low noise signal 34 is acquired while using the noise intensity signal 31 while using the similarity index 32 and the filter coefficient 33, and therefore, the noise intensity signal 31 can be used as a parameter in the calculation of the low noise signal 34. As a result, the low noise signal 34 can be acquired while reflecting the distribution of the noise of the object signal 30, and therefore, compared with the configuration in which the filter coefficient 33 is calculated only from the object signal 30, the low noise signal 34 with reduced noise can be acquired with high accuracy.

[0185] In addition, in the present embodiment, as described above, in step 115 of calculating the filter coefficient 33, the filter coefficient 33 is calculated using the object signal 30 while using the similarity index 32. Thus, for example, the filter coefficient 33 can be calculated based on the similarity index 32 and the filter coefficient calculated from the object signal 30 itself. As a result, the filter coefficient calculated from the object signal 30 itself is further used as a parameter used in the calculation of the filter coefficient 33, thereby making it possible to improve the certainty of the filter coefficient 33.

[0186] In addition, in the present embodiment, as described above, step 111 is further included, in which step 111, an auxiliary signal 35 is acquired for the same object as the object for which the object signal 30 is acquired, and the auxiliary signal 35 is a signal acquired using a device different from the device for acquiring the object signal 30, an acquisition condition different from the condition when the object signal 30 is acquired, or a processing condition different from the object signal 30, and the auxiliary signal 35 is used for processing in at least any one of the following cases: when estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, when calculating the similarity index 32 in step 114 of calculating the similarity index 32, when calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, and when acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34.

[0187] The auxiliary signal 35 is a signal acquired by a different device, acquisition condition, or processing condition than the object signal 30, and is a signal containing information different from that of the object signal 30 for the same object as the object signal 30. Therefore, for example, when estimating the noise intensity signal 31 from the object signal 30 having a low resolution, the noise intensity signal 31 can be estimated while easily saving the high-frequency signal 30b as the object by using a high-resolution signal as the auxiliary signal 35. Therefore, the estimation accuracy of the noise intensity signal 31 can be improved. In addition, for example, by using a signal having a higher SN ratio than the SN ratio of the noise intensity signal 31 as the auxiliary signal 35 to calculate the similarity index 32, the accuracy of the similarity index 32 can be improved compared with the case where the similarity index 32 is calculated from the noise intensity signal 31. In addition, for example, by using the auxiliary signal 35 when calculating the filter coefficient 33, the number of parameters used in the calculation of the filter coefficient 33 can be increased, and therefore the certainty of the filter coefficient 33 can be improved. In addition, for example, by synthesizing the auxiliary signal 35 with a high resolution with the low noise signal 34 when acquiring the low noise signal 34, it is possible to reduce noise while improving the visibility of the boundary line between the object and the background. As a result, by using the auxiliary signal 35 in any of the processing steps, the processing accuracy of the step using the auxiliary signal 35 can be improved compared to the case where the auxiliary signal 35 is not used, so that the noise can be reduced with high accuracy while suppressing the damage of the high frequency components of the object signal 30.

[0188] In addition, in the present embodiment, as described above, the auxiliary signal 35 includes the low noise signal 34, and the low noise signal 34 is used as the auxiliary signal 35 for processing in at least any of the following cases: when estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, when calculating the similarity index 32 in step 114 of calculating the similarity index 32, when calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, and when acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34. Thus, even when it is difficult to acquire a signal using a device different from the device that acquired the target signal 30, an acquisition condition different from the condition when the target signal 30 was acquired, or a processing condition different from the target signal 30, the low noise signal 34 can be used as the auxiliary signal 35. As a result, even when it is difficult to acquire a signal using a device different from the device that acquired the target signal 30, an acquisition condition different from the condition when the target signal 30 was acquired, or a processing condition different from the target signal 30, it is possible to reduce noise with high accuracy while suppressing the loss of high-frequency components of the target signal 30.

[0189] In addition, in the present embodiment, as described above, in at least any of the cases of estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, calculating the similarity index 32 in step 114 of calculating the similarity index 32, calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, and acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34, processing based on the reference data set 51 is performed. Thus, for example, by replacing the processing of the step with a low processing speed with the processing using the reference data set 51, the processing speed can be increased without changing the noise reduction processing that does not use the reference data set 51 and the overall structure of the noise reduction processing. As a result, the processing speed in the step with a low processing speed can be improved, and thus the processing speed of the overall noise reduction processing can be increased.

[0190] In addition, in the present embodiment, as described above, step 118 is further included, in which the object signal 30 and the low noise signal 34 are synthesized in a weighted state to obtain a low noise synthesis signal 36 to adjust the noise reduction effect. Thus, by adjusting the weights of the object signal 30 without noise reduction and the low noise signal 34 with noise reduction, the noise reduction effect can be easily adjusted. As a result, compared with a structure in which a series of processes are performed again to obtain the low noise signal 34 based on the object signal 30 after changing the parameters, the time required for the adjustment process of the noise reduction effect can be shortened.

[0191] In addition, in the present embodiment, as described above, in step 118 of acquiring the low noise composite signal 36, the object signal 30 and the low noise signal 34 are composited in a weighted state to acquire the low noise composite signal 36. Thus, by adjusting the weighting of the object signal 30 without noise reduction and the weighting of the low noise signal 34 with noise reduction, the noise reduction effect can be adjusted more easily.

[0192] In addition, in the present embodiment, as described above, in step 118 of acquiring the low noise synthesis signal 36, the low noise synthesis signal 36 is acquired by using the noise intensity signal 31 while using the object signal 30 and the low noise signal 34. Thus, the object signal 30 and the low noise signal 34 can be synthesized in a state where the distribution of noise is reflected based on the noise intensity signal 31, so that the noise reduction effect can be adjusted in more detail.

[0193] In addition, in the present embodiment, as described above, in step 118 of acquiring the low noise synthesis signal 36, the low noise synthesis signal 36 is acquired by using at least one of the similarity index 32 and the filter coefficient 33 while using the object signal 30 and the low noise signal 34. Thus, at least one of the similarity index 32 and the filter coefficient 33 acquired based on the object signal 30 can be used as a coefficient when calculating the low noise synthesis signal 36. As a result, unlike a configuration in which the similarity index 32 and the filter coefficient 33 are not used, the coefficient based on the object signal 30 can be used as a parameter, so that the accuracy of the low noise synthesis signal 36 can be improved.

[0194] In addition, in the present embodiment, as described above, step 112 is further included, in which an auxiliary signal 35 is acquired for the same object as the object for which the object signal 30 is acquired, and the auxiliary signal 35 is a signal acquired using a device different from the device for acquiring the object signal 30, an acquisition condition different from the condition when the object signal 30 is acquired, or a processing condition different from the object signal 30, and the auxiliary signal 35 is used for processing in at least any one of the following cases: when estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, when calculating the similarity index 32 in step 114 of calculating the similarity index 32, when calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, when acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34, and when acquiring the low noise synthetic signal 36 in step 118 of acquiring the low noise synthetic signal 36.

[0195] Thus, when the auxiliary signal 35 is used when estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, when calculating the similarity index 32 in step 114 of calculating the similarity index 32, when calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, and when acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34, as described above, it is possible to reduce the noise with high accuracy while suppressing the damage to the high frequency components of the object signal 30. In addition, when the auxiliary signal 35 is used when acquiring the low noise synthesis signal 36, for example, by synthesizing the auxiliary signal 35 with high resolution and the low noise signal 34, it is possible to reduce the noise while improving the visibility of the boundary line between the object and the background. As a result, in step 118 of acquiring the low noise synthesis signal 36, it is possible to reduce the noise with high accuracy while suppressing the damage to the high frequency components of the object signal 30, compared with the case where the auxiliary signal 35 is not used.

[0196] In addition, in the present embodiment, as described above, in at least any of the following cases, when estimating the noise intensity signal 31 in step 113 of estimating the noise intensity signal 31, when calculating the similarity index 32 in step 114 of calculating the similarity index 32, when calculating the filter coefficient 33 in step 115 of calculating the filter coefficient 33, when acquiring the low noise signal 34 in step 116 of acquiring the low noise signal 34, and when acquiring the low noise composite signal 36 in step 118 of acquiring the low noise composite signal 36, processing based on the reference data set 51 is performed. Thus, for example, by replacing the processing of the step with a low processing speed with the processing using the reference data set 51, the processing speed can be increased without changing the noise reduction processing that does not use the reference data set 51 and the overall structure of the noise reduction processing. As a result, the processing speed in the step with a low processing speed can be improved, so that the processing speed of the noise reduction processing and the processing of adjusting the noise reduction effect as a whole can be increased.

[0197] [Modifications]

[0198] The embodiments and examples disclosed herein are illustrative and non-restrictive in all aspects. The scope of the present invention is not indicated by the description of the embodiments and examples described above but by the claims, and includes all changes (modifications) within the meaning and scope equivalent to the claims.

[0199] For example, in the above embodiment, an example of a structure in which the noise intensity signal estimation unit 20 uses a smoothing filter to estimate the noise intensity signal 31 is shown, but the present invention is not limited thereto. For example, the noise intensity estimation unit may also be configured to estimate the noise intensity signal 31 by solving the optimization problem of the following equation (20).

[0200] [Number 17]

[0201]

[0202] y is the target signal 30. R(·) is a regularization function. The regularization function includes, for example, any one of a quadratic function, a Huber function, and a total variation function.

[0203] In addition, in the above-mentioned embodiment, an example of a structure in which the noise intensity signal estimation unit 20 estimates the noise intensity signal 31 by multiplying the high-frequency signal 30b by the noise model 50 after applying the second filter 41, but the present invention is not limited to this. For example, the noise intensity signal estimation unit 20 may not multiply the noise intensity signal 31 by the noise model 50. However, if the noise intensity signal 31 is not multiplied by the noise model 50, it is impossible to estimate the noise intensity signal 31 based on the distribution of noise caused by the acquisition principle when acquiring the image signal and the noise caused by the device for acquiring the image signal, and the estimation accuracy of the noise intensity signal 31 decreases. Therefore, the noise intensity signal estimation unit 20 is preferably configured to multiply the noise intensity signal 31 by the noise model 50.

[0204] In addition, in the above-mentioned embodiment, an example of a structure is shown in which the noise reduction unit 2 uses the low noise signal 34 as the object signal 30 to repeat a series of processes of step 113 of estimating the noise intensity signal 31, step 114 of calculating the similarity index 32, step 115 of calculating the filter coefficient 33, and step 116 of acquiring the low noise signal 34, but the present invention is not limited to this. For example, the noise reduction unit 2 may not repeat the processes of steps 113 to 116.

[0205] In addition, in the above-mentioned embodiment, an example of a structure in which the noise reduction unit 2 repeats the processing of step 113 to step 116 as a series of processing is shown, but the present invention is not limited to this. Fig.19 As shown, the noise reduction unit may be configured to repeat the processing of steps 113 to 116 and the processing of step 118 as a series of processing. That is, it is sufficient as long as the determination processing based on the repetition of step 117 can be performed after step 118. Fig.19 Each step in the flowchart shown is a step for performing the same processing as each step in the above-described embodiment.

[0206] In addition, in the above-mentioned embodiment, an example of a structure in which the filter processing unit 23 uses the noise intensity signal 31 to obtain the low noise signal 34 while using the filter coefficient 33 is shown, but the present invention is not limited to this. For example, the filter processing unit may be configured to obtain the low noise signal 34 by using the filter coefficient 33 instead of the noise intensity signal 31. However, when the filter processing unit uses the filter coefficient 33 instead of the noise intensity signal 31 to obtain the low noise signal 34, it is difficult to reduce the noise in the low noise signal 34 with high accuracy. Therefore, the filter processing unit 23 is preferably configured to obtain the low noise signal 34 by using the noise intensity signal 31 while using the filter coefficient 33.

[0207] In addition, in the above-mentioned embodiment, an example of a structure in which the filter coefficient calculation unit 22 uses the object signal 30 to calculate the filter coefficient 33 while using the similarity index 32 is shown, but the present invention is not limited to this. For example, the filter coefficient calculation unit may also be configured to obtain the filter coefficient 33 based on the similarity index 32 without using the object signal 30. However, in the case of a structure in which the filter coefficient calculation unit obtains the filter coefficient 33 based on the similarity index 32 without using the object signal 30, the certainty of the filter coefficient 33 decreases. Therefore, the filter coefficient calculation unit 22 is preferably configured to calculate the filter coefficient 33 using the object signal 30 while using the similarity index 32.

[0208] In addition, in the above-mentioned embodiment, an example of a structure in which the auxiliary signal 35 includes the low-noise signal 34 is shown, but the present invention is not limited thereto. For example, the auxiliary signal 35 may not include the low-noise signal 34. However, in the case of a structure in which the auxiliary signal 35 does not include the low-noise signal 34, it is difficult to obtain the auxiliary signal 35 corresponding to the object signal 30, which is a signal that is difficult to obtain using a device different from the device that obtains the object signal 30, an acquisition condition different from the condition when the object signal 30 is obtained, or a processing condition different from the object signal 30. Therefore, the auxiliary signal 35 preferably includes the low-noise signal 34.

[0209] In addition, in the above-mentioned embodiment, an example of a structure in which the noise reduction unit 2 synthesizes the object signal 30 and the low noise signal 34 in a weighted state to obtain a low noise synthesis signal 36 to adjust the noise reduction effect is shown, but the present invention is not limited to this. For example, the noise adjustment unit may not be configured to obtain the low noise synthesis signal 36. However, in the case where the noise adjustment unit is not configured to obtain the low noise synthesis signal 36, in order to adjust the noise reduction effect, it is necessary to change the parameters and re-perform the noise reduction process. In this case, the time required for the noise reduction process increases. Therefore, the noise reduction unit 2 is preferably configured to obtain the low noise synthesis signal 36.

[0210] In addition, in the above-mentioned embodiment, an example of a structure in which the synthesis processing unit 24 synthesizes the object signal 30 and the low noise signal 34 in a state where they are weighted separately to obtain the low noise synthesis signal 36 is shown, but the present invention is not limited to this. For example, the synthesis processing unit may also be configured to synthesize the object signal 30 and the low noise signal 34 without weighting them separately to obtain the low noise synthesis signal 36. However, in the case of a structure in which the synthesis processing unit synthesizes the object signal 30 and the low noise signal 34 without weighting them separately to obtain the low noise synthesis signal 36, the degree of freedom of adjustment of the noise reduction effect decreases. Therefore, the synthesis processing unit 24 is preferably configured to synthesize the object signal 30 and the low noise signal 34 in a state where they are weighted separately to obtain the low noise synthesis signal 36.

[0211] In addition, in the above-mentioned embodiment, an example of a structure in which the synthesis processing unit 24 uses the noise intensity signal 31 to obtain the low noise synthesis signal 36 while using the object signal 30 and the low noise signal 34, is shown, but the present invention is not limited to this. For example, the synthesis processing unit may be configured to obtain the low noise synthesis signal 36 using the object signal 30 and the low noise signal 34 without using the noise intensity signal 31. However, in the case of a structure in which the synthesis processing unit uses the object signal 30 and the low noise signal 34 to obtain the low noise synthesis signal 36 without using the noise intensity signal 31, it is impossible to synthesize the object signal 30 and the low noise signal 34 in a state in which the distribution of noise is reflected based on the noise intensity signal 31. Therefore, it is difficult to adjust the noise reduction effect of the low noise synthesis signal 36 in detail. Therefore, the synthesis processing unit 24 is preferably configured to obtain the low noise synthesis signal 36 by using the noise intensity signal 31 while using the object signal 30 and the low noise signal 34.

[0212] In addition, in the above-mentioned embodiment, an example of a structure in which the synthesis processing unit 24 uses at least one of the similarity index 32 and the filter coefficient 33 to obtain the low noise synthesis signal 36 while using the object signal 30 and the low noise signal 34, but the present invention is not limited to this. For example, the synthesis processing unit may be configured to obtain the low noise synthesis signal 36 by using the object signal 30 and the low noise signal 34 without using the similarity index 32 and the filter coefficient 33. However, in the case where the synthesis processing unit is configured to obtain the low noise synthesis signal 36 by using the object signal 30 and the low noise signal 34 without using the similarity index 32 and the filter coefficient 33, the accuracy of the low noise synthesis signal 36 decreases. Therefore, the synthesis processing unit 24 is preferably configured to obtain the low noise synthesis signal 36 by using at least one of the similarity index 32 and the filter coefficient 33 while using the object signal 30 and the low noise signal 34.

[0213] In addition, in the above embodiment, an example of a structure in which the low noise signal 34 is used as the auxiliary signal 35 is shown, but the present invention is not limited to this. For example, the low noise synthesis signal 36 may be used as the auxiliary signal. In addition, the low noise synthesis signal 36 may be used as the object signal 30 to repeat the noise reduction process.

[0214] In addition, in the above-mentioned embodiment, an example of a structure in which the filter window (second filter window) of the standard deviation filter (second filter) is larger than the filter window (first filter window) of the smoothing filter (first filter) is shown, but the present invention is not limited thereto. The size of the first filter window and the size of the second filter window can be set according to the object signal 30. That is, the size of the second filter window can also be smaller than the size of the first filter window, and can also be equal to the size of the first filter window.

[0215] In the above embodiment, the object signal 30 shown in Table 52 is only an example, and signals other than those shown in Table 52 may also be included in the object signal 30. In addition, the auxiliary signal 35 shown in Table 53 is only an example, and signals other than those shown in Table 53 may also be included in the auxiliary signal 35.

[0216] In addition, in the above-mentioned embodiment, for the convenience of explanation, the noise reduction processing performed by the noise reduction unit 2 is described using a flow chart of a process-driven type in which processing is performed sequentially according to the processing flow, but the present invention is not limited to this. In the present invention, the noise reduction processing performed by the noise reduction unit 2 can also be performed by an event-driven (event-driven) processing in which processing is performed in units of events. In this case, it can be performed in a completely event-driven manner or in a combination of event-driven and flow-driven.

[0217] [Way]

[0218] Those skilled in the art will appreciate that the above-described exemplary embodiments are specific examples of the following aspects.

[0219] (Item 1)

[0220] A noise reduction processing method for an object signal, the noise reduction processing method comprising the following steps:

[0221] estimating a noise intensity signal based on the object signal, wherein the noise intensity signal represents an intensity distribution of noise of the object signal;

[0222] Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point;

[0223] Calculating a filter coefficient for noise reduction filtering based on the similarity index; and

[0224] applying the noise reduction filter to the object signal based on the filter coefficients to obtain a low noise signal,

[0225] In the step of estimating the noise intensity signal, the following processing is performed: applying a first filter having a first filter window to the object signal to obtain a low-frequency signal; obtaining a high-frequency signal based on the object signal and the low-frequency signal; and applying a second filter having a second filter window independent of the first filter window to the high-frequency signal to estimate the noise intensity signal.

[0226] (Item 2)

[0227] The noise reduction processing method according to item 1, wherein:

[0228] The first filtering includes smoothing filtering,

[0229] The second filtering comprises a standard deviation filtering,

[0230] In the step of estimating the noise intensity signal, the standard deviation filter different from the smoothing filter is applied to the high frequency signal to estimate the noise intensity signal.

[0231] (Item 3)

[0232] The noise reduction processing method according to item 1 or 2, wherein:

[0233] The object signal is an image signal of two dimensions or more.

[0234] The noise reduction processing method further comprises the following steps: acquiring a noise model of the noise distribution based on a noise distribution of any one of noise caused by an acquisition principle when acquiring the image signal and noise caused by a device for acquiring the image signal,

[0235] In the step of estimating the noise intensity signal, the noise intensity signal is estimated by multiplying the high frequency signal by the noise model after applying the second filtering to the high frequency signal.

[0236] (Item 4)

[0237] The noise reduction method according to any one of items 1 to 3, wherein:

[0238] A series of processes of estimating the noise intensity signal, calculating the similarity index, calculating the filter coefficient, and acquiring the low noise signal are repeatedly performed using the low noise signal as the target signal.

[0239] (Item 5)

[0240] The noise reduction method according to any one of items 1 to 4, wherein:

[0241] In the step of acquiring the low noise signal, the noise intensity signal is used while using the filter coefficient to acquire the low noise signal.

[0242] (Item 6)

[0243] The noise reduction method according to any one of items 1 to 5, wherein:

[0244] In the step of calculating the filter coefficient, the filter coefficient is calculated using the object signal while using the similarity index.

[0245] (Item 7)

[0246] The noise reduction method according to any one of items 1 to 6, wherein:

[0247] The method further includes the following steps: acquiring an auxiliary signal for the same object as the object from which the object signal is acquired, wherein the auxiliary signal is acquired using a device different from the device from which the object signal is acquired, an acquisition condition different from the condition when the object signal is acquired, or a processing condition different from that of the object signal;

[0248] The auxiliary signal is used for processing in at least any one of the following cases: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal.

[0249] (Item 8)

[0250] The noise reduction processing method according to item 7, wherein:

[0251] The auxiliary signal includes the low noise signal,

[0252] In at least any one of the following cases, when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal, the low noise signal is used as the auxiliary signal for processing.

[0253] (Item 9)

[0254] The noise reduction method according to any one of items 1 to 8, wherein:

[0255] In at least any one of the following cases, processing based on a reference data set is performed: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal.

[0256] (Item 10)

[0257] The noise reduction method according to any one of items 1 to 9, wherein:

[0258] The method further includes the step of synthesizing the object signal and the low noise signal in a weighted state to obtain a low noise synthesis signal, so as to adjust the noise reduction effect.

[0259] (Item 11)

[0260] A noise reduction processing device for an object signal, the noise reduction processing device comprising:

[0261] an object signal acquisition unit that acquires the object signal; and

[0262] a noise reduction unit for reducing noise of the target signal,

[0263] The noise reduction unit is configured to perform the following processing:

[0264] Applying a first filter having a first filter window to the object signal to obtain a low-frequency signal, acquiring a high-frequency signal based on the object signal and the low-frequency signal, and applying a second filter having a second filter window independent of the first filter window to the high-frequency signal to estimate a noise intensity signal;

[0265] Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point;

[0266] Calculating a filter coefficient for noise reduction filtering based on the similarity index; and

[0267] Based on the filter coefficients, the noise reduction filter is applied to the object signal to obtain a low noise signal.

[0268] (Item 12)

[0269] A noise reduction processing method for an object signal, the noise reduction processing method comprising the following steps:

[0270] estimating a noise intensity signal representing an intensity distribution of noise of the object signal based on the object signal;

[0271] Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point;

[0272] Calculating a filter coefficient for noise reduction filtering based on the similarity index;

[0273] Applying the noise reduction filter to the object signal based on the filter coefficient to obtain a low noise signal; and

[0274] The object signal and the low noise signal are synthesized to obtain a low noise synthesis signal to adjust the noise reduction effect.

[0275] (Item 13)

[0276] The noise reduction processing method according to item 12, wherein:

[0277] In the step of acquiring the low noise composite signal, the object signal and the low noise signal are composited in a state where they are weighted, thereby acquiring the low noise composite signal.

[0278] (Item 14)

[0279] The noise reduction processing method according to item 12 or 13, wherein:

[0280] In the step of acquiring the low noise synthesis signal, the low noise signal is acquired by using the noise intensity signal while using the object signal and the low noise signal.

[0281] (Item 15)

[0282] The noise reduction method according to any one of items 12 to 14, wherein:

[0283] In the step of acquiring the low noise synthetic signal, the low noise synthetic signal is acquired by using at least either the similarity index and the filter coefficient while using the object signal and the low noise signal.

[0284] (Item 16)

[0285] The noise reduction method according to any one of items 12 to 14, wherein:

[0286] The method further includes the following steps: acquiring an auxiliary signal for the same object as the object from which the object signal is acquired, wherein the auxiliary signal is acquired using a device different from the device from which the object signal is acquired, an acquisition condition different from the condition when the object signal is acquired, or a processing condition different from that of the object signal;

[0287] The auxiliary signal is used for processing in at least any one of the following cases: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, when acquiring the low noise signal in the step of acquiring the low noise signal, and when acquiring the low noise synthetic signal in the step of acquiring the low noise synthetic signal.

[0288] (Item 17)

[0289] The noise reduction method according to any one of items 12 to 14, wherein:

[0290] In at least any one of the following cases, processing based on a reference data set is performed: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, when acquiring the low noise signal in the step of acquiring the low noise signal, and when acquiring the low noise synthetic signal in the step of acquiring the low noise synthetic signal.

[0291] (Item 18)

[0292] A noise reduction processing device for an object signal, the noise reduction processing device comprising:

[0293] an object signal acquisition unit that acquires the object signal; and

[0294] a noise reduction unit for reducing noise of the target signal,

[0295] The noise reduction unit is configured to perform the following processing:

[0296] estimating a noise intensity signal representing an intensity distribution of noise of the object signal based on the object signal;

[0297] Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point;

[0298] Calculating a filter coefficient for noise reduction filtering based on the similarity index;

[0299] Applying the noise reduction filter to the object signal based on the filter coefficient to obtain a low noise signal; and

[0300] The object signal and the low noise signal are synthesized to obtain a low noise synthesis signal with an adjusted noise reduction effect, so as to adjust the noise reduction effect.

Claims

1. A method for reducing noise of an object signal, the method comprising the following steps: estimating a noise intensity signal, wherein the noise intensity signal represents an intensity distribution of noise of the object signal; Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; Calculating a filter coefficient for noise reduction filtering based on the similarity index; as well as applying the noise reduction filter to the object signal based on the filter coefficient to obtain a low noise signal, In the step of estimating the noise intensity signal, the following processing is performed: Applying a first filter with a first filter window to the object signal to obtain a low-frequency signal; acquiring a high-frequency signal based on the object signal and the low-frequency signal; as well as A second filter having a second filter window independent of the first filter window is applied to the high frequency signal to estimate the noise intensity signal.

2. The noise reduction processing method according to claim 1, wherein: The first filtering includes smoothing filtering, The second filtering comprises a standard deviation filtering, In the step of estimating the noise intensity signal, the standard deviation filter different from the smoothing filter is applied to the high frequency signal to estimate the noise intensity signal.

3. The noise reduction processing method according to claim 2, wherein: The object signal is an image signal of two dimensions or more. The noise reduction processing method further comprises the following steps: acquiring a noise model of the noise distribution based on a noise distribution of any one of noise caused by an acquisition principle when acquiring the image signal and noise caused by a device for acquiring the image signal, In the step of estimating the noise intensity signal, the noise intensity signal is estimated by multiplying the high frequency signal by the noise model after applying the second filtering to the high frequency signal.

4. The noise reduction processing method according to claim 3, wherein: A series of processes of estimating the noise intensity signal, calculating the similarity index, calculating the filter coefficient, and acquiring the low noise signal are repeatedly performed using the low noise signal as the target signal.

5. The noise reduction processing method according to claim 1, wherein: In the step of acquiring the low noise signal, the noise intensity signal is used while using the filter coefficient to acquire the low noise signal.

6. The noise reduction processing method according to claim 1, wherein: In the step of calculating the filter coefficient, the filter coefficient is calculated by using the object signal while using the similarity index.

7. The noise reduction processing method according to claim 1, wherein: The following steps are also included: acquiring an auxiliary signal for the same object as the object for which the object signal is acquired, the auxiliary signal being a signal acquired using a device different from the device for acquiring the object signal, an acquisition condition different from the condition when the object signal is acquired, or a processing condition different from that of the object signal, The auxiliary signal is used for processing in at least any one of the following cases: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal.

8. The noise reduction processing method according to claim 7, wherein: The auxiliary signal includes the low noise signal, In at least any one of the following cases, when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal, the low noise signal is used as the auxiliary signal for processing.

9. The noise reduction processing method according to claim 1, wherein: In at least any one of the following cases, processing based on a reference data set is performed: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, and when acquiring the low noise signal in the step of acquiring the low noise signal.

10. The noise reduction processing method according to claim 1, wherein: The following steps are also included: The target signal and the low noise signal are synthesized in a weighted state to obtain a low noise synthesis signal, so as to adjust the noise reduction effect.

11. A noise reduction processing device for an object signal, the noise reduction processing device comprising: an object signal acquisition unit that acquires the object signal; and a noise reduction unit for reducing noise of the target signal, in, The noise reduction unit is configured to perform the following processing: Applying a first filter having a first filter window to the object signal to obtain a low-frequency signal, acquiring a high-frequency signal based on the object signal and the low-frequency signal, and applying a second filter having a second filter window independent of the first filter window to the high-frequency signal to estimate a noise intensity signal; Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; Calculating a filter coefficient for noise reduction filtering based on the similarity index; as well as Based on the filter coefficients, the noise reduction filter is applied to the object signal to obtain a low noise signal.

12. A method for reducing noise of an object signal, the method comprising the following steps: estimating a noise intensity signal representing an intensity distribution of noise of the object signal based on the object signal; Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; Calculating a filter coefficient for noise reduction filtering based on the similarity index; Based on the filter coefficients, applying the noise reduction filter to the object signal to obtain a low noise signal; as well as The object signal and the low noise signal are synthesized to obtain a low noise synthesis signal to adjust the noise reduction effect.

13. The noise reduction processing method according to claim 12, wherein: In the step of acquiring the low noise composite signal, the object signal and the low noise signal are composited in a state where they are weighted, thereby acquiring the low noise composite signal.

14. The noise reduction processing method according to claim 12, wherein: In the step of acquiring the low noise synthesis signal, the low noise signal is acquired by using the noise intensity signal while using the object signal and the low noise signal.

15. The noise reduction processing method according to claim 12, wherein: In the step of acquiring the low noise synthetic signal, the low noise synthetic signal is acquired by using at least either the similarity index and the filter coefficient while using the object signal and the low noise signal.

16. The noise reduction processing method according to claim 12, wherein: The following steps are also included: acquiring an auxiliary signal for the same object as the object for which the object signal is acquired, the auxiliary signal being a signal acquired using a device different from the device for acquiring the object signal, an acquisition condition different from the condition when the object signal is acquired, or a processing condition different from that of the object signal, The auxiliary signal is used for processing in at least any one of the following cases: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, when acquiring the low noise signal in the step of acquiring the low noise signal, and when acquiring the low noise synthetic signal in the step of acquiring the low noise synthetic signal.

17. The noise reduction processing method according to claim 12, wherein: In at least any one of the following cases, processing based on a reference data set is performed: when estimating the noise intensity signal in the step of estimating the noise intensity signal, when calculating the similarity index in the step of calculating the similarity index, when calculating the filter coefficient in the step of calculating the filter coefficient, when acquiring the low noise signal in the step of acquiring the low noise signal, and when acquiring the low noise synthetic signal in the step of acquiring the low noise synthetic signal.

18. A noise reduction processing device for an object signal, the noise reduction processing device comprising: an object signal acquisition unit that acquires the object signal; and a noise reduction unit for reducing noise of the target signal, in, The noise reduction unit is configured to perform the following processing: estimating a noise intensity signal representing an intensity distribution of noise of the object signal based on the object signal; Calculating, based on the object signal and the noise intensity signal, a similarity index between a prescribed signal acquired at a first signal point and a signal acquired at a second signal point adjacent to the first signal point; Calculating a filter coefficient for noise reduction filtering based on the similarity index; Based on the filter coefficients, applying the noise reduction filter to the object signal to obtain a low noise signal; as well as The object signal and the low noise signal are synthesized to obtain a low noise synthesis signal to adjust the noise reduction effect.

Images