Image processing apparatus, image processing method, image processing program, and recording medium

Abstract

In the noise estimation step S2, a noise image included in the object image including a linear noise extending along the first direction is estimated. At this time, using an evaluation function including a first term indicating a difference between a result of performing a differentiation process in the second direction and a low-frequency component extraction process in the first direction on the object image and a result of performing a differentiation process in the second direction and a low-frequency component extraction process in the first direction on the noise image, a noise image in which a value of the evaluation function becomes minimum is found. In the noise reduction step S3, from the object image and the noise image, a noise-reduced image is generated from the object image. Thus, an image processing method or the like capable of processing an object image including a linear noise extending along one direction and generating a noise-reduced image maintaining quantitativeness of the object image is realized.

Description

Technical Field

[0001] This invention relates to an image processing apparatus, an image processing method, an image processing program, and a recording medium. Background Technology

[0002] Several techniques are known for generating phase differential images from one or more interferometric images acquired using a device employing differential interferometry, and then obtaining phase images based on these phase differential images. In these techniques, phase images can be generated by integrating or deconvolving the phase differential images. These techniques are suitable, for example, for acquiring phase images of cells.

[0003] The resulting phase images sometimes contain linear noise (linear artifacts) extending in one direction, often consisting of multiple parallel linear noise lines. Images containing linear noise extending in one direction exist not only in phase images obtained using apparatus employing differential interferometry microscopy, but also in other types of images.

[0004] Non-Patent Document 1 describes a technique for processing an object image containing linear noise extending in one direction to generate an image with reduced noise.

[0005] Existing technical documents

[0006] [Non-patent literature]

[0007] [Non-patent document 1] MRARNISON et al., "Using the Hilbert transform for 3Dvisualization of differential interference contrast microscope images", Journal of Microscopy, Vol. 199Pt1, pp. 79-84 (2000) Summary of the Invention

[0008] The problem that the invention aims to solve

[0009] However, in the technology described in Non-Patent Document 1, although it is possible to generate an image with reduced noise from an object image, even if the object image has quantity, it will lose quantity in the image after noise reduction processing.

[0010] The purpose of this embodiment is to provide an image processing apparatus, image processing method, image processing program, and recording medium capable of processing an object image containing linear noise extending in one direction to generate a noise-reduced image that maintains the quantitative characteristics of the object image.

[0011] Methods for solving problems

[0012] The embodiment is an image processing apparatus. The image processing apparatus is a device for generating a noise-reduced image by processing an object image containing linear noise extending along a first direction. It includes: a noise estimation unit that estimates a noise image contained in the object image; and a noise reduction unit that generates a noise-reduced image from the object image based on the object image and the noise image. The noise estimation unit uses an evaluation function to determine the noise image whose value is minimized. This evaluation function includes a first term representing the difference between the result of differential processing of the object image in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction, and the result of differential processing of the noise image in the second direction and low-frequency component extraction processing in the first direction.

[0013] The implementation method is an image processing method. The image processing method is a method for generating a noise-reduced image by processing an object image containing linear noise extending along a first direction, comprising: a noise estimation step, estimating the noise image contained in the object image; and a noise reduction step, generating a noise-reduced image from the object image based on the object image and the noise image. In the noise estimation step, an evaluation function containing a first term is used to find the noise image whose value is minimized. This first term represents the difference between the result of differential processing in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction on the object image, and the result of differential processing in the second direction and low-frequency component extraction processing in the first direction on the noise image.

[0014] The implementation method is an image processing program. An image processing program is a program used to cause a computer to perform the steps of the image processing method described above.

[0015] The implementation method is a recording medium. The recording medium is a computer-readable medium that records the above-described image processing program.

[0016] Invention Effects

[0017] The image processing apparatus, image processing method, image processing program, and recording medium according to the embodiments are capable of processing an object image containing linear noise extending in one direction to generate a noise-reduced image that maintains the quantitative characteristics of the object image. Attached Figure Description

[0018] Figure 1 This is a diagram showing the structure of an image processing apparatus 1 according to one embodiment.

[0019] Figure 2 This is a flowchart illustrating one implementation of an image processing method.

[0020] Figure 3 (a) is a graph representing the phase differential image. Figure 3 (b) is a diagram representing the phase image.

[0021] Figure 4 (a) is a graph representing a noisy image. Figure 4 (b) is a graph representing the image after noise reduction processing.

[0022] Figure 5 (a) is a graph representing the image of object x. Figure 5 (b) indicates that... Figure 5 The image Dx is a graph of the object image x that has undergone differential processing in the second direction D, and... Figure 5 (c) indicates that... Figure 5 The image L1Dx is the result of low-frequency component extraction processing L1 in the first direction of image Dx (b).

[0023] Figure 6 (a) is a graph representing the noise image y during the calculation process. Figure 6 (b) indicates that... Figure 6 The image Dy is the result of the second-direction differential processing D on the noisy image y of (a). Figure 6 (c) indicates that... Figure 6 The image of (b) Dy is the result of low-frequency component extraction processing L1 in the first direction, and the image L1Dy is a graph.

[0024] Figure 7 It means to Figure 6 The image L2y is the result of high-frequency component extraction processing L2 in the first direction of the noise image y during the calculation process of (a).

[0025] Figure 8 (a) is a graph showing the noise image obtained when using the evaluation function E(x,y) of equation (4). Figure 8 (b) is a graph representing the image after noise reduction processing obtained when using the evaluation function E(x,y) of equation (4).

[0026] Figure 9 (a) is a graph representing the noise image obtained using the evaluation function E(x,y) in equation (5). Figure 9(b) is a graph showing the image after noise reduction processing obtained by using the evaluation function E(x,y) in equation (5).

[0027] Figure 10 (a) is a graph representing the noise image obtained using the evaluation function E(x,y) in equation (6). Figure 10 (b) is a graph showing the image after noise reduction processing obtained by using the evaluation function E(x,y) in equation (6).

[0028] Figure 11 (a) is a graph showing the noise image obtained when using the evaluation function E(x,y) in equation (7). Figure 11 (b) is a graph showing the image after noise reduction processing obtained by using the evaluation function E(x,y) in equation (7).

[0029] Figure 12 (a) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.335. Figure 12 (b) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.342.

[0030] Figure 13 (a) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.349. Figure 13 (b) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.362. Detailed Implementation

[0031] Hereinafter, embodiments of the image processing apparatus, image processing method, image processing program, and recording medium will be described in detail with reference to the accompanying drawings. Furthermore, in the description of the drawings, the same reference numerals are used to denote the same elements, and repeated descriptions are omitted. The present invention is not limited to these illustrations.

[0032] Figure 1 This diagram illustrates the structure of the image processing apparatus 1 according to this embodiment. The image processing apparatus 1 includes a control unit 10, a target image generation unit 11, a noise calculation unit 12, a noise reduction unit 13, an input unit 14, a storage unit 15, and a display unit 16. The image processing apparatus 1 may be a computer. The control unit 10 controls the operation of each of the target image generation unit 11, the noise calculation unit 12, the noise reduction unit 13, the input unit 14, the storage unit 15, and the display unit 16, and includes a CPU.

[0033] The object image generation unit 11, noise calculation unit 12, and noise reduction unit 13 perform image processing, including processing devices such as CPU, DSP, or FPGA. The input unit 14 inputs the data of the image to be processed, and the image processing conditions are input via keyboard or mouse.

[0034] The storage unit 15 stores various image data, including hard disk drives, flash memory, RAM, and ROM. Furthermore, the object image generation unit 11, noise calculation unit 12, noise reduction unit 13, and storage unit 15 can also be configured using cloud computing. The display unit 16 displays images to be processed, images in progress, and processed images, and may include, for example, a liquid crystal display (LCD).

[0035] The storage unit 15 also stores programs for instructing the object image generation unit 11, noise calculation unit 12, and noise reduction unit 13 to perform each step of image processing. This image processing program can be stored in the storage unit 15 during the manufacture or at the time of shipment of the image processing apparatus 1, or it can be a program obtained after shipment via a communication line and stored in the storage unit 15. Alternatively, a program recorded on a computer-readable recording medium 2 can also be stored in the storage unit 15. The recording medium 2 can be any medium such as a floppy disk, CD-ROM, DVD-ROM, BD-ROM, or USB memory.

[0036] Figure 2 This is a flowchart illustrating the image processing method of this embodiment. The image processing method of this embodiment includes an object image generation step S1, a noise estimation step S2, and a noise reduction step S3.

[0037] The object image generation step S1 is the processing performed by the object image generation unit 11. The noise estimation step S2 is the processing performed by the noise estimation unit 12. The noise reduction step S3 is the processing performed by the noise reduction unit 13. As an example, the case where a phase image is generated from the phase differential image in the object image generation step S1 will be explained.

[0038] In the object image generation step S1, the object image generation unit 11 generates a phase image by performing integration or deconvolution processing on the phase differential image. The phase image generated here is the object image that becomes the target of noise reduction processing.

[0039] In noise estimation step S2, the noise estimation unit 12 estimates the noise image contained in the object image (phase image). In noise reduction step S3, the noise reduction unit 13 generates an image (noise-reduced image) based on the object image and the noise image, with the noise reduced from the object image. Specifically, the noise-reduced image can be generated by subtracting the noise image from the object image. Hereinafter, specific image examples will be used to explain each step in detail.

[0040] Figure 3 (a) is a diagram showing the phase differential image. This phase differential image was generated from an interference image obtained using a device employing a differential interference microscope. The shearing direction in the differential interference microscope is left-right in this diagram. In addition to the several cells being observed, the phase differential image also shows a background region around them with approximately the same phase (i.e., a phase differential of approximately 0).

[0041] Figure 3 (b) is a diagram representing the phase image. This phase image is generated by the object image generation unit 11 in the object image generation step S1. Figure 3 The phase differential image of (a) is generated by integral processing. Specifically, the phase image to be obtained is set as x, the differential processing of the first direction of the phase image x (the left and right direction and the shearing direction in the figure) is set as A, the phase differential image is set as b, and the positive constant is set as λ. The phase image x can be obtained by solving the optimization problem expressed by the following equation (1).

[0042] [Formula 1]

[0043]

[0044] In the above, λ is set to 10, for example. -5 ~10 -2 The range of values. Under the constraint that the phase image x is greater than or equal to 0, the phase image x can be obtained as the image with the smallest difference between the result Ax of the first-direction differential processing A on the phase image x and the phase differential image b. As shown in the figure, the phase image x contains linear noise (linear artifacts) extending along the first direction (left and right direction in the figure).

[0045] Figure 4 (a) is a graph representing a noise image. This noise image is calculated by the noise calculation unit 12 in noise calculation step S2 and includes... Figure 3 The image in (b) is the phase image (the object image to which the noise reduction processing is performed). Details of the noise image calculation processing in noise calculation step S2 will be described later.

[0046] Figure 4 (b) is a diagram showing the image after noise reduction processing. This image after noise reduction processing is obtained from the noise reduction unit 13 in noise reduction step S3. Figure 3 (b) Object image (phase image) minus Figure 4 The image was generated from the noisy image in (a). As shown in the figure, the noise reduction process reduces the noise contained in the object image while maintaining the quantity of the object image.

[0047] Next, the processing of the noise image contained in the object image in the noise estimation step S2 will be explained in detail. Let the object image be x, and let the noise image contained in the object image x be y.

[0048] The low-frequency component extraction process for the first direction of the image (left-right direction in the figure) is set as L1, and the high-frequency component extraction process for the first direction of the image is set as L2. The differential processing for the second direction of the image (up-down direction orthogonal to the first direction in the figure) is set as D, and the processing for extracting the background region in the image is set as M. In addition, positive constants are set as λ and μ.

[0049] In the above, for example, λ is set to 10. -3 ~10 -1 The range of values ​​for μ is set to 10. -2 The value is in the range of ~1. For the evaluation function E(x,y) expressed by the following equation (3), the noisy image y is calculated by solving the optimization problem expressed by the following equation (2).

[0050] [Equation 2]

[0051]

[0052] [Formula 3]

[0053]

[0054] The first term of the evaluation function E(x,y) expressed by equation (3) above represents the difference between L1Dx, the result of performing differential processing D in the second direction and low-frequency component extraction processing L1 in the first direction on the object image x, and L1Dy, the result of performing differential processing D in the second direction and low-frequency component extraction processing L1 in the first direction on the noisy image y. The order of differential processing D and low-frequency component extraction processing L1 for each image is arbitrary. Alternatively, differential processing D and low-frequency component extraction processing L1 can also be performed on the difference between the object image x and the noisy image y.

[0055] The linear noise (linear artifacts) contained in the object image extends along the first direction (left-right direction) of the object image, so the noise variation is greater along the second direction, and the spatial frequency of the noise is lower along the first direction. Therefore, the noisy image y can be obtained as the image that minimizes the first term of the evaluation function E(x,y).

[0056] Figure 5 and Figure 6 This is a graph illustrating an example of the first term of the evaluation function E(x,y). Figure 5 (a) is a graph representing the image of object x. Figure 5 (b) indicates that... Figure 5The image Dx is a graph of the result of differentiating the object image x in the second direction D. In this image Dx, the noise becomes more apparent, but information about the observed object (high-frequency components) is also present.

[0057] Figure 5 (c) indicates that... Figure 5 The image L1Dx is the result of low-frequency component extraction processing L1 in the first direction of image Dx (b). Low-frequency components are extracted from this image L1Dx.

[0058] Figure 6 (a) is a graph representing the noise image y during the calculation process. Figure 6 (b) indicates that... Figure 6 The image Dy is the result of the second-direction differential processing D on the noisy image y in (a). Ideally, there is no information about the observed object (high-frequency components) in the noisy image at the end of the extrapolation process, but information about the observed object (high-frequency components) is present in the image Dy during the extrapolation process.

[0059] Figure 6 (c) indicates that... Figure 6 Image L1Dy is the result of low-frequency component extraction processing L1 in the first direction of image Dy (b). Since there is ideally no information about the observed object (high-frequency components) in the noisy image at the end of the extrapolation process, this image L1Dy is approximately the same as image Dy.

[0060] The first term of the evaluation function E(x,y) represents Figure 5 Image L1Dx of (c) and Figure 6 The difference between the image L1Dy and the image (c) is calculated. The image y with the smallest difference is found.

[0061] The second term of the evaluation function E(x,y) expressed by equation (3) above represents the difference between the result Mx of background region extraction processing M on the object image x and the result My of background region extraction processing M on the noisy image y. Alternatively, background region extraction processing M can be performed on the difference between the object image x and the noisy image y. In both the object image x and the noisy image y, there is no information about the observed object in the background region; only noise exists. Therefore, the noisy image y can be obtained as the image whose second term of the evaluation function E(x,y) is minimized.

[0062] The third term of the evaluation function E(x,y) expressed by equation (3) above represents the result L2y of the high-frequency component extraction processing L2 performed on the noisy image y in the first direction. Along the first direction, the spatial frequency of the noise is lower than the spatial frequency of the information of the observed object, and it does not have a high component. Therefore, the noisy image y can be obtained as the image whose third term of the evaluation function E(x,y) is minimized. Figure 7 It means to Figure 6 The image L2y is the result of high-frequency component extraction processing L2 in the first direction of the noise image y during the calculation process of (a).

[0063] Generally, it is impossible to find a noisy image y that simultaneously minimizes the first, second, and third terms of the evaluation function E(x,y) expressed by equation (3) above. Therefore, as in equation (3) above, constants λ and μ are used to minimize the evaluation function E(x,y) expressed by the linear sum of the first, second, and third terms, and the optimization problem expressed by equation (2) above is solved, thereby deriving the noisy image y.

[0064] In addition, the evaluation function E(x,y) needs to include the first term in equation (3) above, but it may not include either or both of the second and third terms in equation (3) above.

[0065] That is, the evaluation function E(x,y) can also be expressed by setting the value of the constant λ or μ to 0 in the above equation (3), and can be represented by any one of the following equations (4), (5), and (6).

[0066] [Formula 4]

[0067]

[0068] [Formula 5]

[0069]

[0070] [Formula 6]

[0071]

[0072] Next, the noise image and the image after noise reduction processing obtained using the evaluation functions E(x,y) of equations (3) to (6) above will be explained. In any case, Figure 3 The phase image shown in (b) is set as the object image x to be the object of noise reduction processing.

[0073] Using the evaluation function E(x,y) from equation (3) above, in noise estimation step S2, we obtain... Figure 4 The noisy image y shown in (a) is obtained in noise reduction step S3. Figure 4 Image (b) shows the image after noise reduction processing. Here, λ = 1 × 10⁻⁶. -2 μ=1×10 -1 The noise reduction process significantly reduces the noise contained in the object image while maintaining the quantitative nature of the object image.

[0074] Using the evaluation function E(x,y) from equation (4) above, in noise estimation step S2, we obtain... Figure 8 The noisy image y shown in (a) is obtained in noise reduction step S3. Figure 8 Image (b) shows the noise-reduced image. Here, μ = 1 × 10⁻⁶. -1 In this case, the resulting noisy image y contains information about the observed object. Therefore, although the noise is reduced after noise reduction processing, some information about the observed object is lost, but the quantitative nature of the object image is maintained relatively well.

[0075] Using the evaluation function E(x,y) from equation (5) above, in noise estimation step S2, we obtain... Figure 9 The noisy image y shown in (a) is obtained in noise reduction step S3. Figure 9 Image (b) shows the image after noise reduction processing. Here, λ = 1 × 10⁻⁶. -2 In this case, the noisy image y exhibits computational errors in the portion indicated by the arrows in the figure. Therefore, although the noise reduction is not complete, the image after noise reduction processing maintains the quantitative characteristics of the object image relatively well.

[0076] Using the evaluation function E(x,y) from equation (6) above, in noise estimation step S2, we obtain... Figure 10 The noisy image y shown in (a) is obtained in noise reduction step S3. Figure 10 The image shown in (b) is the image after noise reduction processing. In this case, the resulting noisy image y contains information about the observed object. Therefore, although the noise is reduced and some information about the observed object is lost, the quantitative nature of the object image is maintained relatively well after noise reduction processing.

[0077] As a comparative example, when using the evaluation function E(x,y) of equation (7) below, which does not contain the first term in equation (3) above, in the noise estimation step S2, we obtain Figure 11 The noisy image y shown in (a) is obtained in noise reduction step S3. Figure 11 Image (b) shows the image after noise reduction processing. Here, λ = 1 × 10⁻⁶. -1When μ = 1, the noise in the resulting noisy image y cannot be inferred in the region where the observed object exists. Therefore, the noise reduction image does not reduce noise in the region where the observed object exists.

[0078] [Formula 7]

[0079]

[0080] Thus, by solving the optimization problem represented by equation (2) above, any one of the evaluation functions E(x,y) in equations (3) to (6) above is minimized, enabling high-precision estimation of the noisy image y. Moreover, the noise-reduced image obtained based on the object image x and the noisy image y reduces the noise contained in the object image while maintaining the quantifiability of the object image. Furthermore, the use of the evaluation function E(x,y) in equation (3) above is the most preferred option.

[0081] Figure 12 and Figure 13 This is a graph showing the noise-reduced image obtained by the image processing method of this embodiment when the refractive index of the solution containing the cells being observed is used as each value. The refractive index of the solution is adjusted by adjusting the concentration of BSA (bovine serum albumin) contained in the solution. The refractive index of the solution is measured using an Abbe refractometer from ATAGO Corporation. The evaluation function E(x,y) of equation (3) above is used. Here, λ = 1 × 10 -2 μ=1×10 -1 .

[0082] Figure 12 (a) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.335. Figure 12 (b) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.342. Figure 13 (a) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.349. Figure 13 (b) is a graph showing the image after noise reduction when the refractive index of the solution is set to approximately 1.362.

[0083] As these figures show, regardless of the solution's refractive index, the noise-reduced images all reduce the noise contained in the object image while maintaining its quantifiability. As the solution's refractive index increases, the phase difference between the cells and the solution decreases.

[0084] like Figure 13As shown in (b), even when the phase difference between the cell and the solution is small, the resulting noise-reduced image sufficiently reduces the noise contained in the object image. The refractive index of the solution when the phase difference between the cell and the solution is zero can be accurately determined as the refractive index of the cell.

[0085] The image processing apparatus, image processing method, image processing program, and recording medium are not limited to the above-described embodiments and structural examples, and can be modified in various other ways.

[0086] The image processing apparatus of the above embodiment is an apparatus for processing an object image containing linear noise extending along a first direction to generate a noise-reduced image. It includes: a noise estimation unit that estimates a noise image contained in the object image; and a noise reduction unit that generates a noise-reduced image from the object image based on the object image and the noise image. The noise estimation unit uses an evaluation function to find the noise image whose value is minimized. The evaluation function includes a first term representing the result of performing differential processing in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction on the object image, and the difference between the result of performing differential processing in the second direction and low-frequency component extraction processing in the first direction on the noise image.

[0087] In the image processing apparatus described above, the noise estimation unit may also be configured to use an evaluation function that includes a second term representing the difference between the background region in the object image and the background region in the noise image, and obtain the noise image with the minimum value of the evaluation function.

[0088] In the image processing apparatus described above, the noise estimation unit may also be configured to use an evaluation function that includes a third term representing the result of high-frequency component extraction processing of the noise image in the first direction, and calculate the value of the evaluation function to obtain the noise image with the minimum value.

[0089] The image processing apparatus described above may also be configured to further include an object image generation unit, which performs integration or deconvolution processing on the differential image in the first direction to generate an object image.

[0090] The image processing apparatus described above can also be configured to use the phase image generated by performing integration or deconvolution processing on the phase differential image in the first direction as the object image.

[0091] The image processing method described above is a method for processing an object image containing linear noise extending along a first direction to generate a noise-reduced image, comprising: a noise estimation step, estimating a noise image contained in the object image; and a noise reduction step, generating a noise-reduced image from the object image based on the object image and the noise image. In the noise estimation step, an evaluation function is used to find the noise image whose value is minimized. The evaluation function includes a first term representing the result of differential processing in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction on the object image, and the difference between the result of differential processing in the second direction and low-frequency component extraction processing in the first direction on the noise image.

[0092] The image processing method described above can also be configured such that, in the noise estimation step, an evaluation function that also includes a second term representing the difference between the background region in the object image and the background region in the noise image is used to obtain the value of the evaluation function as the noise image with the minimum value.

[0093] The image processing method described above can also be configured such that, in the noise estimation step, an evaluation function containing a third term representing the result of high-frequency component extraction processing in the first direction of the noise image is used to obtain the value of the evaluation function as the noise image with the minimum value.

[0094] The image processing method described above can also be configured to include an object image generation step that further includes performing integration or deconvolution processing on the differential image in the first direction to generate the object image.

[0095] The image processing method described above can also be configured to use the phase image generated by performing integration or deconvolution processing on the phase differential image in the first direction as the object image.

[0096] The image processing program described in the above embodiments is a program used to cause a computer to execute each step of the above image processing method.

[0097] The recording medium in the above embodiments is a computer-readable medium that records the above image processing program.

[0098] [Industry availability]

[0099] The embodiments can be used as image processing apparatus, image processing method, image processing program, and recording medium capable of processing an object image containing linear noise extending in one direction to generate a noise-reduced image that maintains the quantitative characteristics of the object image.

[0100] Explanation of reference numerals in the attached figures

[0101] 1…Image processing device

[0102] 2…Recording media

[0103] 10…Control Department

[0104] 11…Object Image Generation Department

[0105] 12…Noise Estimation Department

[0106] 13…Noise Reduction Section

[0107] 14… Input Section

[0108] 15… Storage Department

[0109] 16… Display section.

Claims

1. An image processing apparatus, characterized in that: It is an apparatus for processing an object image containing linear noise extending along a first direction to generate an image with reduced noise. include: A noise estimation unit that estimates the noise image contained in the object image; and The noise reduction unit generates an image with reduced noise from the object image based on the object image and the noise image. The noise estimation unit uses an evaluation function that includes a first term to find the noise image whose value is minimized. The first term represents the difference between the result of performing differential processing in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction on the object image, and the result of performing differential processing in the second direction and low-frequency component extraction processing in the first direction on the noise image. The phase image generated by performing the integration or deconvolution process in the first direction on the phase differential image is used as the object image.

2. The image processing apparatus as claimed in claim 1, characterized in that: The noise estimation unit uses the evaluation function, which also includes a second term representing the difference between the background region in the object image and the background region in the noise image, to calculate the noise image whose value is minimized.

3. The image processing apparatus as described in claim 1 or 2, characterized in that: The noise estimation unit uses the evaluation function, which also includes a third term representing the result of high-frequency component extraction processing in the first direction of the noise image, to calculate the noise image whose value is minimized.

4. An image processing method, characterized in that: This is a method for processing an object image containing linear noise extending along a first direction to generate an image with reduced noise. include: The noise estimation step involves estimating the noise image contained in the object image; and The noise reduction step involves generating an image with reduced noise from the object image, based on the object image and the noisy image. In the noise estimation step, an evaluation function containing the first term is used to find the noise image whose value is minimized. The first term represents the difference between the result of performing differential processing in a second direction orthogonal to the first direction and low-frequency component extraction processing in the first direction on the object image, and the result of performing differential processing in the second direction and low-frequency component extraction processing in the first direction on the noise image. The phase image generated by performing the integration or deconvolution process in the first direction on the phase differential image is used as the object image.

5. The image processing method as described in claim 4, characterized in that: In the noise estimation step, the evaluation function, which also includes a second term representing the difference between the background region in the object image and the background region in the noise image, is used to find the noise image whose value is minimized.

6. The image processing method as described in claim 4 or 5, characterized in that: In the noise estimation step, the noise image whose value is minimized is obtained by using the evaluation function, which also includes a third term representing the result of high-frequency component extraction processing in the first direction of the noise image.

7. An image processing program product, characterized in that: Including image processing programs, When the image processing program is executed by a computer, it performs each step of the image processing method according to any one of claims 4 to 6.

8. A computer-readable recording medium, characterized in that: The image processing program of claim 7 is recorded.

Images