KR102126916B1 - Apparatus and method for reconstructing blurred image - Google Patents

Abstract

Disclosed is an apparatus and method for reconstructing blurred images. The image reconstruction apparatus according to an embodiment of the present invention, while moving a window in an image to be reconstructed including a blurring noise, rotates filtering using an inverse filter for pixels included in the window a filtering unit that performs (rotate filtering); And a determination unit determining whether to perform the rotation filtering on the reconstructed image based on the filtering result by the rotation filtering, and wherein the filtering unit includes the window and By changing the size of the inverse filter, additional rotation filtering is performed on the reconstructed image.

Description

Apparatus and method for restoring blurred images{APPARATUS AND METHOD FOR RECONSTRUCTING BLURRED IMAGE}

Embodiments of the present invention relate to an image reconstruction technique for removing blurring noise included in a blurred image.

In the field of forensic science, which has recently gained importance, research on reconstruction of criminal cue images with low visibility and low visibility is actively underway.

An inverse filtering technique exists as one of the basic methods for reconstructing a digital image. This is to predict the deterioration function that damages the original image and divide it into inverses to obtain the original image through inverse Fourier transform. If the deterioration function can be accurately designed, high-performance image restoration is possible. However, this technique is based on the assumption that the blurring noise of the damaged image is homogeneous and is damaged by one of the complex blurring noises (vertical blurring noise, horizontal blurring noise, and rotating blurring noise). As the blur model must be classified and filtered one by one, such as a blur, a two-dimensional blur, a Gaussian blur, etc., there is a disadvantage that it cannot flexibly cope with the complex blurring noise characteristics of an actual reconstructed image.

Republic of Korea Registered Patent No. 10-0990791 (2010.10.22.

Embodiments of the present invention are to provide an image reconstruction apparatus and method for removing blurring noise included in a blurred image.

The image reconstruction apparatus according to an embodiment of the present invention, while moving a window in an image to be reconstructed including a blurring noise, rotates filtering using an inverse filter for pixels included in the window a filtering unit that performs (rotate filtering); And a determination unit determining whether to perform the rotation filtering on the reconstructed image based on the filtering result by the rotation filtering, and wherein the filtering unit includes the window and By changing the size of the inverse filter, additional rotation filtering is performed on the reconstructed image.

The filtering unit may determine the size of the window and the inverse filter based on the number of times the rotation filtering is performed.

When the rotation filtering performed on the image to be reconstructed is the n-th rotation filtering (where n is a natural number of n≥1), the filtering unit sets the size of the window to (2n+1)×(2n+1). Can be set.

The filtering unit may set the size of the inverse filter to (2n-1)×(2n+1) when rotation filtering performed on the reconstructed target image is an n-th rotation filtering.

The filtering unit may set a filter coefficient of the inverse filter by using Equation 1 below when rotation filtering performed on the reconstructed target image is n-th rotation filtering.

[Equation 1]

(At this time, C _i,j is the filter coefficient)

When the rotation filtering performed on the restoration target image is the n-th rotation filtering, the filtering unit may set the size of the window to 1×(2n+1).

The filtering unit may set a filter coefficient of the inverse filter using Equation 2 below when rotation filtering performed on the reconstructed target image is an n-th rotation filtering.

[Equation 2]

(At this time, C _j is the filter coefficient)

The determining unit calculates an average mean square error (AMSE) between a pixel value of each pixel included in the reconstructed target image and rotation filtering result values for each pixel, and further performs the rotation filtering based on the AMSE Can determine whether

The determination unit is performed on the AMSE calculated using the rotation filtering result values of the n-1 (in this case, n is a natural number of n≥2) rotation performed on the restoration target image and the restoration target image It may be determined whether the rotation filtering is additionally performed based on an absolute error between AMSEs calculated using the rotation filtering result values by the n-th rotation filtering.

The AMSE can be calculated using Equation 3 below.

[Equation 3]

는 상기 회전 필터링 결과 값)(In this case, n is the number of times the rotation filtering is performed, x is the pixel value,

Is the rotation filtering result value)

When the n-th rotation filtering is performed on the restoration target image, the filtering unit calculates an average value of rotation filtering result values for each pixel of the restoration target image using Equation 4 below, and the n-th rotation When the filtering is rotation filtering performed last on the reconstructed image, the average value may be determined as a pixel value of the reconstructed image.

[Equation 4]

은 상기 평균 값,

는 상기 회전 필터링 결과 값)(At this time,

Is the average value,

Is the rotation filtering result value)

In an image restoration method according to an embodiment of the present invention, rotation filtering using an inverse filter for pixels included in the window while moving a window in a restoration target image including blurring noise (rotate filtering); Determining whether the rotation filtering is additionally performed on the reconstructed target image based on a result of filtering by the rotation filtering; And when it is necessary to additionally perform the rotation filtering, changing the size of the window and the inverse filter to perform additional rotation filtering on the reconstructed image.

The size of the window and the inverse filter may be determined based on the number of times the rotation filtering is performed.

The step of performing the rotation filtering and the step of performing the additional rotation filtering may include the size of the window when the rotation filtering performed on the reconstructed target image is an n (where n is a natural number with n≥1) th rotation filtering. Can be set to (2n+1)×(2n+1).

In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the reconstructed target image is the n-th rotation filtering, the size of the inverse filter is (2n-1)×(2n+) It can be set as 1).

In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the image to be reconstructed is the n-th rotation filtering, filter coefficients of the inverse filter using Equation 1 below ( coefficient).

[Equation 1]

(At this time, C _i,j is the filter coefficient)

In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the restoration target image is the n-th rotation filtering, the size of the window may be set to 1×(2n+1). have.

In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the reconstructed target image is the n-th rotation filtering, filter coefficients of the inverse filter using Equation 2 below ( coefficient).

[Equation 2]

(At this time, C _j is the filter coefficient)

In the determining step, an average mean square error (AMSE) between a pixel value of each pixel included in the reconstructed image and a rotation filtering result value for each pixel is calculated, and the rotation filtering is performed based on the AMSE. It can be determined whether or not to perform additional.

In the determining, the AMSE calculated using the rotation filtering result values of the n-1 (where n is a natural number of n≥2) rotation performed on the restoration target image and the restoration target image It is possible to determine whether to perform the rotation filtering further based on the absolute error between AMSE calculated using the rotation filtering result values obtained by the n-th rotation filtering.

The AMSE can be calculated using Equation 3 below.

[Equation 3]

는 상기 회전 필터링 결과 값)(In this case, n is the number of times the rotation filtering is performed, x is the pixel value,

Is the rotation filtering result value)

In the image reconstruction method, when n-th rotation filtering is performed on the restoration target image, calculating an average value of rotation filtering result values for each pixel of the restoration target image using Equation 4 below; And if the n-th rotation filtering is rotation filtering performed last on the reconstructed target image, determining the average value as a pixel value of the reconstructed image.

[Equation 4]

은 상기 평균 값,

는 상기 회전 필터링 결과 값)(At this time,

Is the average value,

Is the rotation filtering result value)

According to embodiments of the present invention, by removing various blurring noises included in an image to be reconstructed, such as vertical blurring noise, horizontal blurring noise, and rotation blurring noise, through rotation filtering, included in the reconstruction target image Flexible noise reduction is possible due to the characteristics of blurring noise.

1 is a configuration diagram of an image restoration apparatus according to an embodiment of the present invention
2 is an exemplary diagram for explaining an example of rotation filtering performed for a restoration target for the first time;
3 is an exemplary diagram for explaining an example of rotation filtering performed for a second restoration target;
4 is an exemplary view for explaining an example of rotation filtering performed for a third time for a restoration target
5 is an exemplary view for explaining another example of rotation filtering performed for a second restoration target;
FIG. 6 is an exemplary diagram for explaining another example of rotation filtering performed for a third restoration target;
7 is a flowchart of an image restoration method according to an embodiment of the present invention
8 is a flowchart illustrating a detailed procedure of an image restoration method according to an embodiment of the present invention
9 is a block diagram illustrating and illustrating a computing environment including a computing device suitable for use in exemplary embodiments of the present invention

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The following detailed description is provided to aid in a comprehensive understanding of the methods, devices and/or systems described herein. However, this is only an example and the present invention is not limited thereto.

In describing the embodiments of the present invention, when it is determined that a detailed description of known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to a user's or operator's intention or practice. Therefore, the definition should be made based on the contents throughout this specification. The terminology used in the detailed description is only for describing embodiments of the present invention and should not be limiting. Unless expressly used otherwise, a singular form includes a plural form. In this description, expressions such as “including” or “equipment” are intended to indicate certain characteristics, numbers, steps, actions, elements, parts or combinations thereof, and one or more other than described. It should not be interpreted to exclude the presence or likelihood of other features, numbers, steps, actions, elements, or parts or combinations thereof.

1 is a configuration diagram of an image restoration apparatus according to an embodiment of the present invention.

Referring to FIG. 1, an image reconstruction apparatus 100 according to an embodiment of the present invention includes a filtering unit 110 and a determination unit 120.

In an embodiment of the present invention, the image reconstruction apparatus 100 is for generating a reconstructed image in which blurring noise is removed from an image to be reconstructed that includes blurring noise. In this case, the blurring noise included in the reconstructed target image may include at least one of a vertical blurring noise, a horizontal blurring noise, and a rotating blurring noise.

The filtering unit 110 performs rotation filtering using an inverse filter on pixels included in the window while moving the window in the reconstructed image. In addition, as described later, the filtering unit 110 performs additional rotation filtering on the image to be reconstructed by changing the size of the window and the inverse filter when additional rotation filtering is required.

At this time, the inverse filter uses the fact that the process of deteriorating the original image consists of filtering between the original image and the transfer function that generates blur. After predictive modeling the transfer function that generates blur, the predicted modeled transfer function and inverse relationship To remove the blurring noise included in the degraded image by performing filtering using the function in the transfer function.

Meanwhile, according to an embodiment of the present invention, the rotation filtering performed by the filtering unit 110 performs filtering for each pixel included in the window while rotating the inverse filter clockwise or counterclockwise within the window. However, it refers to a filtering operation in which filtering is performed while sequentially moving a window within a predetermined interval within a restoration target image to remove blurring noise included in the restoration target image. At this time, the filtering unit 110, for example, after setting the window in the upper left of the image to be restored, the window can be sequentially moved in the X-axis and Y-axis directions in the restoration target image at regular intervals.

On the other hand, according to an embodiment of the present invention, after the rotation filtering for the reconstructed target image is completed, when additional rotation filtering is required, the filtering unit 110 changes the size of the window and inverse filter to the reconstructed target image Additional rotation filtering can be performed. That is, the filtering unit 110 may repeatedly perform rotation filtering on the reconstructed image while changing the size of the window and inverse filter, where the size of the window and the inverse filter perform rotation filtering on the reconstructed image It can be determined based on the number of times.

Specifically, according to an embodiment of the present invention, the filtering unit 110 increases the size of the window when rotation filtering to be performed on an image to be restored is the nth rotation filtering (where n is a natural number of n≥1). It can be set as (2n+1)×(2n+1). That is, the filtering unit 110 sets the size of the window to 3×3 when performing rotation filtering on the reconstructed image for the first time, and when performing rotation filtering on the reconstructed image for the second time, the size of the window Can be set to 5×5.

Meanwhile, according to an embodiment of the present invention, the filtering unit 110 determines the size of the inverse filter when rotation filtering to be performed on an image to be reconstructed is the nth rotation filtering (where n is a natural number with n≥1). It can be set as (2n-1)×(2n+1). In this case, the filtering unit 110 may set the filter coefficient of the inverse filter using Equation 1 below.

[Equation 1]

In Equation 1, C _i,j represents the filter coefficient of the inverse filter, and i and j are indexes representing rows and columns when the inverse filter is viewed as a matrix.

That is, when performing rotation filtering on the reconstructed image for the first time, the filtering unit 110 may set the size of the inverse filter to 1×3 and set the filter coefficient to 1/3. Also, when performing rotation filtering on the reconstructed image for the second time, the filtering unit 110 may set the size of the inverse filter to 3×5 and set the filter coefficient to 1/15.

Meanwhile, according to another embodiment of the present invention, the filtering unit 110 increases the size of the inverse filter when the rotation filtering to be performed on the restoration target image is the nth rotation filtering (where n is a natural number with n≥1). Can be set to 1×(2n+1). In this case, the filtering unit 110 may set the filter coefficient of the inverse filter using Equation 2 below.

[Equation 2]

That is, when performing rotation filtering on the reconstructed image for the first time, the filtering unit 110 may set the size of the inverse filter to 1×3 and set the filter coefficient to 1/3. Also, when performing rotation filtering on the reconstructed image for the second time, the filtering unit 110 may set the size of the inverse filter to 1×5 and the filter coefficient to 1/5.

Specifically, FIG. 2 is an exemplary diagram for explaining an example of rotation filtering performed for a restoration object for the first time.

The circled portion in FIG. 2 represents pixels of an image to be reconstructed included in a window.

Referring to FIG. 2, when rotation filtering is first performed, the size of the window may be set to 3×3, and accordingly, 9 pixels may be included in the window. Also, the size of the inverse filter 200 may be set to 1×3, and the filter coefficient of the inverse filter 200 may be set to 1/3. In addition, the filtering unit 110 may rotate the inverse filter 200 in a clockwise direction within a window, and perform filtering on pixels in the window as illustrated.

3 is an exemplary view for explaining an example of rotation filtering performed for a second restoration target.

The circled portion in FIG. 3 represents pixels of an image to be reconstructed included in a window.

Referring to FIG. 3, when rotation filtering is performed a second time, the size of the window may be set to 5×5, and accordingly, 25 pixels may be included in the window. In addition, the size of the inverse filter 300 may be set to 3×5, and the filter coefficient of the inverse filter 300 may be set to 1/15. Also, as illustrated, the filtering unit 110 may rotate the inverse filter 300 in a clockwise direction within the window, and perform filtering on pixels in the window.

4 is an exemplary view for explaining an example of rotation filtering performed for a third time to a restoration target.

The circled portion in FIG. 4 represents pixels of an image to be reconstructed included in a window.

Referring to FIG. 4, when rotation filtering is performed a third time, the size of the window may be set to 7×7, and accordingly, 49 pixels may be included in the window. In addition, the size of the inverse filter 400 may be set to 5×7, and the filter coefficient of the inverse filter 400 may be set to 1/35. Also, the filtering unit 110 may rotate the inverse filter 400 in a clockwise direction within the window, and perform filtering on pixels in the window as illustrated.

5 is an exemplary view for explaining another example of rotation filtering performed for a second restoration target.

The circled portion in FIG. 5 represents pixels of an image to be reconstructed included in a window.

Referring to FIG. 5, when rotation filtering is performed a second time, the size of the window may be set to 5×5, and accordingly, 25 pixels may be included in the window. In addition, the size of the inverse filter 500 may be set to 1×5, and the filter coefficient of the inverse filter 500 may be set to 1/5. Also, the filtering unit 110 may rotate the inverse filter 500 in a clockwise direction in the window, and perform filtering on pixels in the window.

6 is an exemplary diagram for explaining another example of rotation filtering performed for a third time to a restoration target.

The circled portion in FIG. 6 represents pixels of an image to be reconstructed included in a window.

Referring to FIG. 6, when rotation filtering is performed a third time, the size of the window may be set to 7×7, and accordingly, 49 pixels are included in the window. In addition, the size of the inverse filter 600 is set to 1×7, and the filter coefficient of the inverse filter 600 is set to 1/7. Also, the filtering unit 110 may rotate the inverse filter 600 in a clockwise direction within the window, and perform filtering on pixels in the window as illustrated.

Referring to FIG. 1 again, the determination unit 120 determines whether to perform rotation filtering on the reconstructed target image based on the rotation filtering result performed on the reconstructed target image.

According to an embodiment of the present invention, the determination unit 120 calculates AMSE (Average Mean Square Error) using the pixel values of each pixel included in the reconstructed target image and the results of applying rotation filtering to each pixel, Based on the calculated AMSE, it may be determined whether to perform rotation filtering on the reconstructed target image.

)들이 출력될 수 있다. 이때, 판단부(120)는 예를 들어, 아래의 수학식 3을 이용하여 복원 대상 영상에 포함된 특정 픽셀 x의 픽셀 값과 해당 픽셀 x에 대한 n번째 회전 필터링 결과 값(

)들 사이의 AMSE(

)를 산출할 수 있다.Specifically, as illustrated in FIGS. 2 to 6, when n-th rotation filtering is performed on an image to be reconstructed, filtering is repeatedly performed while an inverse filter rotates within a window. Accordingly, when n-th rotation filtering is performed, 4n-1 rotation filtering result values (for a specific pixel x included in the reconstructed image)

) Can be output. At this time, the determination unit 120, for example, by using Equation 3 below, the pixel value of a specific pixel x included in the reconstructed image and the n-th rotation filtering result value for the pixel x (

) Between AMSE(

).

[Equation 3]

Subsequently, the determination unit 120 performs the AMSE calculated using the n-1 (in this case, n is a natural number of n≥2) rotation filtering performed on the restoration target image and the nth performed on the restoration target image The absolute value error between the calculated AMSEs may be calculated using the rotation filtering result values, and it may be determined whether additional rotation filtering is performed on the reconstructed image based on the calculated absolute value error. Specifically, the determining unit 120 may determine that necessary in this case the calculated absolute error as shown in Equation 4 below, the group is less than the predetermined blurring limit of Δ _t, perform an additional rotation filter of the restoration target image.

[Equation 4]

On the other hand, if it is determined by the determination unit 120 that additional rotation filtering needs to be performed, the filtering unit 110 changes the size of the window and inverse filter as described above, and then performs additional rotation filtering on the reconstructed image. can do.

On the other hand, when it is determined that the additional rotation filtering is unnecessary by the determination unit 110, the filtering unit 110 displays the rotation filtering result values for each pixel obtained by the rotation filtering performed last for the reconstructed image. The average value can be determined as the pixel value of the reconstructed image.

)들의 평균 값(

)을 산출할 수 있다. Specifically, when the n-th rotation filtering is performed on the restoration target image, the filtering unit 110 uses the following equation (4) to calculate the rotation filtering result value for each pixel of the restoration target image (

Mean values of)

).

[Equation 3]

을 복원 영상의 픽셀 값으로 결정할 수 있다.In addition, the filtering unit 110, if the n-th rotation filtering is the last rotation filtering performed on the image to be restored,

Can be determined as the pixel value of the reconstructed image.

7 is a flowchart of an image restoration method according to an embodiment of the present invention.

The method illustrated in FIG. 7 may be performed, for example, by the image restoration apparatus 100 illustrated in FIG. 1.

Referring to FIG. 7, the image reconstruction apparatus 100 first performs rotation filtering using an inverse filter on pixels included in the window while moving the window within the reconstructed target image including the blurring noise (710) .

Thereafter, the image reconstructing apparatus 100 determines whether additional rotation filtering is performed on the reconstructed image based on the filtering result by the rotation filtering (720 ).

Subsequently, when it is necessary to additionally perform rotation filtering on the restoration target image, the image restoration apparatus 100 changes the sizes of the window and inverse filters (730), and then performs steps 710 and 720 again.

8 is a flowchart illustrating a detailed procedure of an image restoration method according to an embodiment of the present invention.

The method illustrated in FIG. 8 may be performed, for example, by the image restoration apparatus 100 illustrated in FIG. 1.

Referring to FIG. 8, the image restoration apparatus 100 first sets the number (n) of rotation filtering performed on the restoration target image to 1 (810).

Thereafter, the image restoration apparatus 100 determines the size of the window and the inverse filter based on the number of times (n) of performing filtering (820 ).

At this time, according to an embodiment of the present invention, the size of the window may be set to (2n+1)×(2n+1).

Further, according to an embodiment of the present invention, the size of the inverse filter may be set as (2n-1)×(2n+1), in which case, the filter coefficient of the inverse filter may be determined according to Equation 1 above. Can.

According to another embodiment of the present invention, the size of the inverse filter may be set to 1×(2n+1), and in this case, the filter coefficient of the inverse filter may be determined according to Equation 2 described above.

Thereafter, the image restoration apparatus 100 performs rotation filtering on the image to be reconstructed using a window and an inverse filter (830 ).

)을 산출한다(840). Subsequently, the image restoration apparatus 100 averages the rotation filtering result values for each pixel included in the restoration target image (

) Is calculated (840).

Thereafter, when the rotation filtering performed in operation 830 is the second or more rotation filtering performed on the restoration target (850), the image restoration apparatus 100 calculates AMSE(e _n ) using the rotation filtering result values. (860).

At this time, if the rotation filtering performed in step 830 is the first rotation filtering performed on the restoration target (850), the image restoration apparatus 100 increases the number of times (n) of rotation filtering by 1 (880), Steps 820 to 850 are performed again.

Thereafter, the image restoration apparatus 100 uses the AMSE(e _n ) calculated using the n-th rotation filtering result values performed on the restoration target image and the n-1-th rotation filtering result values performed on the restoration target image. It is determined whether the absolute error between the calculated AMSE (e _n-1 ) is greater than a preset blurring limit value Δ _t (870).

In this case, when the absolute value error is equal to or less than the blurring limit value, the image restoration apparatus 100 increases the number of times of performing rotation filtering by n (880), and performs steps 820 to 870 again.

을 복원 영상의 필셀 값으로 결정한다(890).On the other hand, if the absolute value error is greater than the blurring limit value, the image restoration apparatus 100

Is determined as the fill cell value of the reconstructed image (890).

Meanwhile, in the flowcharts illustrated in FIGS. 7 and 8, the method is described as being divided into a plurality of steps, but at least some of the steps are performed by changing the order, combined with other steps, omitted together, or as detailed steps. It may be performed separately, or may be performed by adding one or more steps not shown.

9 is a block diagram illustrating and illustrating a computing environment including a computing device suitable for use in example embodiments. In the illustrated embodiment, each component may have different functions and capabilities other than those described below, and may include additional components other than those described below.

The illustrated computing environment 10 includes a computing device 12. In one embodiment, the computing device 12 may be one or more components included in the image restoration device 100 illustrated in FIG. 1.

The computing device 12 includes at least one processor 14, a computer readable storage medium 16 and a communication bus 18. The processor 14 can cause the computing device 12 to operate in accordance with the exemplary embodiment mentioned above. For example, processor 14 may execute one or more programs stored on computer readable storage medium 16. The one or more programs may include one or more computer-executable instructions, which, when executed by processor 14, configure computing device 12 to perform operations according to an exemplary embodiment. Can be.

Computer readable storage medium 16 is configured to store computer executable instructions or program code, program data and/or other suitable types of information. The program 20 stored on the computer readable storage medium 16 includes a set of instructions executable by the processor 14. In one embodiment, the computer-readable storage medium 16 is a memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash Memory devices, other types of storage media that can be accessed by the computing device 12 and store desired information, or suitable combinations thereof.

The communication bus 18 interconnects various other components of the computing device 12, including a processor 14 and a computer readable storage medium 16.

Computing device 12 may also include one or more I/O interfaces 22 and one or more network communication interfaces 26 that provide an interface for one or more I/O devices 24. The input/output interface 22 and the network communication interface 26 are connected to the communication bus 18. The input/output device 24 may be connected to other components of the computing device 12 through the input/output interface 22. Exemplary input/output devices 24 include pointing devices (such as a mouse or trackpad), keyboards, touch input devices (such as touch pads or touch screens), voice or sound input devices, various types of sensor devices, and/or imaging devices. Input devices, and/or output devices such as display devices, printers, speakers, and/or network cards. The exemplary input/output device 24 is a component constituting the computing device 12 and may be included in the computing device 12 or connected to the computing device 12 as a separate device distinct from the computing device 12. It might be.

The present invention has been described in detail through exemplary embodiments, but those skilled in the art to which the present invention pertains can make various modifications to the above-described embodiments without departing from the scope of the present invention. Will understand. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims to be described later, but also by the claims and equivalents.

10: computing environment
12: computing device
14: processor
16: computer readable storage media
18: Communication bus
20: Program
22: I/O interface
24: I/O device
26: network communication interface
100: video restoration device
110: filtering unit
120: judgment unit

Claims (22)

A filtering unit that performs rotation filtering using an inverse filter on pixels included in the window while moving a window in an image to be reconstructed including blurring noise; And
Based on the filtering result by the rotation filtering, and includes a determination unit for determining whether to perform the rotation filtering for the reconstructed image,
The filtering unit performs additional rotation filtering on the image to be reconstructed by changing sizes of the window and the inverse filter when additional rotation filtering is required.
The determining unit calculates an average mean square error (AMSE) between a pixel value of each pixel included in the reconstructed target image and rotation filtering result values for each pixel, and further performs the rotation filtering based on the AMSE Image restoration device for determining whether.
The method according to claim 1,
The filtering unit determines an image size of the window and the inverse filter based on the number of times the rotation filtering is performed.
The method according to claim 2,
When the rotation filtering performed on the image to be reconstructed is the n-th rotation filtering (where n is a natural number of n≥1), the filtering unit sets the size of the window to (2n+1)×(2n+1). Video restoration device to set.
The method according to claim 3,
The filtering unit sets the size of the inverse filter to (2n-1)×(2n+1) when rotation filtering performed on the restoration target image is an n-th rotation filtering.
The method according to claim 4,
The filtering unit, when the rotation filtering performed on the image to be restored is the n-th rotation filtering, an image restoration apparatus that sets a filter coefficient of the inverse filter using Equation 1 below.
[Equation 1]

(At this time, C _i,j is the filter coefficient)
The method according to claim 3,
The filtering unit sets the size of the window to 1×(2n+1) when rotation filtering performed on the restoration target image is n-th rotation filtering.
The method according to claim 6,
When the rotation filtering performed on the restoration target image is the n-th rotation filtering, the filtering unit sets the filter coefficient of the inverse filter by using Equation 2 below.
[Equation 2]

(At this time, C _j is the filter coefficient)
delete The method according to claim 1,
The determination unit is performed on the AMSE calculated using the rotation filtering result values of the n-1 (in this case, n is a natural number of n≥2) rotation performed on the restoration target image and the restoration target image An image restoration apparatus for determining whether to perform the rotation filtering based on an absolute error between AMSEs calculated using rotation filtering result values of the n-th rotation filtering.
The method according to claim 1,
The AMSE is an image reconstruction apparatus calculated using Equation 3 below.
[Equation 3]

(In this case, n is the number of times the rotation filtering is performed, x is the pixel value,

Is the rotation filtering result value)
The method according to claim 1,
When the n-th rotation filtering is performed on the restoration target image, the filtering unit calculates an average value of rotation filtering result values for each pixel of the restoration target image using Equation 4 below, and the n-th rotation An image restoration apparatus that determines the average value as a pixel value of a restoration image when filtering is rotation filtering performed last on the restoration target image.
[Equation 4]

(At this time,

Is the average value,

Is the rotation filtering result value)
Performing rotation filtering using an inverse filter on pixels included in the window while moving a window in a restoration target image including blurring noise;
Determining whether the rotation filtering is additionally performed on the reconstructed target image based on a result of filtering by the rotation filtering; And
If it is necessary to perform the rotation filtering further, changing the size of the window and the inverse filter to perform additional rotation filtering on the reconstructed image,
In the determining step, an average mean square error (AMSE) between a pixel value of each pixel included in the reconstructed image and a rotation filtering result value for each pixel is calculated, and the rotation filtering is performed based on the AMSE. Image restoration method to determine whether to perform additional.
The method according to claim 12,
The size of the window and the inverse filter is determined based on the number of times the rotation filtering is performed.
The method according to claim 13,
The step of performing the rotation filtering and the step of performing the additional rotation filtering may include the size of the window when the rotation filtering performed on the reconstructed target image is an n (where n is a natural number with n≥1) th rotation filtering. How to set (2n+1) × (2n+1).
The method according to claim 14,
In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the reconstructed target image is the n-th rotation filtering, the size of the inverse filter is (2n-1)×(2n+) How to restore video set by 1).
The method according to claim 15,
In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the reconstructed target image is the n-th rotation filtering, the filter coefficient of the inverse filter using Equation 1 below ( coefficient) setting method.
[Equation 1]

(At this time, C _i,j is the filter coefficient)
The method according to claim 14,
In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the restoration target image is the n-th rotation filtering, the size of the window is set to 1×(2n+1). Video restoration method.
The method according to claim 17,
In the step of performing the rotation filtering and the step of performing the additional rotation filtering, when the rotation filtering performed on the reconstructed target image is the n-th rotation filtering, filter coefficients of the inverse filter using Equation 2 below ( coefficient) setting method.
[Equation 2]

(At this time, C _j is the filter coefficient)
delete The method according to claim 12,
In the determining, the AMSE calculated using the rotation filtering result values of the n-1 (where n is a natural number of n≥2) rotation performed on the restoration target image and the restoration target image An image restoration method for determining whether to perform the additional rotation filtering based on an absolute error between AMSEs calculated using the rotation filtering result values obtained by the n-th rotation filtering.
The method according to claim 12,
The AMSE is an image reconstruction method calculated using Equation 3 below.
[Equation 3]

(In this case, n is the number of times the rotation filtering is performed, x is the pixel value,

Is the rotation filtering result value)
The method according to claim 12,
When n-th rotation filtering is performed on the restoration target image, calculating an average value of rotation filtering result values for each pixel of the restoration target image using Equation 4 below; And
And if the n-th rotation filtering is rotation filtering performed last on the reconstructed target image, determining the average value as a pixel value of the reconstructed image.
[Equation 4]

(At this time,

Is the average value,

Is the rotation filtering result value)

Images