KR101426610B1 - Device and Method for Removing Noise of Image - Google Patents

Abstract

A video noise removing method and apparatus are disclosed. The disclosed apparatus includes: a diffusion calculation unit for applying an anisotropic gradient filter to calculate a diffusion energy from a peripheral pixel to a target pixel; A neighboring pixel difference calculation unit for calculating difference value information between the target pixel and neighboring pixels; And a diffusion correction unit for correcting an output value of the diffusion calculation unit based on the difference value information of the neighboring pixel difference calculation unit. According to the disclosed apparatus, there is no need for a large amount of computation for eliminating image noise, and there is an advantage in that Gaussian noise and impulse noise can be removed through a single filter

Description

TECHNICAL FIELD [0001] The present invention relates to a device and method for removing noise,

Embodiments of the present invention relate to image processing, and more particularly, to a method and apparatus for eliminating image noise.

In the image processing, various kinds of noise are accompanied, and removing such noise is an essential task in acquiring a high quality image. Types of noise include Gaussian noise, impulse noise, Poisson noise, and the like. A filter that smoothes an important region (for example, edge region) while removing such noise or removing noise is used for noise processing.

The anisotropic diffusion filter used to remove Gaussian noise is a filter applied to the spiritual processing of the diffusion equation studied in physics. The diffusion flux of the diffusion filter can be defined as Equation 1 below.

In Equation (1), J is a diffusion flux, I is a pixel, and c can be defined by the following equation (2).

In Equation (2), g is an edge stopping function and has the form of a monotone decreasing function.

The continuity equation based on Equation (1) is defined as Equation (3).

Since anisotropic diffusion is energy conservation, s (x, y, t) in Equation (3) becomes 0, and the anisotropic diffusion filter can be defined as Equation (4).

The noise reduction method using this anisotropic diffusion filter is based on the assumption that the pixels of an image are one gas particle and the Gaussian noise is eventually removed when the pixels are diffused while conserving energy.

Since the pixels of the image are discrete values, discretization is required when applying the filter of Equation (4) to the actual image, and the anisotropic diffusion filter to which the discretization is applied can be expressed as Equation (5).

In Equation (5), N (i, j) denotes a window centering on a pixel (i.j), and? Is a time step value.

On the other hand, the median filter is used to remove the impulse noise, and the impulse noise can not be removed by the anisotropic diffusion filter or the bilateral filter for eliminating the Gaussian noise.

Therefore, in order to remove the impulse noise and the Gaussian noise together, it is necessary to remove the noise by using two filters. In the conventional method, only one noise can be removed by one filter. I had to do it.

The present invention proposes an image noise removal apparatus and method that do not require a large amount of calculation.

In addition, the present invention proposes an image noise removal apparatus and method that can remove Gaussian noise and impulse noise through a single filter.

According to another aspect of the present invention, there is provided an image processing apparatus comprising: a diffusion calculation unit for calculating diffusion energy from peripheral pixels to a target pixel by applying an anisotropic fine line filter; A neighboring pixel difference calculation unit for calculating difference value information between the target pixel and neighboring pixels; And a diffusion correction unit for correcting the output value of the diffusion calculation unit based on the difference value information of the neighboring pixel difference calculation unit.

The neighboring pixel difference calculator calculates difference values between the neighboring pixels of the target pixel and the target pixel.

Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit so that the amount of energy diffused from the surrounding pixels to the target pixel is reduced when the sum of difference values calculated by the neighboring pixel difference calculation unit is equal to or less than a preset threshold value.

Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit so that the amount of energy diffused from the surrounding pixels to the target pixel increases when the sum of difference values calculated by the neighboring pixel difference calculation unit is equal to or greater than a predetermined threshold value.

The diffusion correction unit corrects the output value of the diffusion calculation unit using a correction coefficient, and the correction coefficient is set to 1 or more when the sum of difference values calculated by the peripheral pixel difference calculation unit is equal to or less than a preset threshold value.

And the spread correction unit corrects the output value of the diffusion calculation unit by a method of dividing the output value of the diffusion calculation unit by the correction coefficient.

The diffusion corrector calculates the correction coefficient using an edge blocking function that takes a difference value calculated in the neighboring pixel difference calculator as a variable.

The diffusion corrector corrects the diffusion energy according to the following equation.

[Mathematical Expression]

I is the pixel value, g is the edge suppression function, N (i. J) is the window for setting the surrounding pixels,? Is the time step size, and t is the time.

According to another aspect of the present invention, there is provided an anisotropic diffusion filter comprising: (a) calculating diffusion energy from peripheral pixels to a target pixel by applying an anisotropic diffusion filter; (B) calculating difference value information between the target pixel and neighboring pixels; And (c) correcting the diffusion energy of the step (a) based on the difference value information of the step (b).

According to the present invention, there is no need for a large amount of calculation to remove image noise, and Gaussian noise and impulse noise can be removed through a single filter.

1 is a block diagram showing the configuration of a video noise removing apparatus according to an embodiment of the present invention;
2 is a diagram for explaining a method of calculating a difference from neighboring pixels in a neighboring pixel difference calculation unit according to an embodiment of the present invention.
3 is a diagram illustrating an operation for a noise removing apparatus when a difference between a target pixel and a neighboring pixel is not large.
4 is a diagram showing the operation of the noise canceling apparatus when a difference between a target pixel and peripheral pixels is large.
5 is a flowchart showing an overall flow of a noise removing method according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing a configuration of a video noise removing apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the apparatus for removing noise according to an embodiment of the present invention may include a diffusion calculation unit 100, a neighboring pixel difference calculation unit 102, and a diffusion correction unit 104.

The diffusion calculation unit 100 calculates the energy spread from neighboring pixels to a target pixel by applying a known anisotropic diffusion filter. The spreading calculation in the spreading calculator 100 is as shown in Equation (6) below, which is the same as that of the known anisotropic spreading filter.

In Equation (6), I means pixel coordinates and c is the same as defined by Equation (2).

The neighboring pixel difference calculation unit 102 calculates a difference between pixel values of the target pixel and neighboring pixels. Here, the neighboring pixels may be defined as pixels included in the window by setting a predetermined window based on the target pixel.

2 is a diagram for explaining a method of calculating a difference from neighboring pixels in a neighboring pixel difference calculation unit according to an embodiment of the present invention.

In Fig. 2, the circle mark corresponds to the target pixel, and the diamond mark corresponds to the surrounding pixel. Referring to FIG. 2, four pixels corresponding to upper, lower, right, and left sides of a target pixel are set as neighboring pixels, and the sum of the differences between the four neighboring pixels and the target pixel and the difference between the neighboring pixels and the target pixel are And is calculated by the neighboring pixel difference calculation unit 102.

In FIG. 2, the peripheral pixels of the target pixel are shown as pixels located on the upper, lower, left, and right sides of the target pixel, but peripheral pixels may be set differently. For example, a pixel diagonally adjacent to a target pixel may be set as a neighboring pixel.

The sum of the difference values with neighboring pixels among the values calculated by the neighboring pixel difference calculation unit 102 becomes larger as the difference between a large number of neighboring pixels and a target pixel increases.

For example, when a very large impulse noise occurs in a target pixel and a pixel value of neighboring pixels is normal, the sum of difference values output from the neighboring pixel difference calculator 102 becomes very large.

The diffusion correction unit 104 corrects the diffusion energy from the peripheral pixels to the target pixel calculated by the diffusion calculation unit 100 based on the difference value calculated by the peripheral pixel difference calculation unit 102. [

As described above, the filtering performed by the diffusion calculation unit 100 can remove Gaussian noise with an average value of 0, but can not remove the impulse noise. The diffusion correction unit 104 of the present invention corrects the output value of the diffusion calculation unit 100 so that both types of noise can be removed.

Specifically, the diffusion correction unit outputs a correction coefficient? Inversely proportional to the sum of the difference values between the target pixel and the surrounding pixels, and corrects the output value of the diffusion calculation unit 100 based on?. If the sum of the difference values with neighboring pixels is large,? Becomes small, and if the sum of the difference values with neighboring pixels is small,? Becomes large.

The correction coefficient alpha is set so that the sum of the differences between the target pixel and the surrounding pixels is equal to or greater than beta. If the sum of the difference values between the target pixel and the surrounding pixels is less than or equal to?

According to an embodiment of the present invention, the spread correction unit 104 corrects the output value of the spread calculation unit 100 in such a manner that the output value of the spread calculation unit 100 is divided by the correction coefficient?. Therefore, when the difference between the neighboring pixels is less than a predetermined threshold value, alpha is equal to or greater than 1, and thus the diffusion correction unit 104 performs correction so as to reduce the energy diffusion amount. In addition, when the difference value with the neighboring pixels is equal to or greater than a predetermined threshold value,? Becomes equal to or less than 1, so that the diffusion correction unit 104 performs correction so as to increase the energy diffusion amount.

FIG. 3 is a diagram illustrating an operation for a noise removing apparatus when a difference between a target pixel and a neighboring pixel is not large.

In Fig. 3, the left image shows the amount of energy diffused from the neighboring pixel to the target pixel, and the right image is the image showing the pixel value of each pixel.

In Fig. 3, the arrow a indicates the amount of diffusion energy calculated by the diffusion calculation unit, which means the amount of diffusion energy applied to the conventional anisotropic diffusion filter.

FIG. 3 shows the case where the difference between the pixel values of the target pixel and the neighboring pixels as the center pixel is not large (when the sum of the difference values with neighboring pixels is less than a preset threshold value), and the amount of diffusion energy applied by the anisotropic diffusion filter is a. < / RTI >

In the case shown in FIG. 3, when the difference between the pixel values of the target pixel and the neighboring pixels is not large (i.e., no edge exists around the target pixel), the output value of the neighboring pixel difference calculation unit 102 Lt; / RTI > has a very small value.

In this manner, when the output value of the neighboring pixel difference calculator 102 is small, the spread correction unit 104 corrects the diffusion so that the energy diffusion from the surrounding pixels is suppressed.

 In Fig. 3, an arrow marked b indicates the amount of energy to be corrected by the diffusion correction unit 104, which is an amount that suppresses the energy introduced from the surrounding pixels in the diffusion correction unit 104. [

In Fig. 3, arrows indicated by c indicate the final amount of energy diffusion reflecting the amount (b arrow) corrected for the amount of energy diffusion (arrow a) by the anisotropic diffusion filter by arrows.

Referring to FIG. 3, it can be seen that the energy finally introduced through diffusion is smaller than that in the case where the conventional anisotropic diffusion filter is applied. This solves the problem that the conventional anisotropic diffusion filter performs excessive smoothing.

4 is a diagram illustrating an operation of the noise canceling apparatus when a difference between a target pixel and a neighboring pixel is large.

In Fig. 4, the left image shows the amount of energy diffused from the surrounding pixels to the target pixel, and the right image is the image showing the pixel value of each pixel.

As shown in FIG. 3, the arrow a indicates the amount of diffusion energy calculated by the diffusion calculation unit, which means the amount of diffusion energy applied to the conventional anisotropic diffusion filter.

In the conventional anisotropic diffusion filter, when the difference between the target pixel and the neighboring pixel is large, the target pixel is recognized as an edge. In this case, the energy spread from the neighboring pixel to the target pixel is set to be very small, Is relatively small, as indicated by the arrow labeled a.

In the case shown in FIG. 4, only the target pixel has a very different value from the surrounding pixels. In this case, the sum of the differences calculated in the neighboring pixel difference calculator 102 has a very large value, And the sum is larger than a preset threshold value. In the case shown in FIG. 4, the neighboring pixel difference calculator 102 can not output a very large difference value because all neighboring pixels have a large difference from the target pixel. Unlike FIG. 4, if some of the surrounding pixels have pixel values similar to the target pixel, the sum of difference values output from the neighboring pixel difference calculator 102 will be relatively small.

When the sum of difference values calculated by the neighboring pixel difference calculator 102 is larger than a predetermined threshold value, the diffusion correction unit 104 corrects the energy diffusion from peripheral pixels to increase.

In Fig. 4, an arrow marked b indicates the amount of energy to be corrected by the diffusion corrector, which means the amount of energy introduced from the surrounding pixels through the diffusion corrector in the case shown in Fig.

In Fig. 4, an arrow indicated by "c" indicates a final amount of diffusion reflecting the amount (b arrow) corrected to the amount of diffusion (a arrow) by the anisotropic diffusion filter by arrows.

When the noise is removed by the conventional anisotropic diffusion filter, if the impulse noise as shown in FIG. 4 occurs, the impulse noise is recognized as an edge, so that the impulse noise can not be removed.

As shown in FIG. 4, if only the target pixel has a large difference in comparison with neighboring pixels, the anisotropic diffusion filter recognizes the noise as an edge and it is impossible to remove the impulse noise even though the noise is likely to be substantially non-edge noise.

However, according to the present invention, the amount of energy diffused through the spread correction unit 104 is adjusted so that not only Gaussian noise but also impulse noise can be removed.

It can be confirmed that impulse noise is removed by allowing a very large energy to flow into the target pixel through the arrows c in FIG.

The operation of correcting the diffusion amount finally through the diffusion correction unit may be expressed by the following equation (7).

In Equation (7), the portion corresponding to the molecule corresponds to the operation of the existing anisotropic diffusion filter, and the portion corresponding to the denominator is a portion for correcting the molecular value, and the value of the denominator corresponds to the correction coefficient?. In equation (7), A (x, y, t) is an auxiliary function for deriving the sum of spreads.

Since the pixels of the image are discontinuous, it is necessary to discretize Equation (7). The discretization of Equation (7) can be expressed as Equation (8).

In Equation (8), I is a pixel value, g is an edge blocking function, N (i, j) is a window for setting surrounding pixels, and? In Equation (8), the portion corresponding to the denominator (the portion adding the edge suppression function) corresponds to the correction coefficient?.

The edge-blocking function has a small value when the difference between the target pixel and the surrounding pixels is large, and a large value when the difference between the target pixel and the surrounding pixels is small.

If the difference between the target pixel and the neighboring pixel is not large as in the case of FIG. 3, the molecule has a large value due to the edge-blocking function of the molecule. However, the larger value The noise removing apparatus of the present invention operates so that excessive energy inflow is prevented.

If the difference between the target pixel and the neighboring pixel is large as shown in FIG. 4, the molecule has a small value due to the edge blocking function of the molecule, but the small value (less than 1) by the edge blocking function of the denominator, And the noise elimination apparatus of the present invention operates so that the energy spread is made to be a large value.

Equation (8) is computed in an iterative manner, meaning that t + 1 is the next iteration when compared to t.

FIG. 5 is a flowchart showing an overall flow of a noise removing method according to an embodiment of the present invention.

Referring to FIG. 5, an amount of energy diffusion from a peripheral pixel to a target pixel is calculated by an anisotropic diffusion filter method (step 500). The amount of energy diffusion is determined based on the value of the edge inhibition function that takes the difference value from the billet pixel and the difference value with the surrounding pixels as a variable.

When the energy diffusion amount is calculated by the anisotropic diffusion filter method, the difference value between the target pixel and the surrounding pixels is calculated (step 502).

When the difference value between the target pixel and the neighboring pixels is calculated, an edge blocking function value having the difference value with neighboring pixels as a variable is calculated (step 504). As described above, if the difference value from neighboring pixels is small, the edge inhibition function outputs a relatively large value, and if the difference value from neighboring pixels is large, the edge suppression function outputs a relatively small value.

When the edge suppression function value having the difference value with the surrounding pixels as a variable is calculated, the energy diffusion amount calculated using the calculated edge suppression function value is corrected (step 506).

Specifically, the amount of energy diffusion is corrected by dividing the amount of energy diffusion by the sum of the edge-blocking functions. If the sum of the edge-blocking functions is large, the amount of energy diffusion is corrected to be reduced. .

As described above, the present invention has been described with reference to particular embodiments, such as specific elements, and specific embodiments and drawings. However, it should be understood that the present invention is not limited to the above- And various modifications and changes may be made thereto by those skilled in the art to which the present invention pertains. Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described, and all of the equivalents or equivalents of the claims, as well as the following claims, belong to the scope of the present invention .

Claims (16)

A diffusion calculation unit for calculating a diffusion energy from peripheral pixels to a target pixel by applying an anisotropic linear filter;
A neighboring pixel difference calculation unit for calculating difference value information between the target pixel and neighboring pixels; And
And a diffusion correction unit for correcting an output value of the diffusion calculation unit based on difference value information of the neighboring pixel difference calculation unit,
Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit so that the amount of energy diffused from the surrounding pixels to the target pixel increases when the sum of difference values calculated by the neighboring pixel difference calculation unit is equal to or greater than a predetermined threshold value A video noise canceling device. The method according to claim 1,
Wherein the peripheral pixel difference calculation unit calculates a difference value between pixels adjacent to the upper, lower, left, and right sides of the target pixel with respect to the target pixel. The method according to claim 1,
Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit so that the amount of energy diffused from the surrounding pixels to the target pixel is reduced when the sum of difference values calculated by the neighboring pixel difference calculation unit is equal to or less than a preset threshold value A video noise canceling device. delete The method according to claim 1 or 3,
Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit using a correction coefficient, and the correction coefficient is set to 1 or more when the sum of difference values calculated by the neighboring pixel difference calculation unit is equal to or less than a preset threshold value Image noise removal device. 6. The method of claim 5,
Wherein the diffusion correction unit corrects the output value of the diffusion calculation unit by dividing the output value of the diffusion calculation unit by the correction coefficient. The method according to claim 6,
Wherein the diffusion correction unit calculates the correction coefficient using an edge blocking function having a difference value calculated by the neighboring pixel difference calculator as a variable. The method according to claim 6,
Wherein the diffusion corrector corrects the spreading energy according to the following equation.

I is the pixel value, g is the edge suppression function, N (i. J) is the window for setting the surrounding pixels,? Is the time step size, and t is the time. (A) applying an anisotropic diffusion filter to calculate diffusion energy from peripheral pixels to a target pixel;
(B) calculating difference value information between the target pixel and neighboring pixels;
And (c) correcting the diffusion energy of the step (a) based on the difference value information of the step (b)
The step (c)
Wherein the diffusion energy is corrected so that the amount of energy diffused from the surrounding pixels to the target pixel increases when the sum of difference values calculated in step (b) is equal to or greater than a preset threshold value. 10. The method of claim 9,
Wherein the step (b) calculates a difference value between pixels adjacent to the upper, lower, left, and right sides of the target pixel. 11. The method of claim 10,
The step (c)
Wherein the diffusion energy is corrected so that the amount of energy diffused from the surrounding pixels to the target pixel is reduced when the sum of difference values calculated in step (b) is less than or equal to a preset threshold value. delete The method according to claim 9 or 11,
Wherein the step (c) corrects the diffusion energy using a correction coefficient, and the correction coefficient is set to 1 or more when the sum of difference values calculated in step (b) is less than a preset threshold value. Noise removal method. 14. The method of claim 13,
Wherein the step (c) corrects the diffusion energy by dividing the diffusion energy by the correction coefficient. 15. The method of claim 14,
Wherein the step (c) calculates the correction coefficient using an edge blocking function having a difference value calculated in step (b) as a variable. 15. The method of claim 14,
Wherein the step (c) corrects the diffusion energy according to the following equation.

I is the pixel value, g is the edge suppression function, N (i. J) is the window for setting the surrounding pixels, tau is the time step size, and t is the time.

Images