CN111915535B

CN111915535B - Image denoising method and device

Info

Publication number: CN111915535B
Application number: CN202010797044.7A
Authority: CN
Inventors: 王微
Original assignee: Spreadtrum Communications Tianjin Co Ltd
Current assignee: Spreadtrum Communications Tianjin Co Ltd
Priority date: 2015-12-30
Filing date: 2015-12-30
Publication date: 2022-09-27
Anticipated expiration: 2035-12-30
Also published as: CN106934768A; CN111915534B; CN111915534A; CN106934768B; CN111915535A

Abstract

A method and a device for denoising an image carry out recursive filtering on pixels in the image, and adopt the following method to obtain the filtering strength of the current pixel to be processed: calculating edge information on horizontal and vertical chrominance components of the central pixel, and taking the maximum value as a first edge information maximum value; calculating the variance of the image block; calculating a first maximum difference between the central pixel and the selected filtered pixel in the (k-1) th row/(k-1) th column in the image block; calculating edge information on a luminance component of the center pixel; judging whether the frequency area where the central pixel is located is a flat area or not; when the frequency area where the central pixel is located is a flat area, calculating the current brightness of the central pixel; and acquiring a first filtering intensity as the filtering intensity of the central pixel according to the current brightness. By adopting the scheme, the denoising effect of noise can be ensured, and the trailing phenomenon of the denoised image can be avoided.

Description

Image denoising method and device

Technical Field

The present invention relates to the field of image processing, and in particular, to a method and an apparatus for image denoising.

Background

No matter digital cameras, mobile phone cameras or computer cameras are interfered by color noise more or less in the process of image acquisition. Generally, in order to suppress color noise, the most common method is to design a filter to filter out the color noise. For the high frequency characteristics of color noise, a series of low pass filters are usually designed to suppress the color noise. Such filters use a template of a certain size. In order to effectively remove color noise with low frequency, a larger template is often used, and an Infinite Impulse Response (IIR) filter is often used to simulate the denoising function of the larger template due to its characteristic transfer characteristic.

In order to filter color noise, currently, when IIR is used, the following steps are used to perform denoising: and judging whether the image block is a flat area or not according to the edge information and the variance of the horizontal and vertical directions of the image block and the difference between the central pixel of the image block and the chrominance component of the selected pixel in the previous row or column. When the image block is a flat area, determining the filtering strength of the image block according to the brightness of the image block and the difference between the chrominance components of the central pixel and the peripheral pixels of the image block, and then filtering the image block by using the filtering strength.

However, when the image denoising processing is performed by the above method, if the filtering strength is too weak, the denoising effect on the noise may be poor, or if the filtering strength is too strong, the smearing phenomenon may occur on the denoised image.

Disclosure of Invention

The invention solves the problem of how to avoid the trailing phenomenon of the denoised image while ensuring the denoising effect of the noise.

In order to solve the above problem, an embodiment of the present invention provides a method for denoising an image, where recursive filtering is performed on pixels in the image, and the following method is used to obtain filtering strength of a current pixel to be processed, where the method includes:

acquiring an image block with the pixel to be processed as a central pixel, wherein the central pixel is positioned in the kth row and the kth column of the image block;

calculating edge information on horizontal and vertical chrominance components of the central pixel, and taking the maximum value of the edge information on the horizontal and vertical chrominance components as a first edge information maximum value;

calculating the variance of the image block;

calculating the difference maximum value of the central pixel and the selected filtered pixel in the k-1 th row/k-1 th column in the image block as a first difference maximum value;

calculating edge information on a luminance component of the center pixel;

judging whether the frequency area where the central pixel is located is a flat area or not according to the first edge information maximum value, the variance of the image block, the first difference maximum value and the edge information on the brightness component of the central pixel;

when the frequency region where the central pixel is located is a flat region, calculating the current brightness of the central pixel;

and acquiring a first filtering strength as the filtering strength of the central pixel according to the current brightness of the central pixel.

Optionally, the determining, according to the first maximum edge information value, the variance of the image block, the first maximum difference value, and the edge information on the luminance component of the central pixel, whether a frequency region where the central pixel is located is a flat region includes:

determining the frequency of the center pixel as a flat region when all of the following conditions are satisfied:

the maximum value of the first edge information is smaller than a preset first threshold value;

the variance is smaller than a preset second threshold value;

the first difference maximum value is smaller than a preset third threshold value;

and the edge information on the brightness component of the central pixel is smaller than a preset fourth threshold value.

Optionally, values of the first threshold, the second threshold, the third threshold, and the fourth threshold are different according to a different luminance interval in which the current luminance of the central pixel of the image block is located.

Optionally, the method further comprises:

and when the current brightness of the central pixel is lower than a preset dark-area brightness threshold value, determining that the central pixel is in a dark area, and performing dark-area saturation suppression processing on the recursively filtered central pixel.

Optionally, the performing dark-area saturation suppression processing on the recursively filtered center pixel includes:

judging whether the component of the central pixel after the recursive filtering is larger than the component of the central pixel before the recursive filtering;

and when the component of the central pixel after filtering is larger than the component of the central pixel before recursive filtering, adopting the component of the central pixel before recursive filtering as the component of the central pixel after recursive filtering.

Optionally, the performing dark-area saturation suppression processing on the recursively filtered center pixel further includes:

respectively subtracting the components of the central pixel after the recursive filtering and the components of the central pixel before the recursive filtering from a preset fifth threshold value to obtain a first difference value and a second difference value;

judging whether the product of the first difference value and the second difference value is less than zero;

and when the product of the first difference value and the second difference value is less than zero, taking the fifth threshold value as the component of the central pixel after the recursive filtering.

Optionally, the fifth threshold is 128.

Optionally, the method further comprises:

when the edge information on the brightness component of the central pixel is smaller than the fourth threshold, judging whether the edge information on the brightness component of the central pixel is larger than a preset first edge threshold;

when the edge information on the brightness component of the central pixel is larger than the first edge threshold, correcting the first filtering strength according to the edge information on the brightness component of the central pixel to obtain a second filtering strength;

and taking the second filtering strength as the filtering strength of the central pixel.

Optionally, the method further comprises:

when the first difference maximum value is larger than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value;

and correcting the second filtering strength according to the first difference maximum value to obtain a third filtering strength, and taking the third filtering strength as the filtering strength of the central pixel.

Optionally, the method further comprises:

and performing denoising and filtering processing on the first row/column pixels of the acquired image block before judging whether the frequency region where the frequency of the central pixel is located is a flat region.

Optionally, when the current brightness of the center pixel is lower than a preset eighth threshold, the corresponding filtering strength decreases as the current brightness decreases.

Optionally, the method further comprises:

and when the frequency region where the central pixel is located is a non-flat region, performing mean filtering on the central pixel.

The embodiment of the invention provides an image denoising device, which comprises:

the image block acquisition unit is suitable for acquiring an image block with a pixel to be processed as a central pixel, and the central pixel is positioned in the kth row and the kth column of the image block;

a first calculation unit adapted to calculate edge information on horizontal and vertical chrominance components of the central pixel, and take a maximum value of the edge information on the horizontal and vertical chrominance components as a first edge information maximum value;

a second calculation unit adapted to calculate a variance of the image block;

a third calculation unit adapted to calculate a maximum difference value between the central pixel and the filtered pixel selected in the (k-1) th row/(k-1) th column in the image block as a first maximum difference value;

a fourth calculation unit adapted to calculate edge information on a luminance component of the center pixel;

the first judgment unit is suitable for judging whether the frequency area where the central pixel is located is a flat area or not according to the first edge information maximum value, the variance of the image block, the first difference maximum value and the edge information on the brightness component of the central pixel;

a fifth calculating unit, adapted to calculate the current brightness of the central pixel when the frequency region where the central pixel is located is a flat region;

the filtering intensity obtaining unit is suitable for obtaining a first filtering intensity according to the current brightness of the central pixel and taking the first filtering intensity as the filtering intensity of the central pixel;

and the filtering unit is suitable for performing recursive filtering on the central pixel by using the filtering strength of the central pixel.

Optionally, the first determining unit is adapted to determine that the frequency of the central pixel is a flat area when all of the following conditions are satisfied:

the variance is smaller than a preset second threshold;

Optionally, values of the first threshold, the second threshold, the third threshold, and the fourth threshold are different according to a luminance interval in which the current luminance of the center pixel is located.

Optionally, the apparatus further comprises:

and the saturation processing unit is suitable for determining that the central pixel is in a dark area when the current brightness of the central pixel is lower than a preset dark area brightness threshold, and performing dark area saturation suppression processing on the recursively filtered central pixel.

Optionally, the saturation processing unit includes:

a first determining subunit, adapted to determine whether the component of the central pixel after the recursive filtering is greater than the component of the central pixel before the recursive filtering;

a first calculation subunit adapted to adopt, when the first judgment subunit determines that the component of the center pixel after the recursive filtering is larger than the component of the center pixel before the recursive filtering, the component of the center pixel before the recursive filtering as the component of the center pixel after the recursive filtering.

Optionally, the saturation processing unit includes:

the second calculating subunit is adapted to respectively take the difference between the component of the central pixel after the recursive filtering and the component of the central pixel before the recursive filtering and a preset fifth threshold to obtain a first difference value and a second difference value;

a second judgment subunit, adapted to judge whether a product of the first difference and the second difference is smaller than zero;

a third calculating subunit adapted to use the fifth threshold as the component of the recursively filtered center pixel when the second determining subunit determines that the product of the first difference and the second difference is less than zero.

Optionally, the fifth threshold is 128.

Optionally, the apparatus further comprises:

a second judging unit, adapted to judge whether the edge information on the luminance component of the center pixel is greater than a preset first edge threshold value when the edge information on the luminance component of the center pixel is less than the fourth threshold value;

and the first correcting unit is suitable for correcting the first filtering strength according to the edge information on the brightness component of the central pixel to obtain a second filtering strength when the edge information on the brightness component of the central pixel is larger than the first edge threshold value, and the second filtering strength is used as the filtering strength of the central pixel.

Optionally, the apparatus further comprises:

a third judging unit adapted to judge whether the following conditions are satisfied: the first difference maximum value is greater than a preset sixth threshold value and less than a preset seventh threshold value, and the seventh threshold value is less than the third threshold value;

and the second correcting unit is suitable for correcting the second filtering strength according to the first difference maximum value to obtain a third filtering strength when the first difference maximum value is larger than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value, and the third filtering strength is used as the filtering strength of the central pixel.

Optionally, the apparatus further comprises:

and the denoising and filtering unit is suitable for performing denoising and filtering processing on the first row/column of pixels of the acquired image block before judging whether the frequency region where the central pixel is located is a flat region.

Optionally, the apparatus further comprises:

and the mean value filtering unit is suitable for carrying out mean value filtering on the central pixel when the frequency region where the central pixel is located is a non-flat region.

Compared with the prior art, the technical scheme of the invention has the following advantages:

when judging whether the frequency region where the central pixel is located is a flat region, not only the first edge information maximum value, the variance of the central pixel and the first difference maximum value are considered, but also the edge information on the brightness component of the central pixel is considered, and the edge information on the brightness component is more accurate, so that the obtained frequency region information of the central pixel can be more accurate, and more accurate filtering intensity can be determined for the central pixel, so that the filtered image can keep more details, and the noise-removing effect is ensured, and meanwhile, the trailing phenomenon of the denoised image is avoided.

Because the sensitivity of human eyes to color differences is different under different brightness, the values of the first threshold, the second threshold, the third threshold and the fourth threshold are correspondingly adjusted according to the brightness interval where the current brightness of the central pixel of the image block is located, so that the color noise in a flat area can be prevented from being misjudged as a color boundary, or the color boundary is misjudged as flat, and the trailing phenomenon of the denoised image can be further avoided.

Further, when the current brightness of the central pixel is lower than a preset dark-area brightness threshold value, the central pixel is determined to be in a dark area, and dark-area saturation suppression processing is performed on the recursively filtered central pixel, so that in a particularly dark place, even if a color boundary is not detected, a tailing phenomenon of the denoised image can be avoided.

Further, when the edge information on the brightness component of the central pixel is greater than the first edge threshold, the first filtering strength is corrected according to the edge information on the brightness component of the central pixel, so that the phenomenon of layering of the filtered image in some areas can be avoided, and the image quality can be further improved.

Further, when the first maximum difference value is greater than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value, the second filtering strength is corrected through the first maximum difference value, so that the phenomenon that the filtered image is layered in some areas can be avoided, and the image quality can be further improved.

Drawings

FIG. 1 is a schematic flowchart of an image denoising method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of an IIR filter of the prior art;

FIG. 3 is a flow chart illustrating another method for denoising an image according to an embodiment of the present invention;

FIG. 4 is an image block with N being 5 and representing the U component of the central pixel according to an embodiment of the present invention;

FIG. 5 is a block diagram of a size 3 × 3 image according to an embodiment of the present invention;

FIG. 6 is another block size of 3 × 3 in an embodiment of the present invention;

FIG. 7 is a block diagram of another 3 × 3 size image according to an embodiment of the present invention;

FIG. 8 is another block size of 3 × 3 in an embodiment of the present invention;

FIG. 9 is a threshold versus luminance graph in an embodiment of the present invention;

FIG. 10 is a graph of filter intensity versus current brightness for a center pixel in an embodiment of the present invention;

FIG. 11 is a graph of edge information edge of the luminance component of a center pixel versus filter strength according to an embodiment of the present invention;

FIG. 12 is a graph of a first variance maximum versus filter strength for an embodiment of the present invention;

FIG. 13 is a schematic structural diagram of an apparatus for denoising an image according to an embodiment of the present invention;

FIG. 14 is a schematic structural diagram of an apparatus for denoising an image according to an embodiment of the present invention;

fig. 15 is a schematic structural diagram of an apparatus for denoising an image according to another embodiment of the present invention.

Detailed Description

In order to filter color noise, currently, when IIR is used, denoising is performed by the following steps: and judging whether the image block is a flat area or not according to the edge information and the variance of the horizontal and vertical directions of the image block and the difference between the central pixel of the image block and the chrominance component of the selected pixel in the previous row or column. When the image block is a flat area, determining the filtering strength of the image block according to the brightness of the image block and the difference between the chrominance components of the central pixel and the peripheral pixels of the image block, and then filtering the image block by using the filtering strength.

However, when the image denoising processing is performed by the above method, if the filtering strength is not well controlled, a phenomenon of smearing may occur while removing noise. Taking color noise as an example, the appearance of the tailing is that the color in the image is dragged downwards, and like a tail is lengthened, the original place without the color also has the color, and the image quality is influenced. Particularly in a dark area, if the filtering strength is too strong, the downward transmissibility of the IIR becomes strong, and then tailing is generated; if the filtering strength is too weak, downward transmissibility becomes weak, and although no tailing is generated, noise is difficult to remove cleanly.

In order to solve the above problem, an embodiment of the present invention provides an image denoising method, where edge information on a luminance component is relatively accurate, and the method determines whether a frequency region where a center pixel is located is a flat region, by considering not only a first edge information maximum value, a variance of the center pixel, and a first difference maximum value, but also edge information on the luminance component of the center pixel, and the edge information on the luminance component is relatively accurate, so that the obtained frequency region information of the center pixel is more accurate, and thus a more accurate filtering strength can be determined for the center pixel, so that a filtered image retains more details, and a tailing phenomenon of the denoised image is avoided while a denoising effect of noise is ensured.

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.

An image denoising method in the embodiment of the present invention is provided below, and as shown in fig. 1, the method is specifically described below with reference to fig. 1:

s11: and acquiring an image block with the pixel to be processed as a central pixel.

In an implementation, since pixels in neighboring areas affect pixel values of each other, an image block having the to-be-processed pixel as a central pixel may be obtained to obtain more accurate information about the central pixel. For convenience of description, the central pixel may be located at the kth row and the kth column of the image block.

In an embodiment of the present invention, a denoising and filtering process may be performed on the first row/column pixels of the obtained image block, and then the filtered image block is used, so as to perform S12. Thus, at a higher filtering intensity a, if the pixels in the first row/column have higher color noise, the color smear on the image boundary will not be caused in the downward transmission process.

S12: calculating edge information on horizontal and vertical chrominance components of the central pixel, and taking the maximum value of the edge information on the horizontal and vertical chrominance components, the variance of the central pixel, the maximum value of the difference between the central pixel and the filtered pixel selected in the k-1 th row or the k-1 th column in the image block, and the edge information on the luminance component of the central pixel.

For convenience of description, the maximum value of the edge information in the horizontal and vertical directions may be referred to as a first edge information maximum value, and the maximum value of the difference between the central pixel and the filtered pixel selected in the (k-1) th row or the (k-1) th column in the image block may be referred to as a first difference maximum value.

In order to allow more preservation of the filtered image details, edge information of the center pixel may be detected. In an embodiment of the present invention, the edge information of the center pixel may be obtained by calculating edge information on horizontal and vertical chrominance components of the center pixel, and taking a maximum value of the edge information on the horizontal and vertical chrominance components, a variance of the center pixel, a maximum value of a difference between the center pixel and a filtered pixel selected in a (k-1) th row/a (k-1) th column in the image block, and the edge information on a luminance component of the center pixel, and according to results obtained by these calculations.

In an embodiment of the present invention, the IIR filter in the vertical direction is selected, so that the selected filtered pixels may be pixels in a row above the central pixel.

In a specific implementation, when it is determined by the determining that the maximum value of the first edge information is smaller than a preset first threshold, the variance is smaller than a preset second threshold, the maximum value of the first difference is smaller than a preset third threshold, and the edge information on the luminance component of the center pixel is smaller than a preset fourth threshold, it may be determined that the frequency of the input center pixel is a flat area.

Due to the different brightness, the human eye is not as sensitive to color differences. The smaller the brightness, the more sensitive the human eye is to small color differences; the greater the brightness, the more sensitive the human eye is to large color differences. Therefore, if the same first to fourth threshold values are set on all the luminances, a true color boundary cannot be detected in a dark area if the threshold value is too high during setting; if the threshold is too low, color noise in a flat area may be misjudged as a color boundary in a bright area. If the color noise of the flat area is misjudged as a color boundary, the color noise is remained in the image; if the color boundary is erroneously judged to be flat, the filtering strength is too high, and the color in the area in the image is smeared. Therefore, in an embodiment of the present invention, values of the first threshold, the second threshold, the third threshold, and the fourth threshold may be set to be different according to a luminance interval in which the current luminance of the central pixel of the image block is located.

S13: and judging whether the frequency region where the central pixel is located is a flat region or not according to the first edge information maximum value, the variance of the image block, the first difference maximum value and the edge information on the brightness component of the central pixel.

In a specific implementation, since the calculated result values may reflect the frequency region of the central pixel to a certain extent, it may be determined whether the frequency region of the central pixel is a flat region according to the first edge information maximum, the variance of the image block, the first difference maximum, and the edge information on the luminance component of the central pixel. The edge information on the brightness component of the central pixel is often more accurate, and the more accurate result of the frequency region can be obtained by taking the edge information as one of the judgment factors of the frequency region of the central pixel.

When the frequency region where the center pixel is located is a flat region, S14 is performed, otherwise, S16 is performed.

S14: the current brightness of the center pixel is calculated.

The color noise distribution over the image blocks is also different for different luminances. In this way, in the implementation, when the frequency region where the central pixel is located is a flat region, the current brightness of the central pixel can be calculated, and thus, different noise processing parameters can be adopted according to different current brightness.

S15: and acquiring a first filtering intensity according to the current brightness of the central pixel, wherein the first filtering intensity is used as the filtering intensity of the central pixel, and the recursive filtering is performed on the central pixel.

Because the brightness can affect the distribution of noise in the image block, and the sensitivity of human eyes to color difference is different under different brightness, a first filtering strength can be obtained according to the current brightness of the central pixel to serve as the filtering strength of the central pixel, and the central pixel is subjected to recursive filtering by using the first filtering strength.

Generally, color noise is more likely to occur in a place with darker brightness, so when the current brightness of the center pixel is lower than a preset eighth threshold, the filtering strength corresponding to the current brightness may be set to decrease as the current brightness decreases. Thereby, the phenomenon of image smearing in a particularly dark area can be avoided.

Due to the influence of noise, if two adjacent pixel points are judged as a flat area, and one is judged as an edge area, the filtered image is layered in some areas, such as the junction of a human face and hair. In order to avoid such a layering phenomenon, in an embodiment of the present invention, when the edge information on the luminance component of the center pixel is smaller than the fourth threshold, it may be determined whether the edge information on the luminance component of the center pixel is larger than a preset first edge threshold, if the edge information on the luminance component of the center pixel is larger than the first edge threshold, the first filtering strength is modified according to the edge information on the luminance component of the center pixel to obtain a second filtering strength, and then the second filtering strength is used as the filtering strength of the center pixel, and the center pixel is filtered according to the second filtering strength.

In order to further suppress the delamination phenomenon, in another embodiment of the present invention, when the first maximum difference value is greater than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value, the second filtering strength may be further modified according to the first maximum difference value to obtain a third filtering strength, the third filtering strength is used as the filtering strength of the central pixel, and then the central pixel is filtered by using the third filtering strength.

In order to prevent the image smearing phenomenon in a dark place, in an embodiment of the present invention, when the current luminance of the central pixel is lower than a preset dark-area luminance threshold, it may be determined that the central pixel is in a dark area, and the dark-area saturation suppression processing may be performed on the recursively filtered central pixel.

Specifically, the first step: whether the component of the central pixel after the recursive filtering is larger than the component of the central pixel before the recursive filtering can be judged, and when the component of the central pixel after the recursive filtering is larger than the component of the central pixel before the recursive filtering, the component of the central pixel before the recursive filtering is adopted as the component of the central pixel after the recursive filtering, so that the saturation of the central pixel after the filtering is not higher than the saturation of the central pixel before the filtering.

The second step: the components of the center pixel after recursive filtering and the components of the center pixel before recursive filtering may be respectively subtracted from a preset fifth threshold to obtain a first difference value and a second difference value, and then it is determined whether a product of the first difference value and the second difference value is smaller than zero, if the product is smaller than zero, the fifth threshold may be used as the components of the center pixel after recursive filtering, and through this step, it may be ensured that the chromaticity of the center pixel after filtering is not reversed. As described above, with the saturation suppression processing of these two steps, even if the color boundary is not detected, the image smear phenomenon does not occur in a particularly dark place.

In an embodiment of the present invention, the component of the central pixel may be a chrominance U component, and the fifth threshold may be 128.

It is understood that, for the chrominance V component and the luminance Y component, similar methods may be used to perform corresponding processing, and details are not described herein.

S16: and performing mean filtering on the central pixel.

In a specific implementation, if the frequency region where the central pixel is located is not a flat region, mean filtering may be performed on the central pixel, so as to avoid damaging color information of the central pixel itself.

It should be noted that the image denoising method according to the present invention is applicable to color noise removal in YUV space, luminance Y noise removal, and other color space such as Lab space, and for color noise removal in YUV space, it is only applicable to chrominance U component denoising and chrominance V component denoising. The IIR filtering direction does not limit the present invention, and those skilled in the art may select the IIR filtering in the vertical direction or the IIR filtering in the horizontal direction according to the actual requirement. For better understanding and realization of the present invention by those skilled in the art, the method for removing noise on the chrominance U component when using IIR filtering in the vertical direction is used as an example to illustrate the implementation step of the image denoising. IIR filtering methods in other directions and noise removing methods on other components may be implemented with reference to the embodiments described herein, and are not described in detail herein.

In general, low frequency color noise requires a large template to remove, and a large template corresponds to complex computation and high hardware cost. However, IIR filtering can simulate the denoising function of a large template and effectively remove low-frequency color noise at the cost of low cost by using the special transfer characteristics of IIR filtering, simple and quick calculation. A typical IIR filter is shown in fig. 2, and if the formula is used to represent the operational relationship shown in fig. 2, then: y (n) denotes a x y (n-1) + (1-a) x (n), x (n) denotes the current original pixel, y (n) denotes the current filtered pixel, y (n-1) denotes the previous filtered pixel, and a denotes the filtering strength. The transmissibility of IIR filtering is represented by: y (N) depends on y (N-1), … y (N-N), i.e., the output of x (N) depends on x (N-1) … … x (N-N), where N < N.

In order to make those skilled in the art better understand and implement the present invention, a flow chart of another image denoising method in the embodiment of the present invention is further provided below, and referring to fig. 3, the method may specifically include the steps of:

s31: and performing noise pre-filtering on the input image block.

Taking the vertical IIR filter as an example, it can be seen from the above description of the IIR filter that in an image, the output of the pixels in the second row depends largely on the values of the pixels in the first row, and the output of the pixels in the third row depends largely on the values of the pixels in the second row and also largely on the values of the pixels in the first row. Therefore, if the pixel value of the first line has large color noise, the larger the filtering strength a is, the more color smear on the image boundary is caused in the downward transfer process. In general, especially in a dark environment, the luminance value on the image boundary is lower, the color noise is larger, if the filtering strength is too small, the color noise cannot be removed, and if the filtering strength is too large, the color smear on the image boundary is more serious.

Therefore, in an embodiment of the present invention, color noise pre-filtering with a template size of 1 × N may be performed on the pixels in the first row, and taking the U component of the central pixel shown in fig. 4 and the image block with N being 5 as an example, the filtering may be performed by using the following formula:

wherein: u0i satisfies: abs (U0i-U02)<thr10, that is, U0i, is a pixel satisfying that the absolute difference from the chrominance U component U02 of the central pixel of the input image block is less than a preset tenth threshold thr10, and

namely, M is: the absolute difference from the chrominance U component U02 of the central pixel of the input image block is less than the number of pixels of the preset tenth threshold thr 10.

The filtered center pixel U02' may then be subjected to saturation suppression using the following formula:

U02″＝U02′*ratio；

wherein, the ratio is a saturation suppression factor, and the larger the ratio is, the weaker the saturation suppression is; conversely, the stronger the saturation suppression.

After simple filtering and saturation suppression, the color noise on the first line of pixels of the image can be weakened a lot, and the color trailing phenomenon on the image boundary can be greatly suppressed.

S32: and carrying out edge detection on the central pixel, and judging whether the central pixel is in a flat area or not.

In order to preserve the details of the filtered image, in a specific implementation, the chrominance U component may be subjected to edge detection, and filtering is performed only where it is detected as a flat area. Specifically, the edge in both the horizontal and vertical directions can be calculated using an n × m template. In an embodiment of the present invention, taking the 3 × 3 template shown in fig. 5 as an example, the central pixel is U11, and the calculation method for calculating the edge information h on the horizontal chrominance component and the edge information v on the vertical chrominance component is as follows in equations (1) and (2):

h＝|U00+U01+U02-U20-U21-U22| (1)

v＝|U00+U10+U20-U02-U12-U22| (2)

then, a maximum value is selected from the two direction values as a first edge information maximum value max _direc Max (h, v), if this first edge information maximum is smaller than the first threshold th1, the center pixel point may be in a flat area.

In an embodiment of the present invention, taking the 3 × 3 template shown in fig. 5 as an example, the variance of the image block is calculated as follows:

if this variance is not less than the second threshold th2, the center pixel may be considered not to be within a flat area.

Since the final filtering result can be calculated by weighted average, and the filtered result of the previous row of pixels can have a great influence on the current row of pixels, in an embodiment of the present invention, the first maximum difference value of the central pixel U11 can be calculated according to the image block shown in fig. 6, where U00 'represents a U00 filtered value, U01' represents a U01 filtered value, and U02 'represents a U02 filtered value, and in an embodiment of the present invention, the upper three points U00', U01 ', and U02' are given greater weights. The first maximum difference value may be calculated using equations (4) to (7), that is:

max_u’＝max(U00’,U01’,U02’) (4)

min_u’＝min(U00’,U01’,U02’) (5)

u _diff1 ＝|max_u’-u11|，u _diff2 ＝|min_u’-u11| (6)

Umax_diff＝max(u _diff1 ,u _diff2 ) (7)

if the first maximum difference Umax _ diff is smaller than the third threshold th3, the central pixel is considered to be possibly in a flat area.

It can be seen that the above three pieces of edge information are extracted in the chrominance U plane, and in general, the edge information in the luminance Y plane is often more accurate and is worth referring. Therefore, in an embodiment of the present invention, edge information on the luminance component of the center pixel of the luminance Y plane can be calculated. Specifically, taking the image block shown in fig. 7 as an example, the luminance component corresponding to the central pixel is Y11.

Before calculating the edge information on the luminance component of the center pixel, in order to reduce the influence of noise on the edge extraction, the 3 × 3 luminance component shown in fig. 7 may be simply filtered, the filtering manner may be various, and the specific filtering manner does not limit the present invention at all, and here, taking the simplest average filtering as an example, the luminance Y component of the center pixel is filtered by using formula (8):

it can be understood that other pixel points may also be filtered by using the same method to obtain the corresponding filtered luminance components, which is not described herein again, and the image block in fig. 7 after being filtered may generate the image block shown in fig. 8, and then calculate the edge information edge on the luminance component of the central pixel according to the image block shown in fig. 8 by using equations (9) to (11):

edge _x ＝|Y00′+Y01′+Y02′-Y20′-Y21′-Y22′| (9)

edge _y ＝|Y00′+Y10′+Y20′-Y02′-Y12′-Y22′| (10)

edge＝edge _x *edge _x +edge _y *edge _y (11)

if the edge information edge on the luminance component of this center pixel is not less than the fourth threshold y _ thr, the center pixel corresponding to the edge information edge is considered not to be in the flat region.

From the above, it can be seen that if the above conditions are all satisfied, the first maximum edge information value max _ direct is<The first threshold th1, the variance u of the image block _var <The second threshold th2, the first maximum difference Umax _ diff<The third threshold th3, edge information edge on the luminance component of the center pixel<The fourth threshold y _ thr may determine that the central pixel corresponding to the four thresholds is in a flat area, and may filter the central pixel.

Due to the different brightness, the human eye is not as sensitive to color differences. The smaller the brightness, the more sensitive the human eye is to smaller color differences; the greater the brightness, the more sensitive the human eye is to large color differences. If the same threshold value is set on all the brightnesses and the threshold value is too high, the real color boundary cannot be detected in a darker area; if the threshold is too low, color noise in a flat area may be misjudged as a color boundary in a bright area. If the color noise of the flat area is misjudged as a color boundary, the color noise remains in the image; if the color boundary is erroneously judged to be flat, the filtering strength is higher, and the color smearing of the area in the image can be caused.

Therefore, in an embodiment of the present invention, in order to solve the color smear in this case, the above-described first to fourth threshold values may be set in accordance with the luminance change. In short, when the brightness is dark, the threshold value can be set to be small to detect the true color boundary; when the brightness is larger, the color noise is smaller, and filtering is not needed, so that the threshold value can be set smaller; when at the intermediate luminance, the threshold value may be set large to prevent color noise of the flat area from being misjudged as a color boundary.

It has been proven through a lot of practice and experiments that, in an embodiment of the present invention, a chromaticity threshold and luminance curve shown in fig. 9 can be used to set the first to fourth threshold values according to the current luminance of the center pixel. The graph in fig. 9 shows a chroma threshold segmentation method, where the chroma threshold is divided into a plurality of segments, and the chroma threshold is also divided into a plurality of segments as the luminance changes, where y _cur Denotes the current brightness of the central pixel, provided for ease of understanding that the brightness y of the central pixel is _cur At the position in FIG. 9, the corresponding chroma threshold u _thr Is calculated as shown in equation (12):

in addition, the parameter Y in fig. 9 _thr1 、Y _thr2 、Y _thr3 、Y _thr4 、u _thr1 And u _thr2 Can be set by a person skilled in the art according to actual needs. The setting of the first to fourth thresholds may be applied to the curve shown in fig. 9, but only when each threshold is selected, the parameters of the horizontal axis and the vertical axis are changed by itself, for example, if the first threshold is set, the values of the vertical axis are all the parameters of the first threshold, and of course, a person skilled in the art may also divide the horizontal axis or the vertical axis of the curve into a plurality of segments according to actual needs, and a specific segmentation method does not limit the present invention. As long as the mapping relationship of the parameter and the brightness can be reflected.

When the center pixel is in a flat region, performing S33; otherwise, S39 is executed.

S33: and calculating the current brightness of the central pixel, and acquiring a first filtering intensity as the filtering intensity of the central pixel according to the current brightness of the central pixel.

Generally, color noise is more likely to occur in a dark place, so in an embodiment of the present invention, the magnitude of the filtering strength is determined by the brightness. Within a certain range, the smaller the brightness, the stronger the intensity; when this range is exceeded, the strength remains unchanged. FIG. 10 is a diagram illustrating a relationship between a filter strength and a brightness according to an embodiment of the present invention, wherein a first filter strength a can be obtained from FIG. 10 according to a current brightness of the center pixel _y The relationship shown in fig. 10 can also be expressed by the following equation (13):

wherein: the luminance parameter y _ th and the filter strength parameters a1 and a2 can be set correspondingly according to actual needs. In order to prevent the image smearing phenomenon, for a particularly dark area, the filtering strength is weakened, that is, when the current brightness of the central pixel is lower than the eighth threshold Y _ min _ th, the corresponding filtering strength is reduced as the current brightness is reduced, specifically, refer to the filtering strength of the segment from 0 to Y _ min _ th in fig. 10.

S34: for the first filtering strength a _y Correcting to obtain a second filtering strength a _ _yy 。

If two adjacent points are adjacent, one is judged to be flat and one is judged to be an edge due to the effect of noise. The filtered result will be layered in some areas, such as the boundary between human face and hair. To avoid this, in an embodiment of the present invention, the first edge threshold y _ thr1 may be set, and the first edge threshold y _ thr1< the fourth threshold y _ thr, for the value of the edge information edge on the luminance component of the center pixel falling within the [ y _ thr1, y _ thr ] interval, the filtering strength may be modified to make the transition between flat and edge more natural.

In an embodiment of the present invention, can utilizeThe first filter strength is corrected by a correction curve shown in fig. 11, thereby obtaining the second filter strength a \u _yy Wherein, a in FIG. 11 _y Is the first filtering strength calculated from the above-described formula according to the current brightness, and the curve relationship in fig. 11 is described by the formula, which may be as (14):

as described above, the setting of the fourth threshold y _ thr may be different according to the difference of the brightness as shown in fig. 9, so similar to the fourth threshold y _ thr, the setting of the first edge threshold y _ thr1 may also be different according to the difference of the brightness, that is, the first edge threshold y _ thr1 may also be set by using the relationship curve shown in fig. 9, only different threshold parameters are set on the vertical axis, and the specific setting method can be implemented with reference to the above description, and will not be described again.

S35: for the second filtering strength a _yy And correcting to obtain third filtering strength a.

In order to further avoid the delamination phenomenon, when the first maximum difference value Umax _ diff ∈ [ th6, th7], that is, the first maximum difference value is greater than the sixth threshold th6 and smaller than the seventh threshold th7, the second filter strength a _ yy may be corrected according to the first maximum difference value by using a correction curve shown in fig. 12 to obtain a third filter strength a, and the third filter strength a is taken as the filter strength of the center pixel, where the curve in fig. 12 may also be described as follows by using equation (15):

as described above, the setting of the third threshold th3 may refer to the curve relationship shown in fig. 9, that is, it may be different according to the different brightness, so that the setting of the sixth threshold th6 and the setting of the seventh threshold th7 may be different according to the different brightness, similar to the third threshold th 3. Similarly, the curves of the sixth threshold th6 and the seventh threshold th7 may be the same as the method for setting the third threshold th3, but different threshold parameters are set on the vertical axis in fig. 9, and are not described herein again.

S36: and carrying out recursive filtering on the pixels in the image according to the third filtering strength a.

In an embodiment of the present invention, taking the image block shown in fig. 6 as an example, a specific formula for calculating the filtered central pixel U11' is shown as (16):

s37: and judging whether the current brightness of the central pixel is lower than the preset dark area brightness threshold value or not.

And executing S38 when the current brightness of the central pixel is lower than the preset dark-area brightness threshold value, otherwise, ending the process.

S38: and carrying out saturation suppression processing on the filtered central pixel.

In order to prevent color tailing in a dark place, in an embodiment of the present invention, when the current brightness of the central pixel is lower than a preset dark area brightness threshold, it may be determined that the central pixel is in a dark area, and saturation of the color may be suppressed, specifically involving two steps:

the first step is as follows: if the saturation of U11 'is higher than U11, U11' is equal to U11, i.e., the saturation after filtering is guaranteed not to be higher than that before filtering;

the second step is that: if (U11 '-th 5) × (U11-th5) <0, then U11' ═ th5, i.e. it is ensured that the filtered chromaticity is not reversed;

wherein: u11' represents the center pixel after filtering, U11 represents the center pixel before filtering, and the fifth threshold th5 is 128.

It should be noted that the descriptions of the first step and the second step are only for illustration and understanding, and do not set any limit to the specific execution sequence of the saturation suppression processing.

With the two restrictions, that is, the saturation suppression processing in the above two steps, in a particularly dark place, no color tailing occurs even if a color boundary is not detected well.

S39: and performing mean filtering on the central pixel.

Taking the image block shown in fig. 5 as an example, assuming that the absolute value of the difference between N points and U11 is smaller than the preset ninth threshold th9, the final filtering process is as described in equation (17):

wherein: uij is that: the relationship of | Uij-U11| < th9, that is, Uij is a pixel satisfying that the absolute value of the difference value from the central pixel U11 is smaller than the ninth threshold th9, and N is the number of pixels satisfying that the absolute value of the difference value from the central pixel U11 is smaller than the ninth threshold th9, N is satisfied, as shown in formula (18):

wherein: abs () represents taking the absolute value of the parameter in parentheses.

In summary, since the edge information on the luminance component is relatively accurate, when determining whether the frequency region where the central pixel is located is a flat region, the frequency region information of the central pixel can be more accurate by considering not only the first edge information maximum value, the variance of the central pixel, and the first difference maximum value, but also the edge information on the luminance component of the central pixel, so that a more accurate filtering strength can be determined for the central pixel, and thus, more details can be retained in the filtered image, and the noise removal effect of the noise is ensured while the tailing phenomenon of the denoised image is avoided.

In order to make the present invention better understood and realized by those skilled in the art, the following also provides an apparatus which can implement the above-mentioned image denoising method, as shown in fig. 13, the apparatus may include: an image block acquiring unit 81, a first calculating unit 82, a second calculating unit 83, a third calculating unit 84, a fourth calculating unit 85, a first judging unit 86, a fifth calculating unit 87, a filtering strength acquiring unit 88 and a filtering unit 89, wherein:

the image block obtaining unit 81 is adapted to obtain an image block with a pixel to be processed as a central pixel, where the central pixel is located in the kth row and the kth column of the image block;

the first calculating unit 82 is adapted to calculate edge information on horizontal and vertical chrominance components of the central pixel, and take a maximum value of the edge information on the horizontal and vertical chrominance components as a first edge information maximum value;

the second calculating unit 83 adapted to calculate the variance of the image block;

the third calculating unit 84 is adapted to calculate a maximum difference value between the central pixel and the selected filtered pixel in the k-1 st row/k-1 st column in the image block as a first maximum difference value;

the fourth calculation unit 85 adapted to calculate edge information on the luminance component of the center pixel;

the first determining unit 86 is adapted to determine whether a frequency region where the central pixel is located is a flat region according to the first edge information maximum value, the variance of the image block, the first difference maximum value, and the edge information on the luminance component of the central pixel;

the fifth calculating unit 87 is adapted to calculate the current brightness of the central pixel when the frequency region where the central pixel is located is a flat region;

the filtering strength obtaining unit 88 is adapted to obtain a first filtering strength as the filtering strength of the central pixel according to the current brightness of the central pixel;

the filtering unit 89 is adapted to perform recursive filtering on the central pixel by using the filtering strength of the central pixel.

In a specific implementation, the first determining unit 86 is adapted to determine that the frequency of the central pixel is a flat region when all of the following conditions are satisfied: the maximum value of the first edge information is smaller than a preset first threshold value; the variance is smaller than a preset second threshold value; the first difference maximum value is smaller than a preset third threshold value; and the edge information on the brightness component of the central pixel is smaller than a preset fourth threshold value.

In a specific implementation, values of the first threshold, the second threshold, the third threshold, and the fourth threshold are different according to a luminance interval in which the current luminance of the center pixel is located.

Fig. 14 shows another image denoising apparatus in the embodiment of the present invention, which includes, in addition to an image block obtaining unit 81, a first calculating unit 82, a second calculating unit 83, a third calculating unit 84, a fourth calculating unit 85, a first judging unit 86, a fifth calculating unit 87, a filtering strength obtaining unit 88, and a filtering unit 89: a saturation processing unit 91, adapted to determine that the central pixel is in a dark area when the current brightness of the central pixel is lower than a preset dark area brightness threshold, and perform dark area saturation suppression processing on the recursively filtered central pixel.

In a specific implementation, the saturation processing unit 91 includes:

a first determining subunit 911, adapted to determine whether the component of the central pixel after the recursive filtering is greater than the component of the central pixel before the recursive filtering;

a first calculating subunit 912 adapted to, when the first judging subunit 911 determines that the component of the center pixel after the filtering is larger than the component of the center pixel before the recursive filtering, adopt the component of the center pixel before the recursive filtering as the component of the center pixel after the recursive filtering.

In a specific implementation, the saturation processing unit 91 further includes:

a second calculating subunit 913, adapted to difference the component of the center pixel after the recursive filtering and the component of the center pixel before the recursive filtering with a preset fifth threshold respectively, so as to obtain a first difference value and a second difference value;

a second determining subunit 914, adapted to determine whether a product of the first difference and the second difference is smaller than zero;

a third calculating subunit 915 adapted to take the fifth threshold value as the component of the recursively filtered center pixel when the second judging subunit 914 determines that the product is less than zero.

In order to make those skilled in the art better understand and implement the present invention, a schematic diagram of another image denoising apparatus in the embodiment of the present invention is provided below, as shown in fig. 15, in addition to the above-mentioned image block obtaining unit 81, first calculating unit 82, second calculating unit 83, third calculating unit 84, fourth calculating unit 85, first judging unit 86, fifth calculating unit 87, filtering strength obtaining unit 88, filtering unit 89, and saturation processing unit 91, the apparatus further includes a second judging unit 101, a first correcting unit 102, a third judging unit 103, a second correcting unit 104, a denoising filtering unit 105, and a mean filtering unit 106, where:

the second determining unit 101 is adapted to determine whether the edge information on the luminance component of the center pixel is greater than a preset first edge threshold value when the edge information on the luminance component of the center pixel is smaller than the fourth threshold value;

the first correcting unit 102 is adapted to correct the first filtering strength according to the edge information on the luminance component of the center pixel to obtain a second filtering strength when the edge information on the luminance component of the center pixel is greater than the first edge threshold, and use the second filtering strength as the filtering strength of the center pixel.

In a specific implementation, the third determining unit 103 is adapted to determine whether the following conditions are satisfied: the first maximum difference value is larger than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value;

the second correcting unit 104 is adapted to correct the second filtering strength according to the first difference maximum value when the first difference maximum value is greater than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value, so as to obtain a third filtering strength, and use the third filtering strength as the filtering strength of the central pixel.

In a specific implementation, the denoising and filtering unit 105 is adapted to perform denoising and filtering processing on a first row/column of pixels of the acquired image block before determining whether a frequency region where the central pixel is located is a flat region or not

In a specific implementation, when the current brightness of the central pixel is lower than a preset eighth threshold, the corresponding filtering strength decreases as the current brightness decreases.

In a specific implementation, the mean filtering unit 106 is adapted to mean filter the central pixel when the frequency region where the central pixel is located is not a flat region.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.

Although the present invention is disclosed above, the present invention is not limited thereto. Various changes and modifications may be effected therein by one skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for denoising an image is characterized in that recursive filtering is performed on pixels in the image, and the filtering intensity of a current pixel to be processed is obtained by adopting the following method, including:

calculating the variance of the image block;

calculating edge information on a luminance component of the center pixel;

acquiring a first filtering intensity according to the current brightness of the central pixel, wherein the first filtering intensity is used as the filtering intensity of the central pixel;

the image denoising method further comprises the following steps:

when the current brightness of the central pixel is lower than a preset dark-area brightness threshold value, determining that the central pixel is in a dark area, and performing dark-area saturation suppression processing on the recursively filtered central pixel;

the performing dark-area saturation suppression processing on the recursively filtered center pixel includes:

respectively subtracting the components of the central pixel after the recursive filtering and the components of the central pixel before the recursive filtering from a preset fifth threshold value to obtain a first difference value and a second difference value; judging whether the product of the first difference value and the second difference value is less than zero; and when the product of the first difference value and the second difference value is less than zero, taking the fifth threshold value as the component of the central pixel after the recursive filtering.

2. The method of denoising the image according to claim 1, wherein the determining whether the frequency region of the central pixel is a flat region according to the first maximum edge information, the variance of the image block, the first maximum difference, and the edge information on the luminance component of the central pixel comprises:

the variance is smaller than a preset second threshold;

3. The method of image denoising as claimed in claim 2, wherein values of the first threshold, the second threshold, the third threshold and the fourth threshold are different according to a luminance interval in which a current luminance of a center pixel of the image block is located.

4. The method of denoising the image according to claim 2, wherein the fifth threshold is 128.

5. The method of image denoising as claimed in claim 2, further comprising:

6. The method of image denoising as claimed in claim 5, further comprising:

when the first difference maximum value is greater than a preset sixth threshold value and less than a preset seventh threshold value, and the seventh threshold value is less than the third threshold value;

7. The method of image denoising as claimed in claim 1, further comprising:

8. The method of denoising the image according to claim 1, wherein when the current brightness of the central pixel is lower than a preset eighth threshold, the corresponding filtering strength decreases as the current brightness decreases.

9. The method of image denoising as claimed in claim 1, further comprising:

10. An apparatus for denoising an image, comprising:

a second calculation unit adapted to calculate a variance of the image block;

a fourth calculation unit adapted to calculate edge information on the luminance component of the center pixel;

a first judging unit, adapted to judge whether a frequency region where the central pixel is located is a flat region according to the first edge information maximum value, the variance of the image block, the first difference maximum value and edge information on the luminance component of the central pixel;

a filtering unit adapted to recursively filter the center pixel using the filtering intensity of the center pixel;

the image denoising device further comprises:

a saturation processing unit, adapted to determine that the central pixel is in a dark area when the current brightness of the central pixel is lower than a preset dark area brightness threshold, and perform dark area saturation suppression processing on the recursively filtered central pixel;

the saturation processing unit includes: the second calculating subunit is adapted to respectively take the difference between the component of the central pixel after the recursive filtering and the component of the central pixel before the recursive filtering and a preset fifth threshold to obtain a first difference value and a second difference value; a second judgment subunit, adapted to judge whether a product of the first difference and the second difference is smaller than zero; a third calculating subunit adapted to, when the second judging subunit determines that the product of the first difference and the second difference is smaller than zero, take the fifth threshold as the component of the recursively filtered center pixel.

11. The apparatus for denoising the image according to claim 10, wherein the first determining unit is adapted to determine the frequency of the central pixel as a flat region when all of the following conditions are satisfied:

the variance is smaller than a preset second threshold value;

12. The apparatus for denoising the image according to claim 11, wherein values of the first threshold, the second threshold, the third threshold and the fourth threshold are different according to a luminance interval in which a current luminance of the central pixel is located.

13. The apparatus for denoising an image according to claim 11, wherein the fifth threshold is 128.

14. The apparatus for denoising an image according to claim 11, further comprising:

15. The apparatus for denoising an image according to claim 14, further comprising:

a third judging unit adapted to judge whether the following condition is satisfied: the first difference maximum value is greater than a preset sixth threshold value and less than a preset seventh threshold value, and the seventh threshold value is less than the third threshold value;

and the second correcting unit is suitable for correcting the second filtering strength according to the first difference maximum value when the first difference maximum value is larger than a preset sixth threshold value and smaller than a preset seventh threshold value, and the seventh threshold value is smaller than the third threshold value, so as to obtain a third filtering strength, and the third filtering strength is used as the filtering strength of the central pixel.

16. The apparatus for denoising an image according to claim 10, further comprising:

and the denoising and filtering unit is suitable for performing denoising and filtering processing on the first row/column pixels of the acquired image block before judging whether the frequency region where the central pixel is located is a flat region.

17. The apparatus for denoising the image according to claim 10, wherein when the current luminance of the central pixel is lower than a preset eighth threshold, the corresponding filtering strength decreases as the current luminance decreases.

18. The apparatus for denoising an image according to claim 10, further comprising: