JP2008028447A - Method and apparatus for reducing noise in image signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and an apparatus for reducing a noise in an image signal by which the noise in the image signal is reduced so as to produce a color image with a high resolution. <P>SOLUTION: The method of reducing the noise in the image signal from a solid-state imaging element with a color pixel arrangement is provided. This method includes: a first step of selecting a noise reduction object pixel at the prescribed position of the color pixel arrangement as a center pixel and selecting a plurality of pixels around the center pixel as surrounding pixels; a second step of calculating the correction pixel value of the center pixel on the basis of the pixel value of the center pixel and respective pixel values of the surrounding pixel values; and a third step of replacing the original pixel value of the center pixel with the correction pixel value. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to image signal processing, and more particularly, to a noise reduction method and apparatus for color image signals from a solid-state image sensor.

  Conventionally, in an imaging apparatus such as a video camera or a digital camera, a solid-state imaging device such as a CCD image sensor or a CMOS image sensor is generally used. In such an imaging apparatus, in order to perform color imaging, generally, a method of providing a color filter on a solid-state imaging device and obtaining a color image signal by signal processing is widely used.

  However, the color image signal obtained in this way contains a large number of various noises resulting from the structure of the solid-state imaging device. In particular, in a CMOS image sensor having a transistor for each pixel, the transistor characteristics vary and there is a lot of noise. Therefore, when using a solid-state image sensor as an image sensor, it is important to reduce noise.

  Noise is roughly divided into fixed pattern noise (FPN) due to variations in transistor characteristics for each pixel as described above, and random noise (RN) due to irregular fluctuation components included in image signals. There is.

  As a general countermeasure against these noises, a method of passing an image signal through a low pass filter (LPF) is often employed. This is a method of uniformly cutting a component having a frequency equal to or higher than a predetermined frequency set as a cut-off frequency with respect to a high frequency component of an image signal containing the most noise.

  However, in the conventional noise reduction method using the LPF as it is, the high frequency component of the image signal without noise is naturally reduced. Since the high-frequency component of the image signal is related to the sharpness of the image, the sharpness of the image decreases as the noise is removed by the LPF. In other words, the LPF is effective for fixed pattern noise generated in the flat portion of the image, but for fixed pattern noise generated in the edge portion of the image, the data and the noise are uniformly distinguished. The image is cut and the image outline is smoothed. As a result, the sharpness of the image is reduced, resulting in a blurred image. Therefore, the noise reduction method using the LPF uniformly has a drawback of reducing the resolution of the image.

  Since solid-state imaging devices are easily affected by fixed pattern noise, reduction of this fixed pattern noise becomes a major issue. In the method using the conventional LPF uniformly, the cutoff frequency is adjusted to make a trade-off between the noise reduction and the resolution reduction. Therefore, a more effective noise reduction method that can achieve both low noise and high resolution is desired.

  An object of the present invention is to provide an image signal noise reduction method and apparatus capable of reducing noise of an image signal and generating a high-resolution color image.

  In order to achieve the above object, the present invention provides a method for reducing noise of an image signal in a solid-state imaging device having a color pixel array. This method includes a first step of selecting a noise reduction target pixel at a predetermined position of the color pixel array as a central pixel, and selecting a plurality of pixels around the central pixel as peripheral pixels, and the pixel of the central pixel A second step of calculating a correction pixel value of the central pixel from the pixel value of each of the plurality of surrounding pixels and a third step of replacing the original pixel value of the central pixel with the correction pixel value Including.

  In the second step, by adding a value obtained by multiplying a pixel value of the central pixel by a first coefficient and a value obtained by multiplying an average value of pixel values of the plurality of surrounding pixels by a second coefficient. The correction pixel value is obtained. The first coefficient and the second coefficient are determined for each central pixel based on a pixel value of the central pixel and pixel values of the plurality of surrounding pixels.

  Preferably, in the second step, the plurality of surrounding pixels include a first surrounding pixel group and a second surrounding pixel group, and a value obtained by multiplying a pixel value of the center pixel by a first coefficient; , A value obtained by multiplying the average value of the pixel values of the pixels in the first surrounding pixel group by a second coefficient, and a value obtained by multiplying the average value of the pixel values of the pixels in the second surrounding pixel group by a third coefficient. The corrected pixel value is obtained by adding the value. The first coefficient, the second coefficient, and the third coefficient are the pixel value of the central pixel, the pixel value of the pixel in the first surrounding pixel group, and the pixel of the pixel in the second surrounding pixel group. Depending on the value, it is determined for each central pixel.

  The color pixel array is a Bayer array. The solid-state imaging device includes a CMOS image sensor.

  The present invention also provides a noise reduction device for color image signals. The apparatus includes: a solid-state imaging device having a color pixel array; a signal processing unit that processes a color image signal from the solid-state imaging device; and an output unit that outputs the color image signal processed by the signal processing unit. Have. When the signal processing unit reduces noise of the color image signal, first, the noise reduction target pixel at a predetermined position of the color pixel array is selected as a central pixel, and a plurality of pixels around the central pixel are selected. Selecting as a surrounding pixel, and then calculating a correction pixel value of the center pixel from the pixel value of the center pixel and the respective pixel values of the plurality of surrounding pixels, and finally, the original pixel of the center pixel Replace the value with the corrected pixel value.

  The present invention maintains the edge information of the image while deleting the noise, thereby making it possible to achieve both a low noise and a high resolution, and to generate a high-quality color image.

  Next, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Embodiment 1
FIG. 1 is a color pixel array of the solid-state imaging device according to the present embodiment.

  In FIG. 1, G represents a green pixel, R represents a red pixel, and B represents a blue pixel, and the subscript indicates the two-dimensional position of the pixel.

  In the color pixel array as shown in FIG. 1, the number of green pixels is substantially N / 2 and the number of red and blue pixels is substantially N / 4 with respect to the total number of pixels N. Such a color pixel array is called a Bayer color pixel array.

  Here, the reason why the green pixels are doubled is that the spectral sensitivity of the human eye has a peak around green, and the green resolution improves the apparent resolution. Each pixel has a value as image data, and a noise component is included in these values.

でＧ１６を置き換える。 In a general noise reduction method, processing is performed in which data of each pixel including a noise component is replaced with an average value of adjacent pixels. For example, in the five green pixels G10, G12, G16, G20, and G22 in FIG. 1, for the central pixel G16 from which noise is to be removed, the upper left pixel G10, the upper right pixel G12, A value obtained by Equation 1 from the values of the lower left pixel G20 and the lower right pixel G22.
[Outside 1]

To replace G16.

図２及び図３は、この平均値で置き換える方法を概念的に表したものである。

2 and 3 conceptually show a method of replacing with this average value.

  2 and 3, the horizontal axis indicates the arrangement of pixels, and the vertical axis indicates the pixel value (luminance). The small zigzag irregularities in the figure represent noise, and the stepped portions represent edges.

  An edge is a portion that touches the outline or edge of an image, and the formation of the edge is related to a high-frequency component of image data. The greater the difference in the height of the edge portion, the higher the resolution of the image and the sharper the image. On the contrary, if the height difference of the edge portion is small and gentle, the resolution becomes low, the sharpness of the image is lost, and the image becomes blurred. It is a so-called dull picture.

  If there is no edge around the noise as shown in FIG. 2, the method of replacing with the average value is effective. However, since noise is generated uniformly in any part, the noise is close to the edge as shown in FIG. In some cases, as noise is removed, edges are also smoothed. As a result, the sharpness of the image is significantly reduced, resulting in a dull image.

  When noise reduction processing using a simple moving average such as a general low-pass filter is performed, the high frequency components of the image are lost, the resolution is lowered, and the sharpness of the image is lost.

  Therefore, in this embodiment, the following formula 2 is used instead of formula 1.

ここで、α、βは、０≦β≦α≦１及びα≠０の条件を満たす任意の係数である。係数α、βは、数式２を用いたノイズ削減において、周辺の画素値と中心の画素値のどちらに重み付けをして平滑化するかという加重平均の分配比率を表現している。例えば、ノイズ削減を全く行わない状態は、α＝１、β＝０である。周辺画素の平均値が中心画素の値を超えてしまうことがないように、β≦αとしてある。αとβの値を周辺画素の値により、適応的に変化させていくことで、画像のあらゆる部分においてエッジを保ちながらローパスフィルタ処理を施していく。なお、αとβの比β／αを平滑度として定義しておく。

Here, α and β are arbitrary coefficients that satisfy the conditions of 0 ≦ β ≦ α ≦ 1 and α ≠ 0. The coefficients α and β represent a weighted average distribution ratio indicating which of the peripheral pixel value and the central pixel value is weighted and smoothed in the noise reduction using Expression 2. For example, a state where no noise reduction is performed is α = 1 and β = 0. Β ≦ α is set so that the average value of the peripheral pixels does not exceed the value of the central pixel. By adaptively changing the values of α and β according to the values of surrounding pixels, low-pass filter processing is performed while maintaining edges in every part of the image. The ratio β / α between α and β is defined as smoothness.

  In addition, among the central pixel value and the peripheral pixel values, the largest value is Gmax, the smallest value is Gmin, and the difference ΔG is defined as the sharpness, as shown in Equation 3.

ノイズ削減処理は、具体的には以下の通りである。

Specifically, the noise reduction processing is as follows.

(1) When edge information is not included in five pixels of interest Normal LPF processing is performed. That is, when the edge information is not included (that is, the value of the sharpness ΔG is low), the value of the center pixel and the average value of the surrounding pixels are made flat (that is, α: β = 1: 1). To eliminate noise.

(2) When edge information is included in five pixels of interest Depending on the state of the edge, the LPF processing effect is weakened or LPF processing is not performed. That is, as the sharpness ΔG increases, the smoothness β / α is decreased (that is, the value of β with respect to α is decreased). At this time, the noise removal effect decreases, but the edge retention effect increases accordingly. Further, when the sharpness ΔG is larger than a certain value, β = 0 and no noise reduction processing is performed.

  FIG. 4 is a diagram showing the relationship between the sharpness ΔG and the smoothness β / α.

  Experiments show that it is effective to change the sharpness ΔG and the smoothness β / α linearly as shown in FIG. Furthermore, it has been experimentally known that setting the value of the sharpness ΔG with β = 0 as the threshold Th and performing the low-pass filter processing when the threshold Th is exceeded results in better noise reduction processing. .

  This method is effective because humans have the visual characteristic that noise tends to stand out more in a uniform area than in a fine area of image information.

  FIG. 5 is a conceptual diagram showing the necessity of normalization based on brightness of sharpness.

  A solid-state imaging device has a property that noise is likely to occur in a dark portion of an image. In consideration of this, even if the sharpness is the same, it is judged whether the sharpness is at a high position (bright part) or low position (dark part) as seen from the whole pixel level. Normalization of sharpness is performed. When the sharpness is in the bright part of the image, the noise generation rate is low, so the edge retention is prioritized over noise removal. When it is in the dark part, the noise generation rate is high, so the noise removal is prioritized over edge retention. .

  As one specific method of normalization, a value obtained by multiplying the sharpness ΔG by the average value of the five pixels of interest is used as a normalized value. For example, among the five pixel values of G10, G12, G16, G20 and G22 in FIG. 1, the value ΔG16 · Ga16 obtained by multiplying the sharpness ΔG16 obtained by subtracting the minimum value from the maximum value and the average value Ga16 of five pixels is normalized. Used as the value of sharpness. Thereby, the brightness is reflected on the sharpness.

  FIG. 6 is a diagram showing the relationship between the smoothness β / α and the luminance.

  This indicates that it is better to increase the edge retention effect by lowering the low-pass filter processing effect in the higher luminance portion. Experiments have shown that better results can be obtained with linear changes.

  As described above, the calculation of Equation 2 is repeatedly performed for all the green pixels of the image data while changing the values of α and β in a timely manner, thereby reducing noise while retaining the edge information of the image. A color image signal is obtained.

  Further, when the solid-state imaging device is a CMOS sensor, there is a structural defect that a luminance difference is generated between a green pixel adjacent to the red pixel and a green pixel adjacent to the blue pixel. The original process also has the advantage that this defect can be corrected automatically. This processing is effective regardless of the type of noise such as fixed pattern noise or random noise.

  In the present embodiment, among the three primary color pixel (red, green, and blue) arrays on the solid-state imaging device, the green pixel having the largest noise reduction effect and the luminance difference correction effect in the CMOS sensor has been described. This operation can also be performed for red and blue pixels.

  FIG. 7 shows another color pixel arrangement on the solid-state imaging device according to the preferred embodiment of the present invention.

  In FIG. 7, G represents a green pixel, R represents a red pixel, and B represents a blue pixel, and the subscript indicates the two-dimensional position of the pixel. In FIG. 7, when R37 is set as the central pixel by paying attention to the red pixel, when the peripheral pixels are applied to Expression 2 as R17, R35, R39, and R57, Expression 4 is obtained.

同様に青画素に着目して、

Similarly, paying attention to the blue pixel,

となる。以降、緑画素のときと同様の処理を行う。

It becomes. Thereafter, the same processing as that for the green pixel is performed.

  Therefore, the image signal noise reduction method according to the present embodiment is specifically executed as follows.

を計算する。そして、画素Iの補正画素値
[外３]

で画素Iの元の画素値Ｇ_Iを置き換える。また、このような操作を全ての画素に対して繰り返し行う。 First, the pixel values of all the pixels are acquired and stored. Then, predetermined pixels I of the pixel values G _I and the predetermined plurality of pixels located around it (e.g., the center pixel G16 shown in FIG. 1 surrounding pixels G10, G12, G20 and G22) pixel values of Then, the sharpness ΔG is calculated and compared with a predetermined sharpness threshold Th. If ΔG is smaller than Th, noise reduction processing is performed. In this case, the coefficient α is determined by the relationship between the smoothness β / α and the sharpness ΔG shown in FIG. 4 and the relationship between the smoothness β / α and the brightness of the entire image shown in FIG. The value of β is determined, and the corrected pixel value of pixel I according to Equation 2
[Outside 2]

Calculate And the corrected pixel value of pixel I
[Outside 3]

In replace the original pixel value G _I of the pixel I. Such an operation is repeated for all pixels.

  FIG. 8 is a configuration diagram of the imaging apparatus according to the present embodiment.

  As illustrated in FIG. 8, the imaging apparatus includes an imaging device 1 having a color pixel array such as a Bayer array, an AD converter 2, an image data preprocessing unit that performs image data preprocessing 3, and noise reduction processing 4. A processing unit (not shown) including a noise reduction processing unit for performing image processing, a Bayer interpolation processing unit for performing Bayer interpolation processing 5 and an image data post-processing unit for performing image data post-processing 6, and image data processed by the processing unit Output unit (not shown).

  The image sensor 1 converts the received light into an electrical signal to obtain analog optical image data. The AD converter 2 converts analog optical image data obtained by the image sensor into digital image data. Image data preprocessing 3 such as black level adjustment is performed on the digital signal data converted by the AD converter 2, and then noise reduction processing 4 by the above-described noise reduction method is performed. Thereafter, a Bayer interpolation process 5 is performed after the noise reduction process, and various image data post-processes 6 are performed. Finally, the image data processed by the processing unit is output by the output unit.

Embodiment 2
In addition, the number of pixels used when calculating the correction pixel value of the central pixel using the above-described formula 2 is not limited to five, and may be three or seven, and the arrangement is also uniform. It is not limited to. In addition, it is possible to use separate pixel arrays for edge detection and low-pass filter processing, and edge detection is performed with different colors among the three primary color pixel (red, green, blue) arrays, It is also possible to perform low-pass filter processing calculations with different colors.

  In the present embodiment, an array of 13 pixels (center 1 pixel + around 12 pixels) is used for edge detection, and an array of 9 pixels (center 1 pixel + around 8 pixels) is used for low-pass filter processing. Furthermore, in the noise reduction processing of R (red) red pixels and B (blue) pixels, an array of G (green) pixels is used for detection.

  FIG. 9 is a color pixel array of the solid-state imaging device according to the present embodiment.

  The color image array in FIG. 9 is a pixel array for edge detection when G (green) pixel noise is reduced. A part HF (for example, G16, G18, G22, G26, G28) for detecting a high frequency (High Frequency) edge and a part MF for detecting an edge of a medium frequency (Middle Frequency (MF)) while increasing the number of detection pixels. (For example, G10, G12, G14, G20, G24, G30, G32, G34). Along with this, the sharpness becomes the sharpness ΔGH derived from the high frequency part and the sharpness ΔGM derived from the medium frequency part as follows.

図１０は、本実施形態に係る固体撮像素子の他のカラー画素配列である。

FIG. 10 shows another color pixel arrangement of the solid-state imaging device according to the present embodiment.

  The color image array in FIG. 10 is a pixel array that performs low-pass filter processing of G (green) pixels. As apparent from the comparison with FIG. 9, the pixel arrangement for edge detection is different from the arrangement.

  In this case, Expression 2 is expressed as Expression 7 below.

ここで、α、β、γは、０≦γ≦β≦α≦１及びα≠０の条件を満たす任意の係数である。精鋭度ΔＧに対する閾値ＴｈもＴｈ１とＴｈ２（ただし、Ｔｈ１＞Ｔｈ２とする）を設定し、以下のようにノイズ削減処理を行う。

Here, α, β, and γ are arbitrary coefficients that satisfy the conditions of 0 ≦ γ ≦ β ≦ α ≦ 1 and α ≠ 0. The threshold Th for the sharpness ΔG is also set to Th1 and Th2 (where Th1> Th2), and noise reduction processing is performed as follows.

(1) When ΔGH ≧ Th1 It is determined that important edge information is included, the edge holding is prioritized, and the low-pass filter process is not performed. That is, α = 1, β = 0, and γ = 0.

(2) When ΔGH <Th1 and ΔGM ≧ Th2 There is no edge information in the HF detection unit (immediately adjacent to the central pixel), but it is determined that there is edge information in the MF detection unit (around the central pixel). Then, weak low-pass filter processing is performed. That is, processing is performed with γ = 0 and the values of α and β are adaptively changed as described above.

(3) When ΔGH <Th1 and ΔGM <Th2 It is determined that there is no edge information and the image is a flat portion (low frequency portion), and strong LPF processing is performed. In other words, the three coefficients α, β, and γ are processed by changing the values adaptively as described above.

  A similar operation can be performed for red pixels and blue pixels.

  FIG. 11 shows another color pixel arrangement of the solid-state imaging device according to the present embodiment.

  FIG. 11 shows a case where the same operation is applied to the operation performed paying attention to the blue pixel in FIG. Specifically, R37 is set as the central pixel, and R15, R17, R19, R35, R39, R55, R57, and R59 are applied to the peripheral pixels, and the following formula 8 is applied to formula 2.

ここで、エッジ検出用の画素配列として緑画素の配列を用いて、ＨＦ部にＧ２７、Ｇ３６、Ｇ３８、Ｇ４７を、ＭＦ部にＧ１６、Ｇ１８、Ｇ２５、Ｇ２９、Ｇ４５、Ｇ４９、Ｇ５６、Ｇ５８を設定する。検出用に緑画素を用いる理由は、エッジ検出は輝度の変化を見ており、一番輝度の大きな緑画素の変化を検出した方が、精度が向上するからである。

Here, G27, G36, G38, and G47 are set in the HF section, and G16, G18, G25, G29, G45, G49, G56, and G58 are set in the MF section by using the green pixel array as the pixel array for edge detection. To do. The reason for using a green pixel for detection is that edge detection looks at a change in luminance, and the accuracy is improved when the change in the green pixel having the highest luminance is detected.

  FIG. 12 shows another color pixel arrangement of the solid-state imaging device according to the present embodiment.

  FIG. 12 similarly shows a case where attention is paid to blue pixels. B42 is set as the central pixel, and B20, B22, B24, B40, B44, B60, B62, and B64 are applied to the peripheral pixels, and the following formula 9 is applied to formula 2.

エッジ検出用の画素配列には、同様にして緑画素の配列を用いて、ＨＦ部にＧ３２、Ｇ４１、Ｇ４３、Ｇ５２を、ＭＦ部にＧ２１、Ｇ２３、Ｇ３０、Ｇ３４、Ｇ５０、Ｇ５４、Ｇ６１、Ｇ６３を設定する。以降、緑画素のときと同様の処理を行う。

Similarly, for the pixel array for edge detection, an array of green pixels is used, and G32, G41, G43, and G52 are used for the HF part, and G21, G23, G30, G34, G50, G54, G61, and G63 are used for the MF part. Set. Thereafter, the same processing as that for the green pixel is performed.

  In the method according to the present embodiment, the edge detection area is expanded, and the LPF processing is divided into two levels of intensity, so that the conflicting processes of the edge holding process and the low-pass filter process are simultaneously performed in a more balanced and more effective manner. Is possible.

  Note that the imaging apparatus according to the present embodiment is the same as the imaging apparatus according to the first embodiment.

  The preferred embodiment of the present invention has been described above, but the present invention is not limited to this embodiment, and all modifications to the present invention are within the scope of the present invention unless departing from the spirit of the present invention.

It is a color pixel arrangement | sequence of the solid-state image sensor which concerns on suitable embodiment of this invention. It is a conceptual diagram which shows the state which applied the low-pass filter to the state where sharpness is low. The vertical axis direction is the pixel value (luminance), and the horizontal axis direction is the pixel value distribution. It is a conceptual diagram which shows the state which applied the low-pass filter to the state with high sharpness. The vertical axis direction is the pixel value (luminance), and the horizontal axis direction is the pixel value distribution. The step portion in the figure represents the edge of the image. It is a figure which shows the relationship between sharpness and smoothness. It is a conceptual diagram which shows the necessity for the normalization by the brightness | luminance of sharpness. It is a figure which shows the relationship between the brightness | luminance of the whole image, and smoothness. It is another color pixel arrangement | sequence of the solid-state image sensor which concerns on suitable embodiment of this invention. It is a block diagram of the imaging device of this invention. It is a color pixel arrangement | sequence of the solid-state image sensor which concerns on other preferable embodiment of this invention. It is a color pixel arrangement | sequence of the solid-state image sensor which concerns on other preferable embodiment of this invention. It is a color pixel arrangement | sequence of the solid-state image sensor which concerns on other preferable embodiment of this invention. It is a color pixel arrangement | sequence of the solid-state image sensor which concerns on other preferable embodiment of this invention.

Explanation of symbols

G Green pixel R Red pixel B Blue pixel β / α Smoothness ΔG Sharpness
Th Threshold of sharpness 1 Image sensor 2 AD converter 3 Image data preprocessing 4 Noise reduction processing 5 Bayer interpolation processing 6 Image data postprocessing

Claims (8)

A method for reducing noise of an image signal in a solid-state imaging device having a color pixel array,
A first step of selecting a noise reduction target pixel at a predetermined position of the color pixel array as a central pixel and selecting a plurality of pixels around the central pixel as peripheral pixels;
A second step of calculating a correction pixel value of the central pixel from a pixel value of the central pixel and a pixel value of each of the plurality of surrounding pixels;
A third step of replacing the original pixel value of the central pixel with the corrected pixel value;
including,
A noise reduction method for color image signals. In the second step, a value obtained by multiplying a pixel value of the central pixel by a first coefficient and a value obtained by multiplying an average value of pixel values of the plurality of surrounding pixels by a second coefficient are added, To obtain a corrected pixel value,
The color image signal noise reduction method according to claim 1. The first coefficient and the second coefficient are determined for each central pixel by a pixel value of the central pixel and pixel values of the plurality of surrounding pixels.
The color image signal noise reduction method according to claim 2. In the second step,
The plurality of surrounding pixels have a first surrounding pixel group and a second surrounding pixel group;
A value obtained by multiplying a pixel value of the central pixel by a first coefficient, a value obtained by multiplying an average value of pixel values of pixels in the first surrounding pixel group by a second coefficient, and the second surrounding pixel group Adding a value obtained by multiplying the average value of the pixel values of the pixels by a third coefficient to obtain the corrected pixel value;
The color image signal noise reduction method according to claim 1. The first coefficient, the second coefficient, and the third coefficient are the pixel value of the central pixel, the pixel value of the pixel in the first surrounding pixel group, and the pixel of the pixel in the second surrounding pixel group. Depending on the value, it is determined for each central pixel,
The noise reduction method for a color image signal according to claim 4. The color pixel array is a Bayer array.
The color image signal noise reduction method according to claim 1. The solid-state imaging device includes a CMOS image sensor,
The color image signal noise reduction method according to claim 1. A solid-state imaging device having a color pixel array;
A signal processing unit for processing a color image signal from the solid-state imaging device;
An output unit capable of outputting the color image signal processed by the signal processing unit and reducing noise of the color image signal;
Have
When the signal processing unit reduces noise of the color image signal,
Selecting a noise reduction target pixel at a predetermined position of the color pixel array as a central pixel, and selecting a plurality of pixels around the central pixel as peripheral pixels;
Based on the pixel value of the central pixel and the pixel values of the plurality of surrounding pixels, the correction pixel value of the central pixel is calculated,
Replacing the original pixel value of the central pixel with the corrected pixel value;
Noise reduction device for color image signals.

Images