KR100989760B1 - Apparatus For Processing Image Siganl, Method For Reducing Noise Of Image Signal Processing Apparatus And Recorded Medium For Performing Method Of Reducing Noise - Google Patents

Abstract

Disclosed are an image processing apparatus capable of effectively removing noise included in an image, a recording medium recording a noise removing method of the image processing apparatus, and a noise removing method. The image processing apparatus receives Bayer image data from an image sensor, obtains an image boundary value based on pixel values of the same pixel components among pixel values included in the Bayer image data, and the image boundary value is larger than the preset area boundary value. When the color is small based on the Bayer noise canceling unit provided from the Bayer noise canceling unit and the Bayer noise canceling unit which outputs a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as the Bayer noise removing unit. It includes a color interpolation unit for performing interpolation. Therefore, the edge region and the region containing noisy can be distinguished well, and thus, the noise included in the Bayer image can be effectively removed to improve the output image quality.

Image, Noise, Bayer, CSP, ISP

Description

Apparatus For Processing Image Siganl, Method For Reducing Noise Of Image Signal Processing Apparatus And Recorded Medium For Performing Method Of Reducing Noise}

The present invention relates to a signal processing apparatus, and more particularly, to a recording medium in which an image processing apparatus, a noise removing method and a noise removing method of an image processing apparatus which can be applied to a signal processing apparatus for processing a captured image are recorded. .

Image sensors such as Charge Coupled Device (CCD) or Complementary Metal Oxide Semiconductor (CMOS) are composed of two-dimensional integrated pixels and convert electrical signals corresponding to the brightness of incident light into digital signals. To print. Here, the image sensor may be configured in a Bayer pattern, and provides Bayer image data corresponding to brightness of light incident on the Bayer pattern.

Batter image data provided from an image sensor is output as a luminance signal and a color difference signal through various signal processing such as color interpolation, luminance processing, color processing, color format conversion, and the like through an image signal processing apparatus.

1 is a block diagram showing the configuration of a general image signal processing apparatus.

Referring to FIG. 1, a general image signal processing apparatus includes a raw data processor 10, a color interpolator 20, a color corrector 30, a gamma corrector 40, and a format converter 50.

The raw data processing unit 10 receives raw data (RAW), that is, Bayer image data of a Bayer pattern, from an image sensor, and shading caused by defect correction or short focal length of a lens present in the image sensor. Perform preprocessing such as shading correction.

The color interpolator 20 receives data preprocessed by the raw data processor 10, and interpolates each pixel component having one channel component through R (Red), G (Green), After separation into B (Blue) components, these pixels are combined to generate image data in which each pixel has three channel components of R, G, and B.

The color corrector 30 corrects the data spectral characteristics of the image sensor to be close to the sRGB or Adobe RGB spectral characteristics in order to obtain an output image approximating the sRGB or Adobe RGB standard, which is a standard color space.

The gamma correction unit 40 non-linearly transforms the intensity signal of the light by using a nonlinear transfer function in consideration of the fact that human vision reacts non-linearly to brightness.

The format converter 50 converts the image data of the RGB region into a Y component, which is a luminance component, and a Cb and Cr component, which is a chrominance component.

In general, a high sensitivity shot taken by increasing the sensitivity of the image sensor increases the analog gain of the image sensor, thereby increasing the noise, and consequently the noise component in the raw image data of the Bayer pattern output from the image sensor. Will increase.

When the noise is increased due to the high sensitivity shot The conventional image processing apparatus as shown in FIG. 1 performs image processing such as color interpolation, color correction, gamma correction, etc. on a Bayer image including the increased noise. In the process, the processing load is increased, and the processing efficiency is lowered.

In addition, the conventional image processing apparatus has a disadvantage in that the image processing is performed in a state in which the Bayer image contains a lot of noise components, so that the accuracy of the image processing is lowered, and as a result, the quality of the processed image is degraded.

Accordingly, it is a first object of the present invention to provide an image processing apparatus that can effectively reduce noise included in an image.

In addition, a second object of the present invention is to provide a noise removal method that can effectively reduce noise included in an image.

It is also a third object of the present invention to provide a recording medium on which a noise removing method is recorded, which can effectively reduce noise included in an image.

An image processing apparatus according to an aspect of the present invention for achieving the first object of the present invention described above, receives Bayer image data from an image sensor, the pixel of the same pixel components of the pixel values included in the Bayer image data Obtain an image boundary value based on the values, and if the image boundary value is smaller than a preset area boundary value, remove a Bayer noise by removing a correction value corresponding to the image boundary value among the values between the first and second correction values. And a color interpolation unit that performs color interpolation based on the Bayer noise removal value provided from the Bayer noise removal unit. The bayer noise removing unit may output the second correction value as the bayer noise removing value when the image boundary value is larger than the set area boundary value. The Bayer noise removing unit may obtain a Gaussian filtering value by obtaining a center pixel value from the Bayer image data and performing Gaussian filtering on the Bayer image data. The Bayer noise removing unit obtains a horizontal edge component value from the Bayer image data, obtains a vertical edge component value from the Bayer image data, and then extracts the edge component values of the horizontal direction and the vertical edge component values. The image boundary value can be obtained by calculating an average. The Bayer noise removing unit obtains a difference value between pixel values belonging to the same column among pixel values of a predetermined horizontal line and pixel values of an adjacent horizontal line having the same pixel component as the predetermined horizontal line in the Bayer image data. After multiplying, the average of the values multiplied by the weight is obtained to obtain an edge component value for a horizontal line, and then the average of absolute values for each of the edge component values for at least one horizontal line is obtained. Edge components can be obtained. The Bayer noise removing unit is configured to obtain a difference value between pixel values of a predetermined vertical line and pixel values belonging to a same row among pixel values of an adjacent vertical line having the same pixel component as the predetermined vertical line. After multiplying the weights, an average of the values multiplied by the weights is obtained to obtain an edge component value for a vertical line, and then an average of absolute values for each of the edge component values for at least one vertical line is obtained to obtain an average of the vertical directions. Edge components can be obtained. The Bayer noise removing unit may obtain the first correction value based on the center pixel value, the Gaussian filtering value, and a preset low frequency correction intensity value, and obtain the center pixel value, the Gaussian filtering value, and a preset high frequency correction intensity value. The second correction value may be obtained based on. The Bayer noise canceller may include: Equation BNR_out = (Edge_val × hf_out + (BNR_area_thr-Edge_val) × lf_out) / BNR_area_thr (where BNR_out is the Bayer noise removal value, Edge_val, when the image boundary value is smaller than a preset region boundary value). Is the image boundary value, BNR_area_thr is the set area boundary value, lf_out is the first correction value, and hf_out is the second correction value), among the values between the first correction value and the second correction value. A correction value corresponding to the image boundary value may be obtained.

In addition, according to an aspect of the present invention, there is provided a method for removing noise, comprising: receiving Bayer image data having a mask having a predetermined size, and among pixel values included in the Bayer image data. Obtaining an image boundary value based on pixel values of the same pixel components; and corresponding to the image boundary value among values between a first correction value and a second correction value when the image boundary value is smaller than a preset area boundary value. Outputting a correction value to be a Bayer noise removal value and performing color interpolation based on the Bayer noise removal value. The noise removing method may further include outputting the second correction value as the Bayer noise removing value when the image boundary value is larger than the set area boundary value. The step of obtaining an image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data may include obtaining a center pixel value from the Bayer image data and Gaussian filtering the Bayer image data. It may include the step of obtaining a Gaussian filtering value. The step of obtaining an image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data may include obtaining a horizontal edge component value from the Bayer image data, and in the Bayer image data. The method may include obtaining an edge component value in a vertical direction and calculating the average of the horizontal edge component value and the vertical edge component value to obtain the image boundary value. The step of obtaining a horizontal edge component value from the Bayer image data may include: belonging to a same column among pixel values of a predetermined horizontal line and pixel values of an adjacent horizontal line having the same pixel component as the predetermined horizontal line in the Bayer image data. Obtaining a difference value of pixel values, multiplying a predetermined weight, and then averaging the values multiplied by the weight to obtain an edge component value for a horizontal line, and an absolute value for each of the edge component values for at least one of the horizontal lines. Obtaining the horizontal edge component by averaging the values. The step of obtaining the edge component value in the vertical direction from the Bayer image data may include: setting the edge component values in the vertical direction to the same row of pixel values of a predetermined vertical line and pixel values of an adjacent vertical line having the same pixel component as the predetermined vertical line. Obtaining a difference value between pixel values belonging to each other, multiplying a predetermined weight, and then calculating an average of the multiplied weighted values to obtain an edge component value for a vertical line, and for each edge component value for at least one vertical line. Obtaining an average of absolute values to obtain an edge component in the vertical direction. When the image boundary value is smaller than the set area boundary value, outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value may include the center pixel value, the Acquiring the first correction value based on a Gaussian filtering value and a preset low frequency correction intensity value and obtaining the second correction value based on the center pixel value, the Gaussian filtering value, and a preset high frequency correction intensity value It may include a step. When the image boundary value is smaller than the set area boundary value, the step of outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as the Bayer noise removal value may be expressed by Equation BNR_out = ( Edge_val × hf_out + (BNR_area_thr-Edge_val) × lf_out) / BNR_area_thr (where BNR_out is the Bayer noise removal value, Edge_val is the image boundary value, BNR_area_thr is the set area boundary value, and lf_out is the first correction value, hf_out is the above A second correction value) to obtain a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value.

In addition, the recording medium recording the noise removing method according to an aspect of the present invention for achieving the third object of the present invention is provided with the Bayer image data having a mask of a predetermined size, and included in the Bayer image data Obtaining an image boundary value based on pixel values of the same pixel components among the pixel values; and when the image boundary value is smaller than a preset area boundary value, the image among the values between the first correction value and the second correction value. A program for outputting a correction value corresponding to a boundary value as a Bayer noise removal value and performing color interpolation based on the Bayer noise removal value is recorded.

According to the recording medium in which the image processing apparatus, the noise removing method, and the noise removing method of the image processing apparatus are recorded, the edge component is determined by using the image boundary value Edge_val in the Bayer image input with the predetermined mask size. In order to minimize the loss of edge information around the edges, the loss of edge information is minimized. In the flat plane region, the noise reduction intensity is set between the low frequency region correction value (lf_out) and the high frequency region correction value (hf_out) according to the edge boundary value (Edge_val). Adaptive adjustment to remove noise

Therefore, the edge region and the noisy region can be distinguished well, and thus, the noise included in the Bayer image can be effectively removed to improve the output image quality.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals are used for like elements in describing each drawing.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, with reference to the accompanying drawings, it will be described in detail a preferred embodiment of the present invention. In the following description of the present invention, the same reference numerals are used for the same elements in the drawings and redundant descriptions of the same elements will be omitted.

2 is a block diagram illustrating a configuration of an image processing apparatus according to an exemplary embodiment of the present invention, and FIG. 3 is a conceptual diagram illustrating a function of a preprocessor shown in FIG. 2. In FIG. 2, illustration of known functions and configurations that may unnecessarily obscure the subject matter of the present invention are omitted.

2 and 3, the image processing apparatus may include a Bayer noise remover 100 and a color interpolator 200.

The Bayer noise removing unit 100 receives Bayer image data from an image sensor, performs preprocessing on the provided Bayer image data, and determines an edge component. In this case, the noise is removed adaptively according to the edge components by lowering the intensity of removing noise in the edge region of the image contents.

In detail, the Bayer noise remover 100 may include a preprocessor 110 and a noise remover 130.

The preprocessing unit 110 receives Bayer image data having a 5 × 5 mask size from the image sensor, extracts the center pixel value Bc from the provided Bayer image data, and performs Gaussian filtering to generate a Gaussian filtering value GBc. do.

That is, as shown in (a) of FIG. 3, Bayer image data of 5 × 5 mask size provided from an image sensor includes Bayer patterns of any one of four Bayer patterns having R, Gb, Gr, and B as central pixel components. It may be Bayer image data having a, and as shown in FIG. 3B, each pixel value may be represented by P11 to P55 according to the position of the pixel.

As shown in FIG. 3B, the preprocessor 110 extracts a value of P33 (ie, Bc = P33) from the Bayer image data to the central pixel value Bc.

In addition, the preprocessor 110 may include pixels P11, P13, P15, P31, P33, P35, and P51 having the same pixel component as the center pixel P33 of the Bayer image data, as illustrated in FIG. 3C. Gaussian filter is applied to P53 and P55 to obtain Gaussian filtering value GBc.

The Gaussian filtering value GBc may be calculated by Equation 1.

GBc = (P11 + 2 × P13 + P15 + 2 × P31 + 4 × P33 + 2 × P35 + P51 + 2 × P53 + P55) / 16

In addition, the preprocessor 110 obtains an image boundary value Edge_val for Bayer image data having a 5 × 5 mask size provided from the image sensor.

In order to obtain an image boundary value Edge_val, the preprocessing unit 110 first obtains horizontal edge component values in a Bayer pattern of a 5 × 5 mask shown in FIG. That is, in a Bayer pattern having a 5 × 5 mask size as shown in Equation 2, pixel values belonging to the same column among pixel values of a predetermined horizontal line and pixel values of adjacent horizontal lines having the same pixel component as the predetermined horizontal line. The four edge component values for the horizontal line are obtained by multiplying the difference value of the multiplier by multiplying a predetermined weight and averaging the multiplied weights. The edge component values in the horizontal direction are obtained by obtaining an average of absolute values of the four edge component values obtained as described above.

H_EDGE_COMP1 = (2 × (P11-P31) + 4 × (P12-P32) + 4 × (P13-P33) + 4 × (P14-P34)

+ 2 × (P15-P35)) / 16

H_EDGE_COMP2 = (2 × (P21-P41) + 4 × (P22-P42) + 4 × (P23-P43) + 4 × (P24-P44)

+ 2 × (P25-P45)) / 16

H_EDGE_COMP3 = (2 × (P31-P51) + 4 × (P32-P52) + 4 × (P33-P53) + 4 × (P34-P54)

+ 2 × (P35-P55)) / 16

H_EDGE_COMP4 = (2 × (P11-P51) + 4 × (P12-P52) + 4 × (P13-P53) + 4 × (P14-P54)

+ 2 × (P15-P55)) / 16

H_EDGE = (abs (H_EDGE_COMP1) + abs (H_EDGE_COMP2) + abs (H_EDGE_COMP3)

+ abs (H_EDGE_COMP1)) / 4

In addition, the preprocessing unit 110 obtains an edge component in the vertical direction in a Bayer pattern of a 5 × 5 mask. That is, in a Bayer pattern having a 5 × 5 mask size as shown in Equation 3, pixels belonging to the same row among pixel values of a predetermined vertical line and pixel values of adjacent vertical lines having the same pixel component as the predetermined vertical line. The four edge component values for the vertical line are obtained by calculating the difference between the values, multiplying the preset weights, and averaging the values multiplied by the weights. The edge component values in the vertical direction are obtained by obtaining an average of absolute values of the four edge component values obtained as described above.

V_EDGE_COMP1 = (2 × (P11-P13) + 4 × (P21-P23) + 4 × (P31-P33) + 4 × (P41-P43)

+ 2 × (P51-P53)) / 16

V_EDGE_COMP2 = (2 × (P12-P14) + 4 × (P22-P24) + 4 × (P32-P34) + 4 × (P42-P44)

+ 2 × (P52-P54)) / 16

V_EDGE_COMP3 = (2 × (P13-P15) + 4 × (P23-P25) + 4 × (P33-P35) + 4 × (P43-P45)

+ 2 × (P53-P55)) / 16

V_EDGE_COMP4 = (2 × (P11-P15) + 4 × (P21-P25) + 4 × (P31-P35) + 4 × (P41-P45)

+ 2 × (P51-P55)) / 16

V_EDGE = (abs (V_EDGE_COMP1) + abs (V_EDGE_COMP2) + abs (V_EDGE_COMP3)

+ abs (V_EDGE_COMP1)) / 4

The preprocessing unit 110 obtains the horizontal edge component value H_EDGE and the vertical edge component value V_EDGE through Equations 2 and 3, and then, as shown in Equation 4, the horizontal edge component value H_EDGE and An image boundary value Edge_val is obtained by calculating an average of the vertical edge component values V_EDGE.

Edge_val = (H_EDGE + V_EDGE) / 2

The most important factor in removing the noise included in the Bayer image is to distinguish the noise from the edge components well, and then perform the noise reduction process only on the region containing the noise component.

Since the human eye is more sensitive to noise in a flat area than the edges of the image, in an embodiment of the present invention, the edge component is determined by using an image boundary value Edge_val to distinguish the edge from the noise component. By reducing the noise removal intensity around the edges, the loss of edge information is minimized. In the flat areas, the noise reduction intensity is increased to effectively reduce the noise contained in the image.

The noise removing unit 130 may include a center pixel value Bc, a Gaussian filtering value GBc, and an image boundary value Edge_val provided from the preprocessor 110, and a Bayer noise removal enable value BNR_on which is a parameter value provided by the developer. ), The correction values for the low frequency region and the high frequency region are determined based on the region boundary value (BNR_area_thr), the low frequency correction intensity value (BNR_lf_str), and the high frequency correction intensity value (BNR_hf_str). The noise removal intensity is adaptively adjusted according to the image edge value Edge_val to remove noise.

Specifically, the noise removing unit 130 may generate a low frequency region correction value based on the center pixel value Bc and Gaussian filtering value GBc provided from the preprocessor 110 and the low frequency correction intensity value BNR_lf_str provided by the developer. lf_out) is calculated as in Equation 5.

lf_out = ((255-BNR_lf_str) × Bc + BNR_lf_str × GBc) / 256

Here, the low frequency region means a flat region in the image, and the low frequency region correction value lf_out is a Bayer noise removal value BNR_out output for the low frequency region, that is, the region where the image boundary value Edge_val is 0, and is 5 × 5. When each pixel value included in the Bayer pattern having the mask size is composed of 8 bits, the pixel value is normalized to 0 to 255.

In addition, the noise removing unit 130 based on the central pixel value Bc and Gaussian filtering value GBc provided from the preprocessor 110 and the high frequency correction intensity value BNR_hf_str provided by the developer, the high frequency region correction value hf_out. ) Is calculated as in Equation 6.

hf_out = ((255-BNR_hf_str) × Bc + BNR_hf_str × GBc) / 256

Here, the high frequency region means an edge region included in the image, and the high frequency region correction value hf_out corresponds to a high frequency region, that is, when the image boundary value Edge_val has a value larger than the region boundary value BNR_area_thr provided by the developer. A Bayer noise removal value (BNR_out) that is outputted in relation to each other is a value normalized to 0 to 255 when each pixel value included in the Bayer pattern having a 5 × 5 mask size is composed of 8 bits.

In addition, when the image boundary value Edge_val provided from the preprocessing unit 110 is greater than zero and smaller than the area boundary value BNR_area_thr, the noise removing unit 130 adapts the noise removal intensity according to the image boundary value Edge_val. By adjusting it to output Bayer noise reduction value (BNR_out).

The Bayer noise removal value BNR_out when the image boundary value Edge_val is greater than zero and smaller than the area boundary value BNR_area_thr may be calculated by Equation 7.

BNR_out = (Edge_val × hf_out + (BNR_area_thr-Edge_val) × lf_out) / BNR_area_thr

If the Bayer noise removal enable value BNR_on is a value indicating the deactivation, the noise removal unit 130 does not perform a noise removal operation, and replaces the Bayer noise removal value (Bc) with the center pixel value Bc provided to the preprocessor 110. BNR_out).

The color interpolator 200 receives the Bayer noise removal value BNR_out from the Bayer noise canceller 100, and separates each pixel component having one channel component into R, G, and B components through interpolation. These are combined to generate image data in which each pixel has three channel components of R, G, and B.

4 is a conceptual diagram illustrating a noise removing method of an image processing apparatus according to an embodiment of the present invention.

Referring to FIG. 4, the Bayer noise removing unit 100 divides an area for removing noise into two areas based on an area boundary value BNR_area_thr provided from a developer, and calculates an image boundary value calculated through Equation 4 ( When Edge_val) is larger than the area boundary value BNR_area_thr, the high frequency area correction value hf_out calculated through Equation 6 is output as the Bayer noise removal value BNR_out, and the image edge value Edge_val is the area boundary value (BNR_area_thr). If it is smaller than BNR_area_thr), the value corresponding to the image boundary value Edge_val among the values between the low frequency region correction value lf_out and the high frequency region correction value hf_out is calculated by Equation 7 and then Bayer noise removal value BNR_out. Will output

In FIG. 4, when the image boundary value Edge_val is larger than the region boundary value BNR_area_thr, since the image edge corresponds to an edge in the image with a high frequency region, the noise elimination intensity is set low to remove the noise, and the image boundary value ( If the edge_val) is smaller than the area boundary value BNR_area_thr, it may be regarded as a flat surface in the image as the edge value of the image is closer to zero. Therefore, according to the exemplary embodiment of the present invention, as the image boundary value Edge_val approaches 0, the noise elimination intensity is adaptively set to be high to remove the noise, thereby improving the accuracy of the noise elimination.

5 is a flowchart illustrating a noise removing process of an image processing apparatus according to an embodiment of the present invention.

Referring to FIG. 5, first, the Bayer noise removing unit 100 receives Bayer image data having a 5 × 5 mask size from an image sensor (step 501), a center pixel value Bc, a Gaussian filtering value GBc, and an image. The edge value Edge_val is calculated (step 503).

Here, the Gaussian filtering value GBc may be obtained through Equation 1, and the image boundary value Edge_val may be obtained through Equations 2 to 4. The image edge (Edge_val) is a value for determining edge components included in an image, and applies different noise reduction strengths to edge regions and flat surface regions, and adaptively applies noise reduction strengths according to edge components. Is used.

In addition, the Bayer noise removing unit 100 may include a Bayer noise removal enable value BNR_on, an area boundary value BNR_area_thr, a low frequency correction intensity value BNR_lf_str, and a high frequency correction intensity value BNR_hf_str, which are parameters for removing noise from a developer. Is provided (step 505).

In an embodiment of the present invention, for example, the parameter value is provided from a developer, but in another embodiment of the present invention, a control unit (not shown) included in the image processing device or an image processing device included in the system is installed. The controller (not shown) may be configured to adaptively provide the parameter values to the Bayer noise canceller 100 by referring to a preset lookup table according to an image photographing environment such as an exposure value, an image sensor sensitivity, and the like.

Thereafter, the Bayer noise removing unit 100 determines whether the Bayer noise removing enable value BNR_on is a value indicating the activation (for example, 1) (step 507), and the Bayer noise removing enable value BNR_on If it is determined that a value indicating activation is input, the low frequency region correction value lf_out and the high frequency region correction value hf_out are calculated (step 509). Here, the low frequency region correction value lf_out may be calculated through Equation 5, and the high frequency region correction value hf_out may be calculated through Equation 6.

Next, the Bayer noise removing unit 100 determines whether the image boundary value Edge_val acquired in step 503 is smaller than the region boundary value BNR_area_thr (step 511), and the image boundary value Edge_val is the region boundary value ( If it is determined to have a value smaller than BNR_area_thr, the Bayer noise removal value BNR_out is output by adaptively adjusting the noise removal intensity according to the image boundary value Edge_val through Equation 7 (step 513).

In operation 513, the Bayer noise removal value BNR_out may have a value between the low frequency region correction value lf_out and the high frequency region correction value hf_out.

If it is determined in step 511 that the image boundary value Edge_val has a value larger than the area boundary value BNR_area_thr, the Bayer noise removing unit 100 outputs the high frequency region correction value hf_out as the Bayer noise removing value BNR_out. (Step 515).

If the Bayer noise removal enable value BNR_on is a value indicating inactivation in step 507, it means that the Bayer image noise removal is not performed. Therefore, the Bayer noise removal unit 100 is centered on the Bayer noise removal value BNR_out. The pixel value Bc is output.

6 illustrates a noise removal result of an image processing apparatus according to an embodiment of the present invention. FIG. 6A illustrates an image in which noise removal is not performed on a Bayer image, and FIG. The result of performing noise removal on the Bayer image according to an embodiment of the present invention is shown.

As illustrated in FIG. 6, in the noise removing method according to an exemplary embodiment, edge components are determined using an image boundary value Edge_val in units of 5 × 5 masks in a Bayer image, and in an edge periphery 601. Reduce the noise removal intensity to minimize the loss of edge information, and in the flat surface region 603, the noise removal intensity is adapted between the low frequency region correction value (lf_out) and the high frequency region correction value (hf_out) according to the edge boundary value Edge_val. By adjusting the noise, the noise included in the Bayer image can be effectively removed.

Although described with reference to the embodiments above, those skilled in the art will understand that the present invention can be variously modified and changed without departing from the spirit and scope of the invention as set forth in the claims below. Could be.

1 is a block diagram showing the configuration of a general image signal processing apparatus.

2 is a block diagram illustrating a configuration of an image processing apparatus according to an embodiment of the present invention.

3 is a conceptual diagram illustrating a function of a preprocessor shown in FIG. 2.

4 is a conceptual diagram illustrating a noise removing method of an image processing apparatus according to an embodiment of the present invention.

5 is a flowchart illustrating a noise removing process of an image processing apparatus according to an embodiment of the present invention.

6 illustrates a noise removal result of the image processing apparatus according to the exemplary embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

100: Bayer noise removing unit 110: Preprocessing unit

130: noise removing unit 200: color interpolation unit

Claims (17)

An image processing apparatus for removing noise of an image provided from an image sensor, comprising: A Bayer image data is provided from the image sensor, an image boundary value is obtained based on pixel values of the same pixel components among the pixel values included in the Bayer image data, and the image boundary value is smaller than a preset area boundary value. And outputs a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value. When the image boundary value is larger than the set area boundary value, the second correction value is output. A Bayer noise removing unit for outputting a value as the Bayer noise removing value; And And a color interpolation unit which performs color interpolation based on the Bayer noise removal value provided from the Bayer noise removing unit. delete The method of claim 1, wherein the Bayer noise removing unit, And a Gaussian filtering value by obtaining a central pixel value from the Bayer image data and performing Gaussian filtering on the Bayer image data. The method of claim 3, wherein the Bayer noise removing unit, A horizontal edge component value is obtained from the Bayer image data, a vertical edge component value is obtained from the Bayer image data, and then the average of the horizontal edge component value and the vertical edge component value is calculated. And an image boundary value. The method of claim 4, wherein the Bayer noise removing unit, In the Bayer image data, a difference value between pixel values belonging to a same column among pixel values of a predetermined horizontal line and pixel values of an adjacent horizontal line having the same pixel component as the predetermined horizontal line is obtained, multiplied by a preset weight, and then the weighted value. Calculates an edge component value for a horizontal line by averaging the multiplied values, and then obtains an edge component in the horizontal direction by calculating an average of absolute values for each of the edge component values for at least one horizontal line. Image processing apparatus. The method of claim 4, wherein the Bayer noise removing unit, In the Bayer image data, a difference value between pixel values belonging to the same row among pixel values of a predetermined vertical line and pixel values of an adjacent vertical line having the same pixel component as the predetermined vertical line is obtained and multiplied by a preset weight. Obtaining an edge component value for a vertical line by obtaining an average of weighted values, and then calculating an average of absolute values for each of the edge component values for at least one vertical line to obtain an edge component in the vertical direction. Image processing apparatus. The method of claim 3, wherein the Bayer noise removing unit, Obtain the first correction value based on the center pixel value, the Gaussian filtering value, and a preset low frequency correction intensity value, and based on the center pixel value, the Gaussian filtering value, and a preset high frequency correction intensity value, And a correction value is obtained. The method of claim 3, wherein the Bayer noise removing unit, If the image boundary value is smaller than the preset area boundary value, BNR_out = (Edge_val × hf_out + (BNR_area_thr-Edge_val) × lf_out) / BNR_area_thr (Where BNR_out is the Bayer noise removal value, Edge_val is the image boundary value, BNR_area_thr is the set area boundary value, lf_out is the first correction value, and hf_out is the second correction value). And a correction value corresponding to the image boundary value among the values between the correction value and the second correction value. In the noise removing method of the image processing apparatus, Receiving Bayer image data having a mask having a predetermined size; Obtaining an image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data; Outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value when the image boundary value is smaller than a preset area boundary value; Outputting the second correction value as the Bayer noise removal value when the image boundary value is larger than the set area boundary value; And And performing color interpolation based on the Bayer noise removal value. delete The method of claim 9, wherein the obtaining of the image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data includes: Obtaining a center pixel value from the Bayer image data; And And performing Gaussian filtering on the Bayer image data to obtain a Gaussian filtering value. The method of claim 11, wherein the obtaining of the image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data comprises: Obtaining edge component values in a horizontal direction from the Bayer image data; Obtaining edge component values in the vertical direction from the Bayer image data; And And calculating the average of the horizontal edge component value and the vertical edge component value to obtain the image boundary value. The method of claim 12, wherein the obtaining of the horizontal edge component value from the Bayer image data comprises: In the Bayer image data, a difference value between pixel values belonging to a same column among pixel values of a predetermined horizontal line and pixel values of an adjacent horizontal line having the same pixel component as the predetermined horizontal line is obtained, multiplied by a preset weight, and then the weighted value. Obtaining an edge component value for the horizontal line by calculating the average of the multiplied values; And Obtaining the horizontal edge component by averaging the absolute values for each of the edge component values for at least one of the horizontal lines. The method of claim 12, wherein the obtaining of the vertical edge component value from the Bayer image data comprises: In the Bayer image data, a difference value between pixel values belonging to the same row among pixel values of a predetermined vertical line and pixel values of an adjacent vertical line having the same pixel component as the predetermined vertical line is obtained and multiplied by a preset weight. Obtaining an edge component value for the vertical line by averaging the weighted values; And And obtaining an average of absolute values for each of the edge component values for at least one vertical line to obtain the edge component in the vertical direction. The method of claim 11, wherein when the image boundary value is smaller than the set area boundary value, outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value Obtaining the first correction value based on the center pixel value, the Gaussian filtering value, and a preset low frequency correction intensity value; And And obtaining the second correction value based on the center pixel value, the Gaussian filtering value, and a preset high frequency correction intensity value. The method of claim 11, wherein when the image boundary value is smaller than the set area boundary value, outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value , Equation BNR_out = (Edge_val × hf_out + (BNR_area_thr-Edge_val) × lf_out) / BNR_area_thr (Where BNR_out is the Bayer noise removal value, Edge_val is the image boundary value, BNR_area_thr is the set area boundary value, lf_out is the first correction value, and hf_out is the second correction value). And a correction value corresponding to the image boundary value among values between the correction value and the second correction value. In the recording medium in which a program of instructions executable by a digital processing device for removing noise of an image is tangibly implemented, and recording a program that can be read by the digital processing device, Receiving Bayer image data having a mask having a predetermined size; Obtaining an image boundary value based on pixel values of the same pixel components among the pixel values included in the Bayer image data; Outputting a correction value corresponding to the image boundary value among the values between the first correction value and the second correction value as a Bayer noise removal value when the image boundary value is smaller than a preset area boundary value; Outputting the second correction value as the Bayer noise removal value when the image boundary value is larger than the set area boundary value; And And a program for performing color interpolation based on the Bayer noise removal value.

Images