KR101887515B1 - Image error mitigation method using neighboring pixels in noise environment - Google Patents

Abstract

The present invention relates to a method of mitigating image error using surrounding pixel values in a noisy environment. A method of mitigating an image error using neighboring pixel values in a noisy environment according to the present invention includes the steps of: a) generating an error mask through noise detection on a received image to determine a location of a damaged pixel; b) generating an extended error mask that takes into account noise removal at the end of the image of the corrupted pixel where the location is located; c) identifying the corrupted pixel according to the generated extended error mask; d) verifying that neighboring pixels of the corrupted pixel are all noisy; e) determining an average value of the pixel values using the sum of the pixel values and the number of pixels for the normal pixels considering only the normal pixels that are not corrupted by the noise as a result of the checking in the step d); And f) mitigating noise by replacing the averaged value with the pixel value of the corrupted pixel.
According to the present invention, by calculating an average value of error-free pixels around error pixels and assigning them as new pixel values, it is possible to perform better than conventional MF-based techniques in low and high salt and pepper noise environments It is possible to obtain a good noise reduction effect.

Description

[0001] The present invention relates to an image error mitigation method using neighboring pixel values in a noise environment,

The present invention relates to an image error mitigation method, and more particularly, to an image error mitigation method that detects an error pixel of a received image and uses surrounding pixel values in a noise environment that can mitigate an image error by using surrounding pixel values of the error pixel. To an image error mitigation method.

Recently, as the technology of modern society develops, the necessity of image processing using images and images in home appliances, disaster / military video surveillance, and medical fields is increasing.

In general, noise is generated due to various factors in a process of acquiring an analog signal, an analog to digital (A / D) conversion and a transmission process to a digital image or an image. These noises are usually classified as Gaussian noise, impulse noise, etc., and there is salt and pepper noise as one of impulse noise. Salt and Pepper noise is a noise that makes the value of any pixel to the minimum value (0) or maximum value (255), which degrades the original data and degrades the compression efficiency and degrades image processing such as edge detection or segmentation. Which causes the performance deterioration. Also, when the channel environment of the wireless communication or the environment of the satellite communication system is poor, high-density salt and pepper noise may occur. Therefore, research to mitigate this is actively proceeding at home and abroad.

Typical filters used to mitigate salt and pepper noise include Median Filter, Weighted Median Filter (WMF), Center Weighted Median Filter (CWMF), Switching Weighted Median Filter (SWMF), Adaptive Median Filter .

MF selects an intermediate value after aligning the center pixel value and the pixel values in the mask area in order of magnitude using an N × N size mask. Since MF is efficient and easy to implement, several MF-based filters have been proposed.

WMF and CWMF are weighted MF and have good edge preservation characteristics. However, there is a disadvantage that it affects the entire image or image even though the pixels are not damaged by noise. SWMF and AMF have been proposed to overcome this problem. Among them, SWMF processes the pixel values in the image mask with MF and judges whether or not noise is generated according to the difference between the central pixel value and the processed intermediate value. If the difference is larger than the predetermined value, it is judged as noise, and the central pixel value is outputted as the intermediate value. If the difference is smaller than the predetermined value, the normal pixel value is outputted. AMF applies MF only in pixels that are damaged in salt and pepper noise environments. Check the value of all pixels in the image or image and determine the pixel with 0 or 255 value as corrupted pixel and treat it as MF only in that pixel. However, the MF-based filters are disadvantageous in that when the pixel values of more than half are damaged in neighboring pixels including the center pixel, the noise reduction becomes difficult and the performance is degraded.

On the other hand, Japanese Patent Application Laid-Open No. 10-0574862 (Patent Document 1) discloses a method and apparatus for processing an error-generating block of image data, and a method of processing an error- Determining whether an error block exists, and if the error block exists, obtaining an average value of the chrominance information of the pixel in the error block; Obtaining standard deviations with respect to an average value of the obtained error blocks for each of the leftmost and topmost pixels in the error block; Obtaining a standard deviation of an average value of the obtained error blocks with respect to each pixel adjacent to the error block in the left block adjacent to the error block and the upper block; Determining whether a difference between the standard deviation of the block adjacent to the obtained error block and the standard deviation of the obtained error block exceeds a predetermined threshold value; And skipping the error concealment for the error block if the difference between the standard deviation of the error block adjacent block and the standard deviation of the error block obtained does not exceed a predetermined threshold as a result of the determination, And performing error concealment on the block.

In the case of Patent Document 1 as described above, it is possible to minimize the loss of normal data due to the error concealment method, and simultaneously improve the efficiency of computation due to unnecessary error concealment, thereby improving the image quality due to minimization of normal data loss, Although it may be advantageous to improve the efficiency at the same time, it is also based on the MF-based filter method, and when half or more pixel values are damaged in neighboring pixels including the central pixel, it is difficult to mitigate noise, .

Patent Registration No. 10-0574862 (April 21, 2006)

The present invention has been made in view of the above problems of the prior art, and it is an object of the present invention to check the position of error pixels of a received image and sequentially calculate an average value of errorless pixels around the error pixels And a method of mitigating an image error using surrounding pixel values in a noise environment in which an image error can be alleviated by designating a new pixel value.

According to an aspect of the present invention, there is provided an image error mitigation method using neighboring pixel values in a noisy environment,

a) generating an error mask through noise detection on the received image to determine the location of the corrupted pixel;

b) generating an extended error mask that takes into account noise removal at the end of the image of the corrupted pixel where the location is located;

c) identifying a corrupted pixel according to the generated extended error mask;

d) verifying that neighboring pixels of the identified corrupted pixel are all noise;

e) calculating an average value of pixel values using the sum of pixel values and the number of pixels for normal pixels considering only normal pixels not damaged by noise; And

f) mitigating the noise by replacing the obtained average value with the pixel value of the defective pixel.

Here, as a result of the checking in step c), the uncorrelated pixel may further include the step of preserving the pixel value as it is.

Further, as a result of the determination in step d), it may further include the step of taking 127, which is a median value of Salt and Pepper noise (0 or 255), when the center pixel and adjacent pixels are both noises have.

In order to determine the position of the damaged pixel in step a), one image is divided into three areas of red, green, and blue, and the same i * an error mask E (i, j) having a size j can be defined.

At this time, the error mask E (i, j) can be represented by the following equation.

Here, D (i, j) represents an image when an image having an i × j size is damaged by Salt & Pepper noise.

In addition, in the step b), the extended error mask may be generated as three areas of red, green, and blue, and an error mask may be generated for the newly generated part adjacent to the existing error mask. All zero values can be assigned to the corrupted area.

According to the present invention, the position of the error pixels in the received image is confirmed, and an average value of the error-free pixels around the error pixels is calculated and designated as a new pixel value, so that low salt and pepper noise Environment, it is possible to obtain better performance than conventional MF-based techniques, and obtain good noise mitigation effects even in high salt and pepper noise environments.

1 is a flowchart illustrating an image error mitigation method using surrounding pixel values in a noisy environment according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating an outline of a noise mitigation filter technique that is introduced in a method of mitigating image error using surrounding pixel values in a noisy environment according to the present invention.
3 is a diagram illustrating a process of generating an extended error mask in an image error mitigation method using surrounding pixel values in a noisy environment according to the present invention.
FIG. 4 is a diagram illustrating a noise reduction filter introduced in the image error mitigation method using surrounding pixel values in a noisy environment according to the present invention, and a Lena image and a Crew image for evaluating performance of existing filters.
5 is a diagram illustrating regions of neighboring pixels for noise mitigation in an image error mitigation method using neighboring pixel values in a noisy environment according to the present invention.
6 is a diagram illustrating a simulation result of a noise reduction filter introduced in the image error mitigation method using surrounding pixel values in a noisy environment according to the present invention and a Lena image using existing filters.
FIG. 7 is a diagram illustrating a simulation result of a noise reduction filter introduced in a method of reducing image errors using peripheral pixel values in a noisy environment according to the present invention and a Crew image using existing filters.
8 is a graph showing a PSNR graph according to the Salt & Pepper noise density in the Lena image by the noise reduction filter and existing filters introduced in the image error mitigation method of the present invention.
9 is a graph showing a PSNR graph according to Salt & Pepper noise density in a noise reduction filter introduced in the image error mitigation method of the present invention and Crew images by existing filters.
10 is a diagram showing a PSNR comparison table of the noise mitigation filter introduced in the image error mitigation method of the present invention and the Lena image and the Crew image by the existing filters.

The terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary terms and the inventor can properly define the concept of the term to describe its invention in the best way Should be construed in accordance with the principles and meanings and concepts consistent with the technical idea of the present invention.

Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise. Also, the terms " part, "" module, "and" device " Lt; / RTI >

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

1 is a flowchart illustrating an image error mitigation method using surrounding pixel values in a noisy environment according to an embodiment of the present invention.

1, an image error mitigation method using neighboring pixel values in a noisy environment according to the present invention includes an adjacent pixel-based noise mitigation filter (APNMF) that can mitigate Salt and Pepper noise, based noise mitigation filter is introduced to mitigate an image error. First, an error mask is generated through noise detection on a received image to determine the position of a damaged pixel (step S101). Here, in order to grasp the position of the defective pixel, one image is divided into three regions of red, green, and blue, and each region has the same ixj size An error mask E (i, j) can be defined.

At this time, the error mask E (i, j) can be represented by the following equation.

Here, D (i, j) represents an image when an image having an i × j size is damaged by Salt & Pepper noise.

When the position of the damaged pixel is recognized in this way, an extended error mask is generated considering noise removal at the end of the image of the damaged pixel whose position is detected (step S102). Here, the extended error mask may be generated as three areas of red, green, and blue, and all of the error areas of the newly generated area adjacent to the existing error mask are all 0 zero can be assigned to the damaged area. In addition, the extended error mask varies in size according to an area (NxN) of adjacent pixels to be used when noise is mitigated.

Upon completion of the generation of the extended error mask, the corrupted pixels are identified according to the generated extended error mask (step S103). Here, as a result of checking, the pixel value of the uncompacted pixel is stored as it is (step S104).

If it is determined in step S103 that the corrupted pixels are detected, it is determined whether adjacent pixels of the corrupted pixel are all noise (step S105).

As a result of checking in step S105, if there is a pixel damaged by noise, the damaged pixel is ignored, and an average of pixel values is obtained by using the pixel value sum and the pixel number of normal pixels considering only the uncommitted normal pixels (Step S106). Here, the calculation of the average of the pixel values will be described later. As a result of the determination in step S105, if the center pixel and the neighboring pixels are all noise, 127 is taken as the median value of the Salt and Pepper noise value (0 or 255) (step S107). Here, taking the intermediate value 127 as described above is intended to alleviate the noise as much as possible.

Thereafter, the obtained average value is replaced with the pixel value of the damaged pixel to alleviate noise (step S108).

Hereinafter, an image error mitigation method using neighboring pixel values in a noise environment according to the present invention will be described in further detail.

In general, adjacent pixels in an image have similar pixel values. Accordingly, in the image error mitigation method using peripheral pixel values in a noisy environment according to the present invention, a method of mitigating noise by replacing a corrupted pixel value with an average value of adjacent pixels is employed. As shown in FIG. 2, when it is assumed that noise of a pixel P (x, y) that has been damaged is mitigated by using 3 × 3 adjacent pixels, an average value of adjacent pixels P ₁ to P ₈ is calculated to obtain P , y).

<Noise detection>

In the image error mitigation method according to the present invention, in order to mitigate the noise of the corrupted image, the noise of the received image is first detected. Define D (i, j) when an image with an i × j size is corrupted by Salt & Pepper noise. At this time, in order to grasp the position of noise, one image is divided into three areas of red, green, and blue, and error masks E (i, j), and the error mask E (i, j) can be expressed by the following equation (1) as described above.

Salt and Pepper noise represents a pixel value of 0 or 255 when the image is corrupted. Therefore, all the pixel values of D (i, j) are compared, and a pixel having a value of 0 or 255 is determined as a damaged pixel and assigned a value of 0, and a pixel having a value other than 0 is assigned a value of 1 as a normal pixel.

<Extended error mask>

When the noise at the end of the image is relaxed, noise reduction may not be smooth due to size limitations. In order to solve this problem, in the image error mitigation method according to the present invention, an extended error mask is generated as shown in FIG. The extended error mask is generated as three areas of red, green, and blue as in the conventional error mask. In the case of using a 3 × 3 adjacent pixel area for noise reduction, The size of the error mask increases by 2 pixels in both the horizontal and vertical directions. At this time, 0 is allocated to the newly generated part as an error value and designated as the damaged area. This allows more precise noise reduction by excluding this region from the computation when noise is removed at the end of the image.

<Noise Reduction>

The image error mitigation method according to the present invention proceeds according to the following procedure.

Step 1: Check whether the pixel is damaged through extended error mask. If the corresponding pixel is normal, the following Equation 2 is processed.

Step 2: If the pixel is noise, the adjacent eight pixels are checked as shown in FIG. At this time, whether or not the neighboring pixels are damaged through the error mask is extracted, and only values of neighboring normal pixels are extracted and processed as shown in the following Equations (3) and (4).

In Equations (3) and (4), S (i, j) denotes the sum of normal pixel values and C (i, j) denotes the number of normal adjacent pixels. And E represents noise information (0 or 1), and P represents an adjacent pixel value.

Step 3: Substituting the values obtained by applying equations (3) and (4) in step 2 into the following equation (5), an average value of values of adjacent pixels is obtained and output to the degraded pixels. However, if all the adjacent pixels are damaged, the processing is performed as shown in Equation (6). This is to prevent damage to the image as much as possible by taking the Salt & Pepper noise midway between 0 and 255.

Step 4: When noise reduction is completed in all areas of Red, Green, and Blue through a series of processes as described above, integrated image is outputted.

Meanwhile, in order to verify the superiority of the adjacent pixel-based noise mitigation filter (APNMF) introduced for implementing the image error mitigation method using the neighboring pixel values in the noise environment according to the present invention, (Salt & Pepper) noise of 10% to 95% was added to a Lena image of 512 × 512 pixels and a CREW image of 480 × 640 pixels, and the existing MF (Median Filter), AMF Filter) and the neighboring pixel based noise mitigation filter (APNMF) introduced in the present invention. Also, as shown in FIG. 5, when simulation is performed by changing the adjacent pixel region of the adjacent pixel-based noise reduction filter (APNMF) to the sizes of 3x3, 5x5, and 7x7, We also evaluated how the noise reduction performance changed.

The evaluation uses the Peak Signal-to-Noise Ratio (PSNR), which is a measure used in image quality evaluation in many imaging studies. The PSNR can be expressed by the following equations (7) and (8).

In the above equations (7) and (8), M and N are the sizes of the image and image in the horizontal and vertical directions, O is the original, and I is the object to be compared.

FIGS. 6 and 7 illustrate the use of conventional MF and AMF and adjacent pixel-based noise mitigation filters (APNMFs) by adding 80% density Salt and Pepper noise to the first frame image of the Lena and Crew images. Simulation results are shown. (B) is the MF, (c) is the AMF, and (d), (e), and (f) are the undamaged images with Salt and Pepper noise. (APNMF) of 3 × 3, 5 × 5, and 7 × 7 adjacent pixel based noise reduction filters. At high noise densities, MF and AMF remove some noise from the corrupted image, but there are many visual errors. However, in the case of using 3 × 3 adjacent pixel-based noise mitigation filter (APNMF), we can see that much noise is removed, and in the result using APNMF of 5 × 5 and 7 × 7 area, You can see the result similar to the image of.

FIGS. 8 and 9 show PSNR results according to changes in density of Salt and Pepper noise of Lena and Crew images. FIG. AMF shows relatively good performance at low noise density. However, the performance of MFF and AMF is insufficient in high density noise environment as the noise density increases and the number of damaged pixels increases. However, it can be seen that the adjacent pixel-based noise mitigation filter (APNMF) in the 3 × 3 region exhibits excellent overall PSNR results at all noise densities. The APNMF of 5 × 5 and 7 × 7 regions has a slightly lower PSNR than the AMF due to the wide area of neighboring pixels in a low density noise environment. However, in the case of the Lena image, PSNR results are better than those of the 3 × 3 region in noise environments of 50% and 65%, respectively. In the case of Crew images, It can be seen that PSNR results are better than those in the high density noise environment.

10 is a comparison chart of the PSNR results obtained from the simulation in the Lena image and the Crew image.

From the results of Table 1 and Table 2, it can be seen that the result of removing the noise by the adjacent pixel-based noise mitigation filter (APNMF) introduced in the image error mitigation method according to the present invention shows a superior PSNR result . In the Lena image, the conventional MF and AMF exhibit PSNR results of 5.64 [dB] and 6.38 [dB] in 90% density Salt and Pepper noise environment, respectively, and the adjacent pixel- The mitigation filter (APNMF) shows a PSNR of 6.69 [dB]. And the APNMFs in the 5 × 5 and 7 × 7 regions show excellent PSNR results of 16.63 [dB] and 22.39 [dB], respectively. In the Crew image, the conventional MF and AMF exhibit a PSNR of 8.54 [dB] and 8.88 [dB] respectively in 70% density salt and pepper noise environments, and the APNMF of 3 × 3 region is 17.59 [dB ] &Lt; / RTI &gt; The APNMFs in the 5 × 5 and 7 × 7 regions show excellent PSNR results of 23.85 [dB] and 24.55 [dB], respectively.

As described above, the image error mitigation method using the neighboring pixel values in the noise environment according to the present invention confirms the positions of the error pixels of the received image, sequentially calculates the average value of the errorless pixels around the error pixels To assign a new pixel value, it is possible to obtain a superior noise reduction effect in a high salt and pepper noise environment as well as in a low salt and pepper noise environment.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but many variations and modifications may be made without departing from the spirit and scope of the invention. Be clear to the technician. Accordingly, the true scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of the same should be construed as being included in the scope of the present invention.

P _{1 to} P ₈ : Pixels adjacent to pixels that are corrupted by noise

Claims (6)

a) generating an error mask through noise detection on the received image to determine the location of the corrupted pixel;
b) generating an extended error mask that takes into account noise removal at the end of the image of the corrupted pixel where the location is located;
c) identifying a corrupted pixel according to the generated extended error mask;
d) verifying that neighboring pixels of the identified corrupted pixel are all noise;
e) calculating an average value of pixel values using the sum of pixel values and the number of pixels for normal pixels considering only normal pixels not damaged by noise; And
f) mitigating noise by replacing the obtained average value with the pixel value of the defective pixel,
In the step b), the extended error mask is generated as three areas of red, green, and blue, and all of the error areas of the newly generated part adjacent to the existing error mask are all 0 (zero) is assigned to the damaged area,
In order to determine the position of the damaged pixel in step a), one image is divided into three areas of red, green, and blue, and the same ixj size (I, j) having an error mask E (i, j). delete delete delete delete delete

Images