KR100927458B1 - Filtering Method in Spatial Domain by Combining Sort and Average Operations - Google Patents

Abstract

The present invention relates to a filter processing method in a spatial domain by a combination of alignment and averaging, comprising: (a) arranging pixels in a mask in order of pixel values; (b) removing pixel values of a predetermined region from both ends of the aligned pixel values when the pixel values of pixels having an intermediate value among the aligned pixel values are not the same; And (c) calculating a resultant pixel value by averaging the pixel values of the remaining pixels; and (b-1) a pixel of a pixel having an intermediate value among pixel values aligned with a center pixel in the mask. If the values are the same, the method may further include assigning an original pixel value (a median of mask coefficients) to the center pixel in the mask.

According to the present invention, after removing a part of all the pixels in the mask and assigning different weights to the remaining pixels, there is an effect of minimizing the block effect or various noises related to the communication signal generated during the compression of the image.

Image Filter, Spatial Domain, Sort, Average, Block Effect, Noise, Weight

Description

Filtering Method in Spatial Domain by Combining Sort and Average Operations {Method for Filtering by Sorting and Mean Operation in Spatial Domain}

The present invention relates to a method of processing a filter in a spatial domain by a combination of alignment and averaging, and more particularly, by removing a part of all pixels in a mask and assigning different weights to residual pixels, thereby generating an image compression process. The present invention relates to a filter processing method in a spatial domain by a combination of an alignment and an average operation that can effectively mitigate various noises generated in association with a block effect or a communication signal.

Recently, due to the rapid development of mobile communication, high-speed data exchange is possible through the Internet on a mobile communication terminal. However, since the data exchange process necessarily requires compression of video and audio, a certain amount of noise is inevitable even in a communication network providing excellent service quality.

Accordingly, there have been studies related to the filter design of the receiver in order to mitigate the noise related to the loss or the communication signal generated in the compression process of the image as described above. This reflects the reality that the importance of the image filter is continuously increased because the recent communication mainly deals with the multimedia contents, whereas the conventional communication handles only the voice.

The function of the digital image filter is to process the original image and transform the result to suit a particular application. Image processing methods of an image filter include a method of processing in a spatial domain and a method of processing in a frequency domain. Processing in the spatial domain refers to a process of separating, enhancing or weakening a desired component of an image using a mask operation. Therefore, the mask design is an important part in the design of the image filter.

Conventional image filters include a mean filter and a median filter.

Among these, the average filter is a smooth linear spatial filter, and simply outputs an average value of pixels included in neighboring points of the filter mask. By replacing all pixel values in the image with the average intensity values of the neighboring pixels defined by the filter mask, an image having reduced sharp transitions of the intensity levels is generated. In general, since random noise consists of distinct transitions of contrast, the average filter application is mainly limited to the reduction of random noise.

The result of the filter process f (x, y) in the case of using the average filter can be expressed by the following [Equation 1].

In Equation 1, m and n represent the horizontal and vertical sizes of the mask, S _xy represents a pixel set in the mask, and g (s, t) represents a pixel value in the mask.

FIG. 1A illustrates a block effect of an image generated through a discrete cosine transform (DCT) and a compression process, and FIG. 1B illustrates an example of a 5 × 5 average filtered image.

Next, the median filter is not a linear filter such as the above-described average filter, but a nonlinear filter that processes the pixels by sorting them. The median filter replaces any pixel value with the median of the intensity levels that exist around the pixel.

The median filter exhibits less blurring for certain types of noise than the average filter of the same magnitude and provides excellent noise reduction. In other words, it is useful for removing impulse noise such as salt-and-pepper, but it is difficult to predict the application result of the filter in some cases because it is a nonlinear filter.

The result of the filter process f (x, y) in the case of using the median filter can be expressed by the following [Equation 2].

2A and 2B show an example of an impulse noise image and an image of a median-filtered image, respectively.

Although the average filter or the median filter described above has a partially useful result in removing specific noise, it is difficult to cope with various noises and is difficult to be effectively implemented in an environment in which real time communication constraints exist. In addition, there is a problem in that a blurring phenomenon that blurs the filtered image, in particular, a boundary line or a corner of an object, occurs.

That is, the average filter has some usefulness in reducing random noise, but has a problem of blurring in unwanted areas. The median filter has some usefulness in special noise such as impulse noise. There was a problem that the image is not very useful.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and removes a part of all pixels in a mask and assigns different weights to the remaining pixels to relate to a block effect or a communication signal generated in the compression process of an image. Minimize noise.

Another problem to be solved by the present invention is to solve the above problems, the filter processing method that can be implemented significantly improved image quality than the conventional filter by a simple operation, and can be applied to various kinds of damaged image To provide.

The present invention relates to a filter processing method in a spatial domain by a combination of alignment and averaging, comprising: (a) arranging pixels in a mask in order of pixel values; (b) removing pixel values of a predetermined region from both ends of the aligned pixel values when the pixel values of pixels having an intermediate value among the aligned pixel values are not the same; And (c) calculating a result pixel value by averaging the pixel values of the remaining pixels.

Preferably, the average operation of the step (c) is characterized in that the weighted average operation.

Also preferably, the step (c) may include: (c-1) weighting the remaining pixels; And (c-2) calculating an average value of a result value obtained by multiplying each pixel value of the residual pixels by the weight value.

Also preferably, the weight is given to the largest of the remaining pixels.

Also preferably, the weight may be assigned to form a Gaussian distribution based on the pixel value of the center pixel among the remaining pixels.

Also preferably, the average value may be calculated by the following equation.

(In the above equation, m and n represent the horizontal and vertical sizes of the mask, d represents the number of pixel values removed at one end of the aligned pixel values according to step (b), and z represents the ( c) represents the number of remaining pixels in step c), w _i represents a weight of the (d + i) -th pixel among the aligned pixels, and R represents a pixel value calculated as a result of the filtering process.)

delete

And preferably, (b-1) when the pixel value of a pixel having an intermediate value among the pixel values aligned with the center pixel in the mask is the same, the original pixel value (the median value of the mask coefficient) is set to the center pixel in the mask. To give a; characterized in that it further comprises.

According to the present invention, after removing a part of all the pixels in the mask and assigning different weights to the remaining pixels, there is an effect of minimizing the block effect or various noises related to the communication signal generated during the compression of the image.

According to the present invention, it is possible to realize a significantly improved image quality than the image of the conventional filter by a simple operation, there is also an effect that can be applied to various kinds of damaged image.

Before describing the details for carrying out the present invention, it should be noted that configurations that are not directly related to the technical gist of the present invention are omitted within the scope of not distracting the technical gist of the present invention.

In addition, the terms or words used in the present specification and claims are consistent with the technical spirit of the present invention on the basis of the principle that the inventor can appropriately define the concept of the term in order to explain the invention in the best way. It should be interpreted as meaning and concept.

The present invention discloses a filter processing method which can derive a remarkably improved effect over the prior art by designing a digital filter combining the advantages of an average filter and a median filter, which have been proposed in the spatial domain. That is, the pixels in the mask are aligned to remove some pixels at both ends, which are considered noise, and then filter processing is performed on the remaining pixels.

In the conventional filter, the weight is directly assigned to the position of the mask or the filter is applied with a predetermined weight value. However, the present invention discloses a filter processing method based on the pixel value of a pixel rather than depending on the position of the mask. . In other words, the selection of pixels to be applied to the filter is determined according to which pixel values the pixels in the mask have, and weights for each pixel are also specified.

Hereinafter, a method of processing a filter in a spatial domain by a combination of an alignment and an average operation according to a preferred embodiment of the present invention will be described with reference to FIGS. 3 to 6B.

3 is a flowchart illustrating a filter processing method in a spatial domain by a combination of an alignment and an average operation according to a preferred embodiment of the present invention, and FIG. 4 is an exemplary diagram of pixel values (pixel values) aligned in one dimension. FIG. 5 is an exemplary diagram of a result of weighting a selected pixel, FIG. 6A is an exemplary diagram of a 3 × 3 mask, and FIG. 6B is an exemplary diagram of pixel values (pixel values) aligned in the mask.

First, as illustrated in FIG. 3, mn pixels in the m × n mask are arranged in pixel value order (S10).

Next, it is determined whether the pixel value of the pixel having the intermediate value among the aligned pixel values is the same as the aligned pixel values (S20).

Next, as a result of the determination in step S20, when the pixel value of the pixel having the middle value among the aligned pixel values is not the same as the center pixel in the mask, as shown in FIG. 4, at both ends of the aligned pixel values. A part of the pixel value (pixel value) is removed (S30).

Communication noise is characterized by having a significantly lower or higher value than the existing pixel value. Therefore, the effect of noise can be minimized by removing a part of the pixel values (pixel values) at both ends (d by each of the pixels at the ends).

Finally, the filtered pixel values are obtained through the weighted average operation according to Equation 3 below with respect to the remaining pixel values (S40).

Number of pixels for weighted average operation = m × n-2d

Specifically, weights are assigned to form a Gaussian distribution based on the center pixel value (pixel value) weighted with respect to the center pixel among the z residual pixels and weighted to other pixels. An example of a result of weighting the selected pixels in FIG. 5 is illustrated.

As described above, an average value is obtained as shown in Equation 4 with respect to the result value of multiplying each pixel value arranged in one dimension and a weight, and the average value obtained is given as a filtered pixel value.

In Equation 4, w _i represents a weight of the (d + i) -th pixel among the aligned pixels, and R represents a pixel value obtained as a result of the filtering process.

As a result of the determination in step S20, when the pixel value of the pixel having the middle value among the aligned pixel values is the same as the center pixel in the mask, the original pixel value (the median value of the mask coefficient) is originally assigned to the center pixel in the mask ( S50).

For example, when the number 5, which is the center pixel in the mask in FIG. 6A, and the value 4, which is the intermediate value after pixel alignment in FIG. 6B, are the same, the pixel value originally owned is given as the value 5 in the mask. If this is expressed as an equation, Equation 5 below.

Hereinafter, an experimental result of a filter processing method in a spatial domain by a combination of an alignment and an average operation according to a preferred embodiment of the present invention will be described with reference to FIGS. 7A to 9C.

The experiment described above was performed using the images shown in FIGS. 7A and 7B.

FIG. 7A illustrates the original image used in the experiment, and FIG. 7B illustrates the damaged image, which shows a blocking artifact through encoding and quantization (Quantization Parameter = 8) by a binary cosine transform, and shows the performance of the filter. For comparison, the image is imparted with Salt-and-Pepper noise.

8 (a) to 9 (c) illustrate an average filter, a median filter, and a filter processing method according to a preferred embodiment of the present invention for each of the areas indicated by two squares of the damaged image as shown in FIG. 7B. The image applied to the 5x5 mask is shown. After removing 16 pixels at both ends of the total 25 pixels, each weight is assigned to follow the Gaussian distribution for 9 pixels and averaged.

The images of FIGS. 8A and 9A show smoothed results obtained by using a square average filter having a size of 25 pixels. The overall blurring phenomenon was observed over the whole image, and the boundary part was noticeably blurred.

The images of FIGS. 8B and 9B are images obtained using a median filter. As described above, the blurring phenomenon occurring in the average filter is somewhat alleviated, but it can be seen that the interface is not only very rough but also not apparent.

8 (c) and 9 (c) show the results of applying the filter processing method according to the preferred embodiment of the present invention. The roughness of the median filter has been significantly improved, resulting in smoother smoothing and the blurring of the borderline with the average filter. By using the advantage of the average filter, the block effect is reduced, and the advantage of the median filter is used to reduce the noise and blurring.

Additionally, the filter processing method according to the preferred embodiment of the present invention as described above produces different results for Gaussian noise and impulse noise. In addition, as shown in Table 1 below, each weight applied to the remaining pixels after removing a portion of the pixels is also a factor that determines the characteristics of the resultant image.

Weight (9 pixels) Random Noise (PSNR) Impulse Noise (PSNR)  2 2 2 2 5 2 2 2 2 31.0156 31.0233  2 2 3 3 5 3 3 2 2 31.0161 31.0268  1 2 3 4 5 4 3 2 1 31.0166 31.0270  1 3 6 15 20 15 6 3 1 31.0201 31.0276

In Table 1, it can be seen that as the distribution of weight gradually changes in the direction along the Gaussian distribution, the values of random noise or impulse noise also increase. When the weight distribution shows a Gaussian filter coefficient such as [1 3 6 15 20 15 6 3 1], the random or impulse noise of the original image is the highest. In addition, although both the random noise and the impulse noise have been significantly improved by the filter processing method according to the preferred embodiment of the present invention, it can be seen that the effect is more suitable for the removal of the impulse noise than the random noise.

As described above and described with reference to a preferred embodiment for illustrating the technical idea of the present invention, the present invention is not limited to the configuration and operation as shown and described as described above, it is a deviation from the scope of the technical idea It will be understood by those skilled in the art that many modifications and variations can be made to the invention without departing from the scope of the invention. Accordingly, all such suitable changes and modifications and equivalents should be considered to be within the scope of the present invention.

FIG. 1A illustrates an example of an image in which a block effect occurs through a discrete cosine transform (DCT) and a compression process. FIG.

FIG. 1B is an exemplary diagram of an image obtained by average filtering the image of FIG. 1A; FIG.

2A is an exemplary diagram of an image in which an impulse noise is generated.

FIG. 2B illustrates an example of an image obtained by performing a median filter on the image of FIG. 2A; FIG.

3 is an overall flowchart of a filter processing method in a spatial domain by a combination of an alignment and an average operation according to a preferred embodiment of the present invention.

4 is an exemplary diagram of pixel values (pixel values) aligned in one dimension.

5 is an exemplary diagram of a result of weighting a selected pixel.

6A is an illustration of a 3x3 mask.

6B is an illustration of aligned pixel values (pixel values) in a mask.

7A and 7B are original experimental subject images and images in which block effects and impulse noise are added.

8 (a) to 8 (c) are images of which an average filter, a median filter, and a filter processing method according to a preferred embodiment of the present invention are applied to a portion of the image of FIG. 7B, respectively.

9 (a) to 9 (c) are images of which an average filter, a median filter, and a filter processing method according to a preferred embodiment of the present invention are applied to other portions of the image of FIG. 7B.

Claims (7)

In the filter processing method in the spatial domain by the combination of the alignment and the average operation, (a) sorting pixels in the mask in pixel value order; (b) removing pixel values of a predetermined region from both ends of the aligned pixel values when the pixel values of pixels having an intermediate value among the aligned pixel values are not the same; And and (c) calculating a resultant pixel value by averaging the pixel values of the remaining pixels. The method of claim 1, And the average operation of the step (c) is a weighted average operation. The method of claim 2, In step (c), (c-1) weighting the residual pixels; And and (c-2) calculating an average value of a result value of multiplying each pixel value and the weight value of the remaining pixels by using a combination and arithmetic arithmetic combination. The method of claim 3, wherein And the weight is given to the center pixel among the remaining pixels to be the largest. The method of claim 3, wherein And said weights are assigned to form a Gaussian distribution based on pixel values of pixels in the center of the remaining pixels. The method of claim 3, wherein And the average value is calculated by the following equation.

(In the above equation, m and n represent the horizontal and vertical sizes of the mask, d represents the number of pixel values removed at one end of the aligned pixel values according to step (b), and z represents the ( c) represents the number of remaining pixels in step c), w _i represents a weight of the (d + i) -th pixel among the aligned pixels, and R represents a pixel value calculated as a result of the filtering process.) The method of claim 1, (b-1) if a pixel value of a pixel having an intermediate value among the aligned pixel values is the same as the center pixel in the mask, giving the original pixel value (the median of mask coefficients) to the center pixel in the mask; Filter processing method in the spatial domain by the combination and sorting average operation characterized in that it further comprises.

Images