WO2016190698A1

WO2016190698A1 - Weighted median filtering method and device for noise removal

Info

Publication number: WO2016190698A1
Application number: PCT/KR2016/005623
Authority: WO
Inventors: 이동호
Original assignee: 한양대학교 에리카산학협력단
Priority date: 2015-05-27
Filing date: 2016-05-27
Publication date: 2016-12-01

Abstract

A weighted median filtering method and device for noise removal are disclosed. The weighted median filtering method according to one embodiment of the present invention comprises the steps of: detecting noise in an image; determining the size of a weight window to be applied to the image; applying predefined weight values to the weight window of the determined size; changing, to a predetermined first value, the weight value of a pixel detected as the noise among the weight values on the basis of a noise detection result for the image; and performing weighted median filtering by using the weight values including the first value, thereby calculating a median value.

Description

Weighted Median Filtering Method and Apparatus for Noise Reduction

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to weighted median filtering, and more particularly, to a weighted median filtering method and apparatus capable of efficiently removing noise such as impulse noise and improving image quality.

Commonly used devices to remove impulse noise are median filters and weighted media filters. The median filter is a device that applies a window of a specific size to an image, sorts them in order according to the values of pixels in the window, and then outputs the median value among them. In order to improve the performance, a device for measuring the impulse noise and adaptively changing the window size according to the degree has also been proposed. A weighted median filter is proposed as an apparatus for improving the performance of such a median filter. Weighted median filter is a device that increases the probability of being adopted as a median value by weighting the pixels according to their importance in the window and repeating the pixels by the weight when sorting in order.

However, although the conventional median filter or weighted median filter is very efficient in removing impulse noise, there are some side effects or insignificant effects depending on the degree of noise. In particular, the conventional median filter or weighted median filter has a problem that the edge is deformed or distorted when there is noise in the edge portion sensitive to the human eye and the degree is severe.

Embodiments of the present invention provide a weighted median filtering method and apparatus capable of efficiently removing noise such as impulse noise to improve image quality.

Embodiments of the present invention perform weighted median filtering by changing a weight value of a pixel corresponding to an impulse noise to a specific value, thereby weighting a median that can improve image quality by preventing the edge region from being deformed or distorted. It provides a filtering method and apparatus.

Weighted median filtering method according to an embodiment of the present invention comprises the steps of detecting the noise from the image; Determining a size of a weight window to be applied to the image; Applying predefined weighting values to the weighted window of the determined size; Changing a weight value of a pixel detected as noise among the weight values to a first predetermined value based on a noise detection result of the image; And calculating weighted median values by performing weighted median filtering using the weighted values including the first value.

The determining of the size of the weighted window may determine the size of the weighted window based on a noise detection result of the image.

The determining of the size of the weight window may determine the size of the weight window based on the number of noise pixels included in each of a plurality of sizes predefined for the weight window.

The plurality of sizes includes a first size, a second size larger than the first size, and a third size larger than the second size, and the determining of the size of the weighted window applies the weighted window of the first size. Calculate the number of first pixels, the number of noise pixels, and if the calculated number of first pixels is less than or equal to a predefined first reference pixel, determine the first size as the size of the weighted window, and the number of first pixels When the number of reference pixels is greater than the number of reference pixels, the weighted window of the second size is applied to calculate the number of second pixels, which is the number of noise pixels. The size of the window may be determined. If the number of the second pixels is larger than the number of the second reference pixels, the third size may be determined as the size of the weighted window.

The first value may be zero or a predetermined value close to zero.

In accordance with another aspect of the present invention, a weighted median filtering method includes: applying a weight window having a predetermined weight to a video having a predetermined weight value; Changing a weight value of a pixel determined as noise in the weight window to a first predetermined value; And calculating weighted median values by performing weighted median filtering using the weighted values including the first value.

Furthermore, the weighted median filtering method according to another embodiment of the present invention includes detecting noise in the image; And determining a size of a weight window to be applied to the image based on a noise detection result of the image.

Weighted median filtering device according to an embodiment of the present invention includes a detector for detecting noise from the image; A determination unit to determine a size of a weight window to be applied to the image; An application unit which applies predefined weight values to the weight window of the determined size; A changer for changing a weight value of a pixel detected as noise among the weight values to a first predetermined value based on a noise detection result of the image; And a calculation unit configured to calculate the median value by performing weighted median filtering using the weight values including the first value.

The determination unit may determine the size of the weight window based on a noise detection result of the image.

The determination unit may determine the size of the weight window based on the number of noise pixels included in each of the plurality of sizes predefined for the weight window.

The plurality of sizes includes a first size, a second size larger than the first size, and a third size larger than the second size, and the determiner is configured to apply a weight window of the first size to determine the number of noise pixels. Calculate the number of pixels, determine the first size as the size of the weighted window when the calculated number of first pixels is less than or equal to a predefined first reference pixel, and when the number of first pixels is greater than the number of first reference pixels, Calculate a number of second pixels as the number of noise pixels by applying a weight window of a second size, and determine the second size as the size of the weight window when the calculated number of second pixels is less than or equal to a predefined second reference pixel, If the number of the second pixels is greater than the number of the second reference pixels, the third size may be determined as the size of the weight window.

The first value may be zero or a predetermined value close to zero.

According to embodiments of the present invention, weighted median filtering is performed by changing a weight value of a pixel corresponding to impulse noise to a specific value, thereby efficiently removing noise including impulse noise, thereby improving image quality. By preventing the edge area from being broken or distorted, an image of high quality can be obtained.

An image sensor mounted in a camera, for example, the image captured in a dark or lack of light conditions using a CMOS sensor is present in the impulse noise, embodiments of the present invention by effectively removing the noise including the impulse noise In this case, the sharper image can be restored without breaking or distorting the edge region.

Embodiments of the present invention can be applied to all products equipped with a sensor for capturing an image, for example, a device for outputting an image such as a monitor or a TV, a camera or CCTV for generating an image, and a high resolution camera. In addition to devices such as smartphones, it can be applied to many systems for processing images.

1 illustrates an example of a weight value for a weight window applied in weighted median filtering.

2 is a flowchart illustrating an operation of a weighted median filtering method according to an embodiment of the present invention.

FIG. 3 illustrates an exemplary diagram for describing FIG. 2.

4 is a flowchart illustrating an embodiment of operation S220 of FIG. 2.

5 shows an example in which a weighted median filtering device according to the present invention is applied.

6 illustrates a configuration of a weighted median filtering device according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or limited by the embodiments. Also, like reference numerals in the drawings denote like elements.

Embodiments of the present invention perform weighted median filtering by changing a weight value of a pixel corresponding to impulse noise in an image to a specific value, for example, 0, thereby efficiently removing noise including impulse noise and performing edge region. The image which is not broken or distorted may be obtained, and thus image quality may be improved.

The median filtering device is a device for arranging pixel values included in a corresponding window in order and then outputting a median value located in the middle. The median filtering device may output a median value by using Equation 1 below.

[Equation 1]

Y (i, j) = MED {X (i + q, j + t) | (q, t) ∈W}

Here, Y may mean a median value, (i, j) may mean a coordinate or position of a pixel, and W may mean a size of a window.

The weighted median filtering device applies a weight value corresponding to the size of the window according to the size of the window to be applied, and repeats the weighted value of the pixel as shown in Equation 2 below in ordering the size according to the size of the window. Calculate the value.

[Equation 2]

WM (i, j) = MED [X (i + q, j + t) ◇ W (i + q, j + t) | (q, t) ∈W]

K ◇ X = X, X,. . ., X

K times

For example, the weighted median filtering device may apply a weight value as in the example shown in FIG. 1A when the weight window size applied to the image is 3 × 3 size, and the size of the weight window applied to the image is 5 In the case of the x5 size, a weight value as in the example illustrated in FIG. 1B may be applied.

Here, the weight value to be applied is a value determined in advance according to the size of the weight window, and the highest weight value is set at the center of the window, and the weight value is lowered from the center. As described above, conventional weighted median filtering applies weight values that are already determined regardless of whether the corresponding pixel is noise or the like, so that the image quality may be deteriorated when a high weight value is applied to the pixel corresponding to the noise.

According to embodiments of the present invention, a weight value for a pixel determined to be noise, such as an impulse noise, in a predetermined weight value applied to such a weighted median filtering device is set to '0' or a predetermined very small value close to '0'. By performing weighted median filtering after the change, it is possible to effectively remove noise such as an impulse noise and obtain a good quality image without cracking or distorting the edge region.

2 is a flowchart illustrating a weighted median filtering method according to an embodiment of the present invention, and FIG. 3 is a diagram for explaining FIG. 2.

2 and 3, the weighted median filtering method according to an embodiment of the present invention detects noise included in an image when an image including noise such as an impulse noise is received or input (S210).

That is, step S210 can detect whether or not noise for all the pixels constituting the image, the method of detecting the noise included in the image can be known to those skilled in the art in this technology distribution, All methods can be applied.

Next, the size (W × W) of the weighted window for performing weighted median filtering is determined by applying to the received image, and weight values corresponding to the size (W × W) of the determined weighted window are applied (S220 and S230). ).

For example, if the size of the weight window is determined to be 3 × 3 size in step S220, step S230 applies predetermined weight values as shown in FIG. 1A for the 3 × 3 size, and the size of the weight window is 5 × in step S220. If it is determined as 5 size, step S230 applies predetermined weight values as shown in FIG. 1B for 5 × 5 size.

In this case, step S220 may determine the size of the weight window to be applied to the image based on the noise detection result detected by step S210, and the step of determining the size of the weight window may be determined whenever the weight window is applied. Step S220 will be described in detail with reference to FIG. 3.

When the size and weight values of the weight window to be applied are determined by steps S220 and S230, the weight value of the pixel determined as noise such as an impulse noise among pixels in the image area to which the weight window is applied is changed to a first predetermined value ( S240).

That is, in step S240, the weight value applied to the pixel including the noise such as the impulse noise among the weight values applied by the step S230 is a very small value close to the first value, for example, '0' or '0'. By changing this, it is possible to remove or minimize the influence on the pixel determined as noise when performing the weighted median filtering.

In step S240, the weight value of the pixel position determined as noise, such as an impulse noise, among the weight values of the weight window is changed to a first value close to '0' or '0' to complete the weight values applied to the weight window. By performing weighted median filtering using the weighted values, a median value (or weighted median value) is calculated (S250 and S260).

That is, steps S250 and S260 calculate a median value by using the weight value determined by step S240 for an image corresponding to the size of the window centering on the pixel corresponding to the noise to perform weighted median filtering in the image. The frozen value calculates a median value after repeating the pixel value corresponding to the weight value as shown in Equation 2 above.

Through this process, it is possible to remove noise from an image including noise and obtain an image from which the noise is removed. The obtained image is an image in which edge regions are not broken or distorted, and the product or system to which the present invention is applied. Can provide high quality images.

4 is a flowchart illustrating an embodiment of operation S220 of FIG. 2.

4 illustrates a process of determining a size of a weighted window based on a noise detection result of an image. The size of the weighted window is determined based on the number of noise pixels included in each of a plurality of predefined sizes for the weighted window. The decision is made by adaptively applying the size of the weighted window to the noise level.

Referring to FIG. 4, in operation S220, the size of the weighted window may be determined by applying a weighted window of a first size, for example, 3 × 3, to determine the number of noise pixels such as an impulse noise in the weighted window of the first size. 1 pixel number) and compares the calculated first pixel number with a predetermined first reference value (or first reference pixel number) to determine whether the first pixel number is equal to or less than the first reference pixel number (S410 and S420). .

Here, when the first size is 3 × 3 size, the first reference value may be 3.

If it is determined in step S420 that the number of first pixels is less than or equal to the first reference pixel number, the first size is determined as the weight window size for performing weighted median filtering (S430).

On the other hand, if it is determined in step S430 that the number of the first pixels is larger than the number of the first reference pixels, the number of noise pixels such as an impulse noise in the weight window of the second size is applied by applying a weight window of the second size, for example, 5 × 5. Calculate the number of second pixels and compare the calculated number of second pixels with a predetermined second reference value (or number of second reference pixels) to determine whether the number of second pixels is less than or equal to the number of second reference pixels (S440, S450).

Here, when the second size is 5 × 5 size, the second reference value may be 12.

If it is determined in step S450 that the number of second pixels is less than or equal to the second reference pixel number, the second size is determined as the size of the weighted window for performing weighted median filtering (S460).

On the other hand, if it is determined in step S530 that the number of second pixels is larger than the number of second reference pixels, a third size, for example, 7 × 7 size, is determined as the weight window size for weighted median filtering (S470).

As described above, in the embodiments of the present invention, after detecting noise in an image, a weight window of size 3 × 3, which is a first size, is first applied to determine whether to determine the first size according to the number of noises, and to determine the first size. A process of determining whether a second size is determined according to the number of noises by applying a weight window of 5 × 5 size, which is a second size, if not suitable, and determining a size of 7 × 7, which is a third size, when it is inappropriate to determine a second size By performing the sizing process of the adaptive weighting window, the noise included in the image can be effectively removed.

With the development of semiconductor technology, image capturing sensors embedded in cameras are applied with CMOS technology, and high-resolution sensors are showing more and more. When shooting in the dark or lack of light due to the characteristics of the sensor, the captured image includes a lot of noise including an impulse noise, and the present invention can effectively remove noise such as an impulse noise. Applicable to all products or systems.

For example, a weighted median filtering device 500 performing a method according to an embodiment of the present invention is applied after a CMOS image sensor for capturing an image to remove noise as shown in FIG. 5A. An image may be provided, or as shown in FIG. 5B, an image including noise output from a camera may be received to effectively remove noise from the received image, and then display an image from which the noise is removed.

FIG. 6 illustrates a configuration of a weighted median filtering apparatus according to an embodiment of the present invention, and illustrates a configuration of an apparatus for performing the operations of FIGS. 2 to 4.

Referring to FIG. 6, the weighted median filtering apparatus 500 according to an embodiment of the present invention includes a receiver 610, a detector 620, a determiner 630, an applier 640, and a changer 650. And a calculator 660.

The receiver 610 receives an image for performing weighted median filtering of the present invention.

In this case, the receiver 610 may receive an image including noise such as an impulse noise.

The detector 620 detects noise such as an impulse noise in an image received by the receiver 610.

At this time, the detector 620 may detect whether or not all pixels constituting the image is noise, and a method of detecting noise included in the image may be known to those skilled in the art. Any method of detecting can be applied.

The determiner 630 determines the size of the weighted window for performing weighted median filtering by applying to the image.

In this case, the determination unit 630 may determine the size of the weight window to be applied to the image based on the noise detection result detected by the detection unit 620, for example, a plurality of predefined sizes for the weight window, Based on the number of noise pixels included in each of the 3x3 size and the 5x5 size, the size of the weighted window may be determined, for example, one of 3x3 size, 5x5 size, and 7x7 size. The determiner 630 may adaptively determine the size of the weight window according to the noise level.

The determiner 630 calculates the first pixel number, which is the number of noise pixels such as an impulse noise in the weight window of the first size by applying a weight window of the first size, for example, 3 × 3, and calculates the first pixel. If the number is less than or equal to the first reference pixel, the first size is determined as the size of the weighted window for performing weighted median filtering. If the number of the first pixels is larger than the number of the first reference pixels, the second size is, for example, 5 × 5 size. A second weighted number is calculated by applying a weighted window to the number of noise pixels such as an impulse noise in the weighted window of the second size, and when the calculated number of second pixels is less than or equal to the second reference pixel, weighted median filtering is performed on the second size. If the number of the second pixels is larger than the number of the second reference pixels, the size of the weighted window is determined as the size of the weighted window for performing weighted median filtering.

The application unit 640 applies weight values corresponding to the size of the weight window determined by the determination unit 630 to the determined weight window.

The changer 650 determines a first weight value of a pixel determined as noise, such as an impulse noise, among pixels in the image region to which the weight window is applied based on the noise detection result of the image detected by the detector 620. Change it to a value.

That is, the changer 650 may assign a weight value applied to a pixel including noise such as an impulse noise among the weight values applied by the applier 640 to a first value, for example, '0' or '0'. Change it to a very small value that comes close. As described above, the changer 650 changes the weight value of the pixel determined as noise to a very small value close to '0' or '0', so that the weighted median filtering is performed on the pixel determined as noise. The impact can be eliminated or minimized.

The calculation unit 660 changes the weight value of the pixel position determined as noise, such as an impulse noise, among the weight values of the weight window by the changer 650 to a first value close to '0' or '0'. When the weight values applied to are completed, the median value is calculated by performing weighted median filtering using the completed weight values.

In this case, the calculator 660 may repeat the pixel value corresponding to the weight value as shown in Equation 2 and then calculate the median value.

The system or apparatus described above may be implemented with hardware components, software components, and / or combinations of hardware components and software components. For example, the systems, devices, and components described in the embodiments may include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable arrays (FPAs). ), A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions, may be implemented using one or more general purpose or special purpose computers. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of explanation, one processing device may be described as being used, but one of ordinary skill in the art will appreciate that the processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. Or may be permanently or temporarily embodied in a signal wave to be transmitted. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

The method according to the embodiments may be embodied in the form of program instructions that may be executed by various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims that follow.

Claims

Detecting noise from the image;

Determining a size of a weight window to be applied to the image;

Applying predefined weighting values to the weighted window of the determined size;

Changing a weight value of a pixel detected as noise among the weight values to a first predetermined value based on a noise detection result of the image; And

Calculating a median value by performing weighted median filtering using the weighted values including the first value

Weighted median filtering method comprising a.
The method of claim 1,

Determining the size of the weight window

Weighted median filtering method for determining the size of the weighted window based on a noise detection result for the image.
The method of claim 1,

Determining the size of the weight window

Weighted median filtering method for determining the size of the weighted window based on the number of noise pixels included in each of a plurality of predefined sizes for the weighted window.
The method of claim 3,

The plurality of sizes

A first size, a second size greater than the first size, and a third size greater than the second size,

Determining the size of the weight window

Calculate the number of first pixels, the number of noise pixels, by applying the weight window of the first size, and determine the first size as the size of the weight window when the calculated number of first pixels is less than or equal to a predefined first reference pixel. ,

If the number of the first pixels is greater than the number of the first reference pixels, the number of second pixels, which is the number of noise pixels, is calculated by applying the weight window of the second size, and the calculated number of second pixels is less than or equal to a predefined second reference pixel number. And determining the second size as the size of the weight window.

And if the number of the second pixels is greater than the number of the second reference pixels, determining the third size as the size of the weight window.
The method of claim 1,

The first value is

A weighted median filtering method that is zero or a predetermined value near zero.
Applying a weight window of a predetermined size having a predetermined weight value to the image;

Changing a weight value of a pixel determined as noise in the weight window to a first predetermined value; And

Calculating a median value by performing weighted median filtering using the weighted values including the first value

Weighted median filtering method comprising a.
The method of claim 6,

Detecting noise in the image; And

Determining a size of a weight window to be applied to the image based on a noise detection result of the image;

Weighted median filtering method comprising more.
The method of claim 7, wherein

Determining the size of the weight window

Weighted median filtering method for determining the size of the weighted window based on the number of noise pixels included in each of a plurality of predefined sizes for the weighted window.
The method of claim 8,

The plurality of sizes

A first size, a second size greater than the first size, and a third size greater than the second size,

Determining the size of the weight window

Calculate the number of first pixels, the number of noise pixels, by applying the weight window of the first size, and determine the first size as the size of the weight window when the calculated number of first pixels is less than or equal to a predefined first reference pixel. ,

If the number of the first pixels is greater than the number of the first reference pixels, the number of second pixels, which is the number of noise pixels, is calculated by applying the weight window of the second size, and the calculated number of second pixels is less than or equal to a predefined second reference pixel number. And determining the second size as the size of the weight window.

And if the number of the second pixels is greater than the number of the second reference pixels, determining the third size as the size of the weight window.
A detector for detecting noise from an image;

A determination unit to determine a size of a weight window to be applied to the image;

An application unit which applies predefined weight values to the weight window of the determined size;

A changer for changing a weight value of a pixel detected as noise among the weight values to a first predetermined value based on a noise detection result of the image; And

Calculation unit for calculating the median value by performing weighted median filtering using the weighted values including the first value

Weighted median filtering device comprising a.
The method of claim 10,

The determination unit

Weighted median filtering device for determining the size of the weight window based on a noise detection result for the image.
The method of claim 10,

The determination unit

Weighted median filtering device for determining the size of the weighted window based on the number of noise pixels included in each of the plurality of sizes predefined for the weighted window.
The method of claim 12,

The plurality of sizes

A first size, a second size greater than the first size, and a third size greater than the second size,

The determination unit

Calculate the number of first pixels, the number of noise pixels, by applying the weight window of the first size, and determine the first size as the size of the weight window when the calculated number of first pixels is less than or equal to a predefined first reference pixel. ,

If the number of the first pixels is greater than the number of the first reference pixels, the number of second pixels, which is the number of noise pixels, is calculated by applying the weight window of the second size, and the calculated number of second pixels is less than or equal to a predefined second reference pixel number. And determining the second size as the size of the weight window.

And the third size is determined as the size of the weighted window when the number of the second pixels is greater than the number of the second reference pixels.
The method of claim 10,

The first value is

Weighted median filtering device that is a predetermined value near zero or zero.