CN101308573B - Method and apparatus for eliminating noise - Google Patents

Abstract

The invention provides a method to eliminate noise, comprising the following steps: comparing a central pixel point and a neighborhood middle pixel point; determining whether the central pixel point is a noise point based on the comparison result; if the central pixel point is a noise point, replacing the pixel value of the central pixel point with a weighted average of the pixel value of the neighborhood middle pixel point; if not, replacing the pixel value of the central pixel point with a weighted average between the pixel value of the neighborhood middle pixel point and the pixel value of the central pixel point. The method effectively eliminates the noise and well maintains the image edge information.

Description

A kind of method and apparatus of eliminating noise

Technical field

The present invention relates to technical field of image processing, particularly a kind of method and apparatus of eliminating noise.

Technical background

Existing a kind of method simple, that have big reduction noise contributions is to adopt the method for low-pass filter, and when from different point observation, low-pass filter adopts the mean value of observation pixel and the neighbor around it, as the new value of observation pixel.With regard to this method, the value of observation pixel does not have big variation, but has the random noise composition and the noise contribution value of averaging that is included in the surrounding pixel of non-correlation, and the value of this composition is similar to " 0 ".

Therefore, when the method above the employing, the Noise Suppression effect strengthens along with the search area expansion of surrounding pixel,, do average calculating operation with surrounding pixel, the information of image border also is similar to noise and has equally reduced, its result, though noise has reduced, entire image has thickened, i.e. the debase of image.

Summary of the invention

In view of this, the object of the present invention is to provide a kind of method and device thereof of eliminating noise, on the basis of eliminating noise effectively, kept edge of image simultaneously well.

In order to achieve the above object, the invention provides a kind of method of eliminating noise, may further comprise the steps:

Obtain the difference of the pixel value of pixel in the pixel value of central pixel point and the neighborhood, with the absolute value of difference and predetermined threshold value relatively, if comparative result for all greater than, then central pixel point is a noise; Otherwise central pixel point is non-noise;

If central pixel point is a noise, then the pixel value of central pixel point is replaced by the weighted mean of pixel pixel value in the neighborhood; Be specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j ≠ (0,0) promptly belongs to the pixel in the neighborhood, and does not comprise central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood;

Otherwise the pixel value of central pixel point is replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood, is specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j=(0,0) promptly belongs to the pixel in the neighborhood, and comprises central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

As one embodiment of the present of invention, the pixel distance of the space length of the weight of each pixel and this pixel and central pixel point and this pixel and central pixel point is relevant in the neighborhood, and described space length and pixel distance are big more, and the weight of pixel is more little in the neighborhood.

In the above-described embodiments,

$\mathrm{wdd}=\exp (-\frac{{\mathrm{dd}}^2}{2{\left(\mathrm{sigmad}\right)}^2})$

Wherein, sigmad is a threshold parameter, and dd is the space length of pixel in central pixel point and the neighborhood.

In the above-described embodiments,

$\mathrm{wdc}=\exp (-\frac{{\mathrm{dc}}^2}{2{\left(\mathrm{sigmac}\right)}^2})$

Wherein, sigmac is a threshold parameter, and dc is the pixel distance of pixel in central pixel point and the neighborhood.

As one embodiment of the present of invention, dc can obtain by the color component of pixel in central pixel point and the neighborhood,

Or luminance component and chromatic component by pixel in central pixel point and the neighborhood obtain.

The present invention also provides a kind of device of eliminating noise, and this device comprises:

Comparison module is used for obtaining the difference of the pixel value of the pixel value of central pixel point and neighborhood pixel, with the absolute value of difference and predetermined threshold value relatively;

Judge module is used for judging according to relatively result whether central pixel point is noise, if comparative result for all greater than, then central pixel point is a noise; Otherwise central pixel point is non-noise;

Replace module, replace the pixel value of central pixel point according to the judged result of judge module, if central pixel point is a noise, then the pixel value of central pixel point is replaced by the weighted mean of pixel pixel value in the neighborhood; Be specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j ≠ (0,0) promptly belongs to the pixel in the neighborhood, and does not comprise central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood;

Otherwise the pixel value of central pixel point is replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood, is specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j=(0,0) promptly belongs to the pixel in the neighborhood, and comprises central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

As one embodiment of the present of invention, described replacement module further comprises:

The weighted mean acquiring unit is used for obtaining the weighted mean of neighborhood pixel pixel value, or is used for obtaining the weighted mean of neighborhood pixel and central pixel point pixel value;

The weighting function acquiring unit is used for obtaining the space length or the pixel distance of central pixel point and neighborhood pixel.

In the above-described embodiments, the weighting function acquiring unit can obtain pixel distance by the color component of pixel in central pixel point and the neighborhood; Or,

The weighting function acquiring unit can obtain pixel distance by the luminance component and the chromatic component of pixel in central pixel point and the neighborhood.

The present invention is by the comparison of central pixel point and neighborhood territory pixel point, and then judge whether central pixel point is noise, according to the method and the device of judged result replacement central pixel point pixel value, effectively eliminated noise, kept image edge information simultaneously well.

Description of drawings

Fig. 1 is the process flow diagram of method of eliminating audible noise of the present invention;

Fig. 2 is the structured flowchart that the present invention eliminates the noise device;

Fig. 3 is that the present invention eliminates the structured flowchart of replacing module in the noise device.

Embodiment

For making purpose of the present invention clearer, reach specific embodiment now in conjunction with the accompanying drawings and be described.Below by the embodiment that is described with reference to the drawings is exemplary, only is used to explain the present invention, and can not be interpreted as limitation of the present invention.

Fig. 1 is the process flow diagram of method of eliminating audible noise of the present invention.

As shown in Figure 1, the process of noise removing realizes as follows:

S101, pixel in Correlation Centre pixel and the neighborhood;

The coordinate figure of central pixel point can be expressed as (x, y), in the neighborhood coordinate figure corresponding tables of each pixel be shown (x+i, y+j), j represents the line number of distance center pixel uplink and downlink, i represents the line number of the left and right row of distance center pixel; Wherein, neighborhood comprises the pixel that distance center pixel upper and lower, left and right line position all equates; Perhaps, described neighborhood comprises pixel and the equal pixel of left and right line position that distance center pixel uplink and downlink position equates.Such as, the coordinate figure of central pixel point (0,0)-7≤i≤7 ,-7≤j≤7, then pixel is that 1-7 is capable on the distance center pixel in the neighborhood, following 1-7 is capable, all pixels during the capable and right 1-7 of left 1-7 is capable.

The pixel value of each pixel can have multiple expression way in central pixel point and the neighborhood, both can represent by 3 color components of rgb space, also can realize by the distortion in colourity and the brightness space.Therefore, contain based on all should be protection domain of the present invention with the variation that does not break away from inventive concept.

S102 judges according to result relatively whether central pixel point is noise; If central pixel point is a noise, execution in step S103 then, otherwise, execution in step S104;

Pixel in Correlation Centre pixel and the neighborhood, judge according to relatively result whether central pixel point is that noise is: the difference of obtaining the pixel value of pixel in the pixel value of central pixel point and the neighborhood, the absolute value and the predetermined threshold value of difference are compared, if comparative result for all greater than, then central pixel point is a noise; Otherwise central pixel point is non-noise.Also can carry out according to the method that detects noise in the prior art.

S103, the pixel value of central pixel point is replaced by the weighted mean of pixel pixel value in the neighborhood;

The pixel distance of the space length of the weight of each pixel and this pixel and central pixel point and this pixel and central pixel point is relevant in the neighborhood, and described space length and pixel distance are big more, and the weight of pixel is more little in the neighborhood.

Wherein, the process of replacement can be used following formulate:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j ≠ (0,0) promptly belongs to pixel in the neighborhood, and does not comprise central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

$\mathrm{wdd}=\exp (-\frac{{\mathrm{dd}}^2}{2{\left(\mathrm{sigmad}\right)}^2})$

Wherein, sigmad is a threshold parameter, and dd is the space length of pixel in central pixel point and the neighborhood.

$\mathrm{wdc}=\exp (-\frac{{\mathrm{dc}}^2}{2{\left(\mathrm{sigmac}\right)}^2})$

Wherein, sigmac is a threshold parameter, and dc is the pixel distance of pixel in central pixel point and the neighborhood.

Dc can obtain by the color component of pixel in central pixel point and the neighborhood, or luminance component and chromatic component by pixel in central pixel point and the neighborhood obtain.

Below describe the process that central pixel point is replaced in detail:

According to the coordinate figure of each pixel in central pixel point described in the S101 and the neighborhood, then in central pixel point and the neighborhood space length of pixel square be:

Dd ²=i ²+ j ², further can calculate the value of wdd.

If color is the color component in the pixel rgb space, then in central pixel point and the neighborhood pixel distance of pixel square be:

dc ²＝(R(x+i，y+j)-R(x，y)) ²+(G(x+i，y+j)-G(x，y)) ²+(B(x+i，y+j)-B(x，y)) ²

Certainly, can further be translated into the easily Lab space of perception of human eye, expression formula is as follows:

dc ²＝(L(x+i，y+j)-L(x，y)) ²+(a(x+i，y+j)-a(x，y)) ²+(b(x+i，y+j)-b(x，y)) ²

In order to reduce the calculated amount of rgb space, can be undertaken by following formula to the Lab space conversion:

L＝(R+2G+B)/4；a＝G-R；b＝G-B；

In like manner, also can between other color components, luminance component or chromatic component, transform mutually, such as,

Y＝c11*R+c12*G+c13*B+c14；

Cr＝c21*R+c22*G+c23*B+c24；

Cb＝c31*R+c32*G+c33*B+c34；

Therefore, contain based on all should be protection domain of the present invention with the variation that does not break away from inventive concept.

And, give weight to above-mentioned human eye than the dc value in the Lab space of easy perception, expression formula is as follows: dc ²=p1* (L (x+i, y+j)-L (x, y)) ²+ p2* (a (x+i, y+j)-a (x, y)) ²+ p3* (b (x+i, y+j)-b (x, y)) ²Wherein, P1 is the weight to the brightness distance, and p2, p3 are the weights to chrominance distance; General p2=p3, and p1＞p2, p1＞p3; Such as, p1=0.5, p2=0.25, p3=0.25; Because human eye is more responsive to brightness.

Equally, other color spaces can carry out corresponding weights distortion conversion.

According to dc, further can obtain $\mathrm{wdc}=\exp (-\frac{{\mathrm{dc}}^2}{2{\left(\mathrm{sigmac}\right)}^2}),$ In the Lab space, the pixel value of central pixel point is replaced by like this:

S104, the pixel value of central pixel point are replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood.

The pixel distance of the space length of the weight of each pixel and this pixel and central pixel point and this pixel and central pixel point is relevant in the neighborhood, and described space length and pixel distance are big more, and the weight of pixel is more little in the neighborhood.

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood, and i, j=(0,0) promptly belongs to the pixel in the neighborhood, and comprises central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

In the Lab space, the pixel value of central pixel point is replaced by equally:

Be replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood by aforementioned calculation central pixel point pixel value, and the pixel distance of the space length of the weight of each pixel and this pixel and central pixel point and this pixel and central pixel point is relevant in central pixel point and the neighborhood, and described space length and pixel distance are big more, and the weight of pixel is more little in the neighborhood.

Both can be weighted mean to all pixels in the neighborhood to the weighted mean of pixel in the neighborhood among above S103 or the S104, also can be that the pixel on a certain direction in the field by central pixel point is weighted on average.

Fig. 2 is the structured flowchart that the present invention eliminates the noise device.

As shown in Figure 2, eliminating noise device 200 comprises comparison module 210, judge module 220 and replaces module 230.

Wherein, comparison module 210 is used for Correlation Centre pixel and neighborhood pixel; Judge module 220 is used for judging according to result relatively whether central pixel point is noise; Replace module 230, replace the pixel value of central pixel point according to the judged result of judge module.

Fig. 3 is that the present invention eliminates the structured flowchart of replacing module 230 in the noise device.

As shown in Figure 3, replace module 230 and further comprise weighted mean acquiring unit 231 and weighting function acquiring unit 232, wherein:

Weighted mean acquiring unit 231 is used for obtaining the weighted mean of neighborhood pixel pixel value, or is used for obtaining the weighted mean of neighborhood pixel and central pixel point pixel value;

Weighting function acquiring unit 232 is used for obtaining the space length or the pixel distance of central pixel point and neighborhood pixel.

Wherein, weighting function acquiring unit 232 can obtain pixel distance by the color component of pixel in central pixel point and the neighborhood; Or,

Weighting function acquiring unit 232 can obtain pixel distance by the luminance component and the chromatic component of pixel in central pixel point and the neighborhood.

The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement etc., all should be included within protection scope of the present invention.

Claims (8)

1. a method of eliminating noise is characterized in that, may further comprise the steps:

Obtain the difference of the pixel value of pixel in the pixel value of central pixel point and the neighborhood, with the absolute value of difference and predetermined threshold value relatively, if comparative result for all greater than, then central pixel point is a noise; Otherwise central pixel point is non-noise;

If central pixel point is a noise, then the pixel value of central pixel point is replaced by the weighted mean of pixel pixel value in the neighborhood; Be specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j ≠ (0,0) promptly belongs to the pixel in the neighborhood, and does not comprise central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood;

Otherwise the pixel value of central pixel point is replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood, is specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j=(0,0) promptly belongs to the pixel in the neighborhood, and comprises central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

2. the method for elimination noise according to claim 1, it is characterized in that, the pixel distance of the space length of the weight of each pixel and this pixel and central pixel point and this pixel and central pixel point is relevant in the neighborhood, and described space length and pixel distance are big more, and the weight of pixel is more little in the neighborhood.

3. the method for elimination noise according to claim 1 is characterized in that,

$\mathrm{wdd}=\exp (-\frac{{\mathrm{dd}}^2}{2{\left(\mathrm{sigmad}\right)}^2})$

Wherein, sigmad is a threshold parameter, and dd is the space length of pixel in central pixel point and the neighborhood.

4. the method for elimination noise according to claim 1 is characterized in that,

$\mathrm{wdc}=\exp (-\frac{{\mathrm{dc}}^2}{2{\left(\mathrm{sigmac}\right)}^2})$

Wherein, sigmac is a threshold parameter, and dc is the pixel distance of pixel in central pixel point and the neighborhood.

5. the method for elimination noise according to claim 4 is characterized in that, dc can obtain by the color component of pixel in central pixel point and the neighborhood,

Or luminance component and chromatic component by pixel in central pixel point and the neighborhood obtain.

6. a device of eliminating noise is characterized in that, this device comprises:

Comparison module is used for obtaining the difference of the pixel value of the pixel value of central pixel point and neighborhood pixel, with the absolute value of difference and predetermined threshold value relatively;

Judge module is used for judging according to relatively result whether central pixel point is noise, if comparative result for all greater than, then central pixel point is a noise; Otherwise central pixel point is non-noise;

Replace module, replace the pixel value of central pixel point according to the judged result of judge module, if central pixel point is a noise, then the pixel value of central pixel point is replaced by the weighted mean of pixel pixel value in the neighborhood; Be specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j ≠ (0,0) promptly belongs to the pixel in the neighborhood, and does not comprise central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood;

Otherwise the pixel value of central pixel point is replaced by the weighted mean of pixel and central pixel point pixel value in the neighborhood, is specially:

Wherein, (x, y color) are the pixel value after the central pixel point replacement to out, and (x y) is the center pixel point coordinate, and color is color component, chromatic component or the luminance component of pixel;

In (x+i, y+j color) are the pixel value of pixel in the neighborhood, wherein, (x+i y+j) is pixel coordinate in the neighborhood, i, and j ∈ neighborhood and i, j=(0,0) promptly belongs to the pixel in the neighborhood, and comprises central pixel point;

Wdd is the weighting function of pixel and central pixel point space length in the neighborhood, and wdc is the weighting function of pixel and central pixel point pixel distance in the neighborhood.

7. the device of elimination noise according to claim 6 is characterized in that, described replacement module further comprises:

The weighted mean acquiring unit is used for obtaining the weighted mean of neighborhood pixel pixel value, or is used for obtaining the weighted mean of neighborhood pixel and central pixel point pixel value;

The weighting function acquiring unit is used for obtaining the space length or the pixel distance of central pixel point and neighborhood pixel.

8. the device of elimination noise according to claim 7 is characterized in that, the weighting function acquiring unit can obtain pixel distance by the color component of pixel in central pixel point and the neighborhood; Or,

The weighting function acquiring unit can obtain pixel distance by the luminance component and the chromatic component of pixel in central pixel point and the neighborhood.

Images