CN1929552A - Spatial domain pixel data processing method - Google Patents
Spatial domain pixel data processing method Download PDFInfo
- Publication number
- CN1929552A CN1929552A CN 200610021987 CN200610021987A CN1929552A CN 1929552 A CN1929552 A CN 1929552A CN 200610021987 CN200610021987 CN 200610021987 CN 200610021987 A CN200610021987 A CN 200610021987A CN 1929552 A CN1929552 A CN 1929552A
- Authority
- CN
- China
- Prior art keywords
- pixel
- neighborhood
- pixel data
- processing method
- spatial domain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Processing (AREA)
Abstract
This invention relates to digital image signal process aiming at current filter method and discloses one pixel data process method to filter multiple noises. This invention integrates even value linear filter method to compute middle values to combine these two to filter out noises effectively to ensure image details.
Description
Technical field
The present invention relates to digital image signal process aiming, particularly the method for the filtering noise reduction of pixel data and reduction enhancing.
Background technology
One width of cloth digital picture is made up of the pixel (pixel) of limited size, and pixel has reflected the color and the monochrome information of image specific location.In order to handle, generally adopt the such discrete data structure of matrix to express image with computer.We can represent a width of cloth digital picture with the bidimensional matrix, and element wherein is exactly a pixel, and its quantity is integer.Pixel data is corresponding to the brightness of pixel and the quantization level in the chromaticity range.The important set of being made up of some pixels that are adjacent to each other in picture element matrix, we are referred to as neighborhood.
View data is subjected to the influence of the physical characteristic of circuit hardware in image acquisition and storage, the processing procedure, and the influence of factor such as various interference signals in the transmission channel and being polluted, and shows on the image to be exactly to have occurred noise on the picture.For the image variation that reduces these influences and cause thereupon, perhaps in order to recover ruined image, perhaps only be to wish to strengthen image with outstanding wherein useful feature, just need use digital filter view data is handled.
According to different characteristics, picture noise has multiple.(uncorrelated such as additive noise with image intensity signal, usually in transmission course, introduce), multiplicative noise is (relevant with picture signal, often the variation with picture signal changes), quantizing noise (producing in the digital image quantification process), salt-pepper noise (generally by introducing in the processes such as image segmentation, transform domain processing), impulsive noise (also be " impulse noise ", positive pulse one by one that promptly in image, superposes or negative pulse gray scale catastrophe point), Gaussian noise (a kind of intensity is obeyed the random noise of Gauss or normal distribution, for example electronic jamming of video camera) etc.
Common image filtering method has in image residing spatial domain itself and carries out, and also has view data forwarded to later on through conversion such as Fouriers to carry out in the frequency domain.Filtering in its frequency domain need relate to complicated territory translation operation, implements comparatively speaking and can expend more resources and time.
The filter that uses in spatial domain has linear and non-linear branch again.Method commonly used in the linear filter is, all in order to this point certain for the center, or the average gray of several neighborhoods (or the later average gray of weighting) replaces the gray value of this point to all pixels.Method commonly used in the nonlinear filter is, all in order to this point certain for the center, or the gray scale median of several neighborhoods (or the median after the weighting) replaces the gray value of this point to all pixels.In addition, also have the data to all pixels, all getting with it is that the lowest mean square difference of pixel data of one or several neighborhood of central point is the method for its output valve, the Mean Method of getting weight coefficient with Gaussian function, and the method for the repeatedly iteration of mean value or median, or the like.
Practice shows, though above method all has the filtering noise reduction to a certain extent, its limitation is arranged all separately regrettably.The filtering method of asking weighted average such as linearity is apparent in view on loss of detail; The method paired pulses interference of getting median is very effective, and powerless to Gaussian noise; Though it is effective to Gaussian noise to get the method for mean square deviation, the filtering of paired pulses noise can not show a candle to the method for getting intermediate value, and calculates relative complex.
In fact, the noise in the image often with the signal weave in, if Filtering Processing is improper, it is unclear that the details of image itself such as boundary profile, lines etc. are thickened, and reduces picture quality on the contrary.In addition, because all kinds of noise characteristic differences, the influence to image that embodies is also different, and these noises often are not single existence, but several whiles and depositing, so, can not reach satisfied effect if adopt a kind ofly merely at the relatively effective filter of a certain noise like model.
Summary of the invention
Technical problem to be solved by this invention is exactly the filtering method at prior art, and is with strong points, and the shortcoming of integrated filter weak effect provides a kind of picture element data processing method that can leach multiple noise.
The present invention solve the technical problem, and the technical scheme of employing is that spatial domain pixel data processing method may further comprise the steps:
(1) initialization noise decision threshold T1, T2 and filter factor a, b;
(2) judge pixel f according to threshold value T1, T2
xWhether be noise spot;
(3) as judging pixel f
xBe noise spot, enter following steps, otherwise output pixel f
xData;
(4) calculating pixel f
xThe weighted average g1 of the pixel data in the neighborhood;
(5) calculating pixel f
xThe median g2 of the pixel data in the neighborhood;
(6) with a * g1+b * g2 as pixel f
xData output.
The invention has the beneficial effects as follows that can be simultaneously carry out filtering at the noise of number of different types, operand is little, program is simple, can keep image detail preferably.
Description of drawings
Fig. 1 is with f
xSchematic diagram for 3 * 3 rectangular neighborhoods at center;
Fig. 2 is with f
xSchematic diagram for the cross neighborhood at center;
Fig. 3 is with f
xSchematic diagram for the X-shaped neighborhood at center;
Fig. 4 is the flow chart of embodiment.
Embodiment
Below in conjunction with drawings and Examples, describe technical scheme of the present invention in detail.
The present invention has not only taken into full account the validity to gaussian additive noise such as (or weighted mean) linear filter method of averaging; also consider simultaneously and ask the validity of intermediate value non-linear filtering methods such as (or weighted medians) the good protection and the paired pulses salt-pepper noise of details; the two is combined; thereby reach the effectively multiple noise of filtering, can guarantee image detail purpose clearly simultaneously again.In addition, before Filtering Processing, carry out noise earlier and judge, only handle then at noise spot, thus the negative effect of having avoided filtering operation that useful signal is brought.
Technical scheme of the present invention is that spatial domain pixel data processing method may further comprise the steps:
(1) initialization noise decision threshold T1, T2 and filter factor a, b;
(2) judge pixel f according to threshold value T1, T2
xWhether be noise spot;
(3) as judging pixel f
xBe noise spot, enter following steps, otherwise output pixel f
xData;
(4) calculating pixel f
xThe weighted average g1 of the pixel data in the neighborhood;
(5) calculating pixel f
xThe median g2 of the pixel data in the neighborhood;
(6) with a * g1+b * g2 as pixel f
xData output;
Concrete is that described pixel is the pixel of the single picture frame of rest image or moving image;
More particularly, the Y channel data of described pixel data be pixel gray value, luminous value, brightness value or yuv space;
Concrete noise determination methods is calculating pixel f
xOther pixels f in data and its neighborhood
iThe difference of data when difference surpasses T2 greater than the number of pixels of T1, is judged pixel f
xBe noise spot;
In the step (1), the span of recommendation is: noise decision threshold T1:30~60; Noise judgment threshold T2:190~240; Filter factor a, b should satisfy: a/b:0~1, the condition of a+b≤1 simultaneously.
The weight coefficient of concrete weighted average is 1/9 type, 1/16 type, 1/25 type, 1/49 type or Gaussian function type;
Further be that described neighborhood is with pixel f
xRectangular neighborhood for the n * m at center; Wherein n, m are the odd number more than or equal to 3;
Preferred neighborhood is: the square neighborhood of n=m;
Fairly simple neighborhood is with pixel f
xFor the center, symmetrical cross and symmetrical X-shaped neighborhood;
This moment, median g2 was obtained by following formula:
In the following formula, the median of pixel data in " med10 " expression cross neighborhood; Median in " medx " expression X-shaped neighborhood; The maximum in the two is got in " max (med10, medx) " expression; The minimum value in the two is got in " min (med10, medx) " expression.
Embodiment
1, after initialization noise decision threshold T1, T2 and filter factor a, the b, at first chooses a pixel f
x, its pixel data is f
x
2, be the neighborhood territory pixel point with near the pixel comprise in 3 * 3 the rectangular neighborhood eight then, its pixel data is f
i, i=1,2 ... 8.Referring to Fig. 1.
3, obtain difference DELTA
i:
Δ
i=f
i-f
x (i=1,2,...8);
4, accumulative total Δ
iF greater than T1
iNumber N.
5, carrying out noise judges: when 0<N<T2, and remarked pixel f
xBe signaling point, jump to following the 9th step; Otherwise be noise spot, the step below continuing.The value of T2 is relevant with the noise pollution degree.
6, calculate selected pixel f
xWeighting (weighting herein comprises multiple weighting schemes such as 1/9th types, ten sixth types or Gauss's weighting type, also can the adopt different convolution templates) mean value of pixel data in the neighborhood, this weighted average shows with g1.Such as getting 3 * 3 Gaussian weightings, then convolution template and computing formula are:
g1=1/16×(f
1+2×f
2+f
3+2×f
4+4×f
x+2×f
5+f
6+2×f
7+f
8);
7, calculating pixel f
xThe median of corresponding cross shape neighborhood, X-shaped neighborhood is represented with g2 here;
In the following formula, the median of pixel, i.e. f among Fig. 2 in " med10 " expression cross neighborhood
2, f
4, f
5, f
7Median.Median in " medx " expression X-shaped neighborhood, i.e. f among Fig. 3
1, f
3, f
6, f
8Median.The maximum in the two is got in " max (med10, medx) " expression; The minimum value in the two is got in " min (med10, medx) " expression.
8, calculate the last output valve of filtering according to following formula, i.e. pixel f
xPixel data:
f
x=a×g1+b×g2;
A is respectively linearity and the nonlinear filtering coefficient of selecting at different noise types with b, and its value is relevant with the size of N.
9, repeated for first to the 8th step, all pixel datas in the intact spatial domain picture element matrix of scan process.
The preferred parameter value scope of the present invention is:
Noise decision threshold T1:30~60; T2:190~240;
Filter factor a, b satisfy a/b:0~1, simultaneously the condition of a+b≤1.
The calculating of the foregoing description pixel data is to be that example is carried out with 3 * 3 rectangular neighborhood, and along with the expansion of rectangular neighborhood, amount of calculation will increase sharply.Rectangular neighborhood for 5 * 5, the pixel in the neighborhood are 25, almost are 3 times of 3 * 3 rectangular neighborhood pixel.But the pixel of its corresponding cross type and X-shaped neighborhood has only increased by 1 times, so cross neighborhood and X-shaped neighborhood are the simplest a kind of neighborhoods.
This routine program circuit is referring to Fig. 4.Above step can iteration be carried out, till filter effect is satisfied.The pixel data of Cai Yonging all is the preceding later data of filtering each time.
It can be the Y channel data etc. of gray value, luminous value, brightness value or the yuv space of image that the present invention divides pixel data, and wherein pixel can be the pixel that constitutes rest image, also can be the pixel of the single two field picture of moving image.
Claims (10)
1. spatial domain pixel data processing method may further comprise the steps:
(1) initialization noise decision threshold T1, T2 and filter factor a, b;
(2) judge pixel f according to threshold value T1, T2
xWhether be noise spot;
(3) as judging pixel f
xBe noise spot, enter following steps, otherwise output pixel f
xData;
(4) calculating pixel f
xThe weighted average g1 of the pixel data in the neighborhood;
(5) calculating pixel f
xThe median g2 of the pixel data in the neighborhood;
(6) with a * g1+b * g2 as pixel f
xData output.
2. spatial domain pixel data processing method according to claim 1 is characterized in that, described pixel is the pixel of the single picture frame of rest image or moving image.
3. spatial domain pixel data processing method according to claim 1 is characterized in that, described pixel data is the Y channel data of gray value, luminous value, brightness value or the yuv space of pixel.
4. spatial domain pixel data processing method according to claim 1 is characterized in that, described step (2) is: calculating pixel f
xOther pixels f in data and its neighborhood
iThe difference of data when difference surpasses T2 greater than the number of pixels of T1, is judged pixel f
xBe noise spot.
5. spatial domain pixel data processing method according to claim 1 is characterized in that, in the step (1), the span of described noise decision threshold T1, T2 is: T1:30~60; T2:190~240; Filter factor a, b should satisfy: a/b:0~1, the condition of a+b≤1 simultaneously.
6. spatial domain pixel data processing method according to claim 1 is characterized in that, in the step (4), the weight coefficient of described weighted average is 1/9 type, 1/16 type, 1/25 type, 1/49 type or Gaussian function type.
7. according to any described spatial domain pixel data processing method of claim of claim 1-6, it is characterized in that described neighborhood is with pixel f
xRectangular neighborhood for the n * m at center; Described n, m are the odd number more than or equal to 3.
8. spatial domain pixel data processing method according to claim 7 is characterized in that, described n=m.
9. according to any described spatial domain pixel data processing method of claim of claim 1-6, it is characterized in that described neighborhood is with pixel f
xFor the center, symmetrical cross and symmetrical X-shaped neighborhood.
10. spatial domain pixel data processing method according to claim 9 is characterized in that, in the step (5), described median g2 is obtained by following formula:
Wherein, the median of pixel data in " med10 " expression cross neighborhood; Median in " medx " expression X-shaped neighborhood; The maximum in the two is got in " max (med10, medx) " expression; The minimum value in the two is got in " min (med10, medx) " expression.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006100219870A CN100454970C (en) | 2006-09-30 | 2006-09-30 | Spatial domain pixel data processing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006100219870A CN100454970C (en) | 2006-09-30 | 2006-09-30 | Spatial domain pixel data processing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1929552A true CN1929552A (en) | 2007-03-14 |
CN100454970C CN100454970C (en) | 2009-01-21 |
Family
ID=37859322
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB2006100219870A Expired - Fee Related CN100454970C (en) | 2006-09-30 | 2006-09-30 | Spatial domain pixel data processing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100454970C (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101472058B (en) * | 2007-12-29 | 2011-04-20 | 比亚迪股份有限公司 | Apparatus and method for removing image noise |
CN102447817A (en) * | 2010-09-30 | 2012-05-09 | 瑞昱半导体股份有限公司 | Image processing device and space image noise eliminating method |
CN103714522A (en) * | 2014-01-06 | 2014-04-09 | 海信集团有限公司 | Image denoising method and image denoising system |
CN104994397A (en) * | 2015-07-08 | 2015-10-21 | 北京大学深圳研究生院 | Adaptive video pre-processing method based on time domain information |
CN106022277A (en) * | 2016-05-26 | 2016-10-12 | 湖南明康中锦医疗科技发展有限公司 | Filtering method and device for ventilator |
WO2017088391A1 (en) * | 2015-11-24 | 2017-06-01 | 乐视控股(北京)有限公司 | Method and apparatus for video denoising and detail enhancement |
CN110544261A (en) * | 2019-09-04 | 2019-12-06 | 东北大学 | Blast furnace tuyere coal injection state detection method based on image processing |
CN110738621A (en) * | 2019-10-17 | 2020-01-31 | 内蒙古工业大学 | Linear structure filtering method, device, equipment and storage medium |
CN113570507A (en) * | 2020-04-29 | 2021-10-29 | 浙江宇视科技有限公司 | Image noise reduction method, device, equipment and storage medium |
CN116977228A (en) * | 2023-09-25 | 2023-10-31 | 广东匠芯创科技有限公司 | Image noise reduction method, electronic device and storage medium |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE69329670T2 (en) * | 1992-09-14 | 2001-03-15 | Thomson Multimedia Sa | Method and device for noise reduction |
CN1288916C (en) * | 2004-12-27 | 2006-12-06 | 北京中星微电子有限公司 | Image dead point and noise eliminating method |
CN1328901C (en) * | 2005-01-26 | 2007-07-25 | 北京中星微电子有限公司 | A method for removing image noise |
-
2006
- 2006-09-30 CN CNB2006100219870A patent/CN100454970C/en not_active Expired - Fee Related
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101472058B (en) * | 2007-12-29 | 2011-04-20 | 比亚迪股份有限公司 | Apparatus and method for removing image noise |
CN102447817A (en) * | 2010-09-30 | 2012-05-09 | 瑞昱半导体股份有限公司 | Image processing device and space image noise eliminating method |
CN103714522A (en) * | 2014-01-06 | 2014-04-09 | 海信集团有限公司 | Image denoising method and image denoising system |
CN104994397A (en) * | 2015-07-08 | 2015-10-21 | 北京大学深圳研究生院 | Adaptive video pre-processing method based on time domain information |
CN104994397B (en) * | 2015-07-08 | 2018-11-09 | 北京大学深圳研究生院 | Adaptive video preprocess method based on time-domain information |
WO2017088391A1 (en) * | 2015-11-24 | 2017-06-01 | 乐视控股(北京)有限公司 | Method and apparatus for video denoising and detail enhancement |
CN106022277A (en) * | 2016-05-26 | 2016-10-12 | 湖南明康中锦医疗科技发展有限公司 | Filtering method and device for ventilator |
CN106022277B (en) * | 2016-05-26 | 2019-08-27 | 湖南明康中锦医疗科技发展有限公司 | A kind of filtering method and device for ventilator |
CN110544261A (en) * | 2019-09-04 | 2019-12-06 | 东北大学 | Blast furnace tuyere coal injection state detection method based on image processing |
CN110544261B (en) * | 2019-09-04 | 2023-08-29 | 东北大学 | Method for detecting coal injection state of blast furnace tuyere based on image processing |
CN110738621A (en) * | 2019-10-17 | 2020-01-31 | 内蒙古工业大学 | Linear structure filtering method, device, equipment and storage medium |
CN110738621B (en) * | 2019-10-17 | 2022-05-17 | 内蒙古工业大学 | Linear structure filtering method, device, equipment and storage medium |
CN113570507A (en) * | 2020-04-29 | 2021-10-29 | 浙江宇视科技有限公司 | Image noise reduction method, device, equipment and storage medium |
CN113570507B (en) * | 2020-04-29 | 2024-03-26 | 浙江宇视科技有限公司 | Image noise reduction method, device, equipment and storage medium |
CN116977228A (en) * | 2023-09-25 | 2023-10-31 | 广东匠芯创科技有限公司 | Image noise reduction method, electronic device and storage medium |
CN116977228B (en) * | 2023-09-25 | 2024-02-09 | 广东匠芯创科技有限公司 | Image noise reduction method, electronic device and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN100454970C (en) | 2009-01-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1929552A (en) | Spatial domain pixel data processing method | |
CN1223181C (en) | Noise reduction | |
CN1302436C (en) | Device for contrast compensation and its method | |
CN1218561C (en) | Noise filtering image sequence | |
CN1665298A (en) | Method of removing noise from digital moving picture data | |
CN1766927A (en) | Methods and systems for locally adaptive image processing filters | |
CN103985097B (en) | A kind of image enchancing method of Nonuniform illumination | |
CN1472954A (en) | Circuit and method for improving image quality by fram correlation | |
CN107590779B (en) | Image denoising and deblurring method based on image block clustering dictionary training | |
KR101341099B1 (en) | Method and apparatus for noise reduction | |
CN1250012C (en) | Method and appts. of removing blocking artifact of MPEG signal | |
CN111105371B (en) | Enhancement method of low-contrast infrared image | |
CN1633159A (en) | A method for removing image noise | |
CN111210395B (en) | Retinex underwater image enhancement method based on gray value mapping | |
CN1218560C (en) | Method and system for sharpness enhancement for coded video | |
CN100342710C (en) | Structure method for enhancing image | |
CN2838184Y (en) | Border-based bar chart equalizer | |
CN101431606A (en) | Self-adapting denoising processing method based on edge detection | |
CN1761285A (en) | Method for removing isolated noise point in video | |
CN106981052B (en) | Adaptive uneven brightness variation correction method based on variation frame | |
CN1741617A (en) | Handle the equipment and the method for the shoot artifacts of picture signal | |
CN1679070A (en) | Adaptive noise reduction for digital display panels | |
CN1750044A (en) | Equalizing method for truncating histogram | |
CN1649395A (en) | Image processing apparatus and method | |
CN1266943C (en) | Device and process for estimating noise level, noise reduction system and coding system comprising such a device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090121 Termination date: 20170930 |