CN102281386B - Method and device for performing adaptive denoising on video image - Google Patents
Method and device for performing adaptive denoising on video image Download PDFInfo
- Publication number
- CN102281386B CN102281386B CN201010197007.9A CN201010197007A CN102281386B CN 102281386 B CN102281386 B CN 102281386B CN 201010197007 A CN201010197007 A CN 201010197007A CN 102281386 B CN102281386 B CN 102281386B
- Authority
- CN
- China
- Prior art keywords
- pixel
- value
- filtering
- module
- motion vector
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a method and device for performing adaptive denoising on a video image. The device comprises a video image reading module, an edge detection module, a filtering mode selection module and a filtering module; the method comprises the following steps of: reading in a frame of video image from a buffer zone, respectively calculating motion vectors between each pixel in the frame of image and the pixel at the same location in a neighboring frame, and respectively executing the edge detection on each pixel in the frame of image; respectively selecting the filtering mode having corresponding intensity for each pixel according to the edge detection result and the calculation result of the motion vector and of each pixel, and then, correspondingly filtering each pixel by a Gaussian similarity filter. The method and the device have obvious denoising effect on various noises of different scene sequences, and can effectively remove noises by combination of a time domain and a space domain.
Description
Technical field
The present invention relates to video image processing, video communication technology, relate in particular to a kind of method and device that video image is carried out to self adaptation denoising.
Background technology
The collection of video image, compressed encoding, transmission, decoding, demonstration etc. are the major functions of Video Applications system, in these processes, can inevitably introduce various noises.Containing noisy video image, not only the subjective quality of image can be directly affected, and the multiple subsequent treatment task of video image can be affected, such as: storage, encoding and decoding, transmission, target recognition and tracking etc.
Video sequence not only will be paid close attention to the visual effect of each two field picture, also will pay close attention to the visual experience of whole sequence.Due to the intrinsic reason of electronic product itself, noise is inevitable, and noise can affect visual effect, therefore in order to improve people's visual experience and to be beneficial to follow-up Video processing, using filtering technique to carry out Denoising disposal to video image has become very necessary technological means.
In denoising, keep edge and the detailed information of image, otherwise visual effect after denoising may, not as good as the effect of removing before noise, even can lose some important information as far as possible.
Traditional Denoising disposal method has neighborhood averaging, gradient weighting method reciprocal, linear interpolation method, median filtering method etc.Above-mentioned algorithm simply and is easily realized, but is not very desirable to the denoising effect of image, and it is poor that the image Relative Fuzzy after processing and details keep.Utilize in addition image local average and variance to judge the method for denoising, but because amount of calculation is larger, be not suitable for the real-time processing of image.Based on the morphologic Image denoising method using wavelet method of random geometry, be also another large focus, but its amount of calculation is also larger, and Soft Thresholding also can damage the flat site of image, therefore be not also suitable for the real-time processing of image.
Summary of the invention
The invention provides a kind of method and device that video image is carried out to self adaptation denoising, to solve the defect that denoise algorithm amount of calculation is large and denoising effect is bad existing in prior art.
For addressing the above problem, the invention provides a kind of method of video image being carried out to self adaptation denoising, comprising:
From buffering area, read in a frame video image, calculate respectively in this two field picture in each pixel and consecutive frame and the motion vector between the pixel of this pixel in same position and respectively each pixel in this two field picture is carried out to rim detection; Utilize respectively the result of calculation of motion vector of each pixel and edge detection results to select the filtering mode of respective strengths for each pixel, then utilize Gauss's similarity filter to carry out corresponding filtering to each pixel.
Further, said method specifically comprises the following steps:
A, from buffering area, read in a frame video image;
B, from this two field picture, select the pixel that filtered ripple is processed;
C, calculate in this pixel and consecutive frame and the pixel of this pixel in same position between motion vector, and this pixel is carried out to rim detection;
D, the result of calculation of motion vector of utilizing this pixel and edge detection results are the filtering mode that this pixel is selected respective strengths, and the filtering mode of then selecting according to this utilizes Gauss's similarity filter to carry out filtering to this pixel;
E, judge in this two field picture whether all pixels are all processed after filtering, if not, perform step B.
Further, said method also can comprise:
Each pixel in this two field picture has been carried out, after filtering processing, the view data after denoising being sent into display buffer.
Further, said method also can have following characteristics:
Motion vector between pixel in same position in this pixel of described calculating and consecutive frame refers to: calculate in the pixel value of this pixel and consecutive frame the absolute value with the difference of the pixel value of the pixel of this pixel in same position.
Further, said method also can have following characteristics:
Describedly this pixel is carried out to rim detection refer to: utilize Laplce's Gaussian transformation (LoG) to detect operator the edge details of this pixel is detected, obtain rim detection value.
Further, said method also can have following characteristics:
Described LoG check operator is
Further, said method also can comprise:
Preset 4 threshold values, comprising: the first threshold value, the second threshold value, the 3rd threshold value and the 4th threshold value;
The result of calculation of the described motion vector that utilizes this pixel and edge detection results are that this pixel selects the filtering mode of respective strengths to refer to, this pixel is carried out to following operation:
A, the motion vector value that judges this pixel whether is less than described the first threshold value or whether rim detection value is greater than described the 4th threshold value, if so, selects time-domain filtering mode; Otherwise execution next step;
B, the motion vector value that judges this pixel whether is less than described the second threshold value or whether rim detection value is greater than described the 3rd threshold value, if so, selects weak filtering mode; Otherwise, select strong filtering mode.
Further, said method also can have following characteristics:
The span of described the first threshold value is 2～5, and the span of described the second threshold value is 10～15.
Further, said method also can have following characteristics:
The described Gauss's of utilization similarity filter carries out corresponding filtering to this pixel and specifically comprises:
1) calculate the similarity value Gs between this pixel and neighborhood territory pixel;
2) all pixels identical with current pixel neighborhood of a point position in the frame of adjacent front and back are carried out to similarity computing;
3) according to formula
calculate the pixel value after this pixel denoising;
Wherein, f (i, j, t) represents to be positioned in t frame the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and N is odd number, and works as
time be present frame; i
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel; While adopting different filtering modes, the value of n is different, and n
time-domain filtering< n
weak filtering< n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filter ripplethe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
Further, said method also can have following characteristics:
N
time-domain filtering=1; n
weak filtering=5; n
weak filtering=7
The present invention also provides a kind of device that video image is carried out to self adaptation denoising, comprising: video image read module, video image read module, rim detection module, filtering mode are selected module and filtration module;
Described video image read module is for reading in a frame video image from buffering area;
Described motion vector computation module is used for calculating the motion vector in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position, and sends to described filtering mode to select module result of calculation;
Described rim detection module is used for respectively each pixel of described video image being carried out to rim detection, and sends to described filtering mode to select module testing result;
Described filtering mode selection module is the filtering mode that each pixel is selected respective strengths for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and which is sent to described filtration module;
Described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel.
Further, said apparatus also can have following characteristics:
Described motion vector computation module refers to for the motion vector that calculates in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position: described motion vector computation module is used for calculating the absolute value of the pixel value of each pixel and consecutive frame and the difference of the pixel value of the pixel of this pixel in same position.
Further, said apparatus also can have following characteristics:
Described rim detection module refers to for respectively each pixel of described video image being carried out to rim detection: described rim detection module is used for utilizing Laplce's Gaussian transformation (LoG) detection operator respectively the edge details of each pixel to be detected, and obtains rim detection value.
Further, said apparatus also can have following characteristics:
Described filtering mode is selected to have preset 4 threshold values in module, comprising: the first threshold value, the second threshold value, the 3rd threshold value and the 4th threshold value;
Described filtering mode selection module is that each pixel selects the filtering mode of respective strengths to refer to for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and described filtering mode selects module for respectively each pixel being carried out to following operation:
A, the motion vector value that judges this pixel whether is less than described the first threshold value or whether rim detection value is greater than described the 4th threshold value, if so, selects time-domain filtering mode; Otherwise execution next step;
B, the motion vector value that judges this pixel whether is less than described the second threshold value or whether rim detection value is greater than described the 3rd threshold value, if so, selects weak filtering mode; Otherwise, select strong filtering mode.
Further, said apparatus also can have following characteristics:
Described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel referring to:
Described filtration module is for calculating the similarity value Gs between this pixel and neighborhood territory pixel; Also for the frame all pixels identical with current pixel neighborhood of a point position in adjacent front and back are carried out to similarity computing; Also for according to formula
calculate the pixel value after this pixel denoising;
Wherein, f (i, j, t) represents to be positioned in t frame the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and N is odd number, and works as
time be present frame; i
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel; While adopting different filtering modes, the value of n is different, and n
time-domain filtering< n
weak filtering< n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filter ripplethe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
Beneficial effect of the present invention is as follows:
1) each noise like of different sequence of scenes is all had to obvious denoising effect, utilize the method that time domain and spatial domain combine can effectively remove noise;
2) utilize LoG detector to carry out rim detection, obtain strong and weak edge and flat site, according to consecutive frame motion vector, divide large motion, little motion, rest point, based on above-mentioned situation, determine filtering strength, and then adaptive-filtering, guarantee the edge details of image, avoided the generation of motion blur phenomenon;
3) by the power of judgement noise, take different filtering strengths, carry out the filtering of Gauss's similarity, regulate adaptively filtering weighting, on the basis that guarantees definition, effectively removed noise;
4) algorithm complex is low, is easy to software and hardware and realizes, and real-time is higher, has met well the requirement of the real-time systems such as mobile communication, video conference.
Accompanying drawing explanation
Fig. 1 carries out the handling process of self adaptation denoising method to video image in the embodiment of the present invention;
Fig. 2 is sequence consecutive frame neighborhood of pixel points diagram in the embodiment of the present invention;
Fig. 3 carries out the structure drawing of device of self adaptation denoising to video image in the embodiment of the present invention.
Embodiment
Basic conception of the present invention is: from buffering area, read in a frame video image, calculate respectively in this two field picture in each pixel and consecutive frame and the motion vector between the pixel of this pixel in same position and respectively each pixel in this two field picture is carried out to rim detection; Utilize respectively the result of calculation of motion vector of each pixel and edge detection results to select the filtering mode of respective strengths for each pixel, then utilize Gauss's similarity filter to carry out corresponding filtering to each pixel.
As shown in Figure 1, the present invention mainly comprises the following steps:
S1, from buffering area, read in a frame video image;
S2, from this two field picture, select the pixel that filtered ripple is processed,
S3, calculate in this pixel and consecutive frame and the pixel of this pixel in same position between motion vector, and this pixel is carried out to rim detection, can be, but not limited to utilize LoG (Laplacianof a Gaussian, Laplce's Gaussian transformation) to detect operator detects the edge details of this pixel;
S4, utilize above-mentioned motion vector computation result and edge detection results to select the filtering mode of respective strengths for this pixel, then utilize Gauss's similarity filter to carry out corresponding filtering to this pixel;
S5, judge in this two field picture that whether all pixels are all processed after filtering, if so, carry out next step; Otherwise, execution step S2;
S6, the view data after denoising is sent into display buffer.
After handling this two field picture, can repeated using said method be processed by the image of subsequent frame, until handle all video images.
In described step S3, according to pixel identical with this pixel position in consecutive frame, calculate the motion vector of this pixel, current pixel point can be divided into rest point, little motor point, large motor point according to motion state, specifically comprise the following steps:
1) utilize in the pixel value of current pixel point and former frame and carry out subtraction with the pixel value of the pixel of current pixel point in same position, and get its absolute value ABS; ?
ABS＝|f(i，j)-f′(i，j)|
Wherein, f (i, j) is illustrated in present frame and is positioned at (i, j) pixel value of the pixel at coordinate position place, this pixel is current pixel point, f ' (i, j) be illustrated in the pixel value that is positioned at the pixel at (i, j) coordinate position place in the former frame of present frame.
2) the absolute value ABS of the difference obtaining and specified thresholds TH1 and TH2 are compared, and according to comparative result, set the motion state PointFlag of current point, if ABS is less than TH1, set the PointFlag value representation rest point of current point, if ABS is between TH1 and TH2, set the little motor point of PointFlag value representation of current point, otherwise set the large motor point of PointFlag value representation of current point.Through dissimilar noise video experiment, obtain preferably, the span that the span of TH1 is 2～5, TH2 is 10～15.
In described step S3, utilize LoG to detect the edge details that operator detects current pixel, the state of current pixel is divided to strong edge, weak edge, flat site, specifically comprise the steps:
1) adopt LoG to detect operator current pixel point is carried out to rim detection, first current pixel point is carried out to smothing filtering, then by the computing of filtering result process Laplacian, obtain the rim detection value Edge of this pixel; Wherein, LoG operator can be:
2) the rim condition EdgeFlag that this rim detection value Edge and default threshold value TH4 and TH3 is compared to set current pixel point, if Edge value is greater than threshold value TH4, the rim condition depending on current pixel point is strong edge; As Edge value at TH3 between TH4, the rim condition depending on current pixel point is weak edge; Otherwise, depending on the rim condition of current pixel point, be flat site.
In described step S4, utilizing above-mentioned motion vector computation result and edge detection results to select the filtering mode of respective strengths for this pixel specifically comprises:
1) whether the motion vector value that judges this pixel is rest point or is strong edge, if so, selects time-domain filtering mode; Otherwise execution next step;
2) judge the whether little motor point of this pixel or weak edge, if so, select weak filtering mode; Otherwise, select strong filtering mode;
In described step S4, utilizing Gauss's similarity filter to carry out corresponding filtering to current pixel point specifically comprises:
1) pixel value of the pixel value of current pixel point and neighborhood territory pixel point around thereof is carried out to difference operation, take absolute value, gained absolute value, by gaussian filtering template, is obtained to the similarity value Gs of current pixel point;
Wherein, f (i, j) represents the pixel value of current pixel point, and f (m, n) represents the pixel value of the neighborhood territory pixel point of current pixel point, and Gauss H is Gaussian parameter.Neighborhood is often referred to the square area of putting a formed k * k centered by current pixel point, and generally, the value of k is odd number.Neighborhood territory pixel refers to other pixels except current pixel point in this region.
2) as shown in Figure 2, all pixels identical with neighborhood position, current pixel point place in the frame of adjacent front and back are carried out to similarity computing, according to formula
calculate the pixel value after current pixel point denoising;
Wherein: f (i, j, t) represents that t framing bit is in the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and the number N of the frame of getting is odd number; And work as
time, represent present frame; As shown in Figure 2, when N=3, represent to calculate a rear two field picture of the former frame of present frame, present frame and present frame.I
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel.While adopting different filtering modes, the value of n is different, and n
time-domain filtering< n
weak filtering< n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filteringthe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
While adopting time-domain filtering, the value of n is desirable 1, and while adopting weak filtering, the value of n is desirable 5, while adopting strong filtering, and the value of n desirable 7.The level and smooth strength factor Gauss of the gaussian filtering H wherein strong filtering being adopted can be made as 100, and other situations can adopt 50 level and smooth strength factor.
In addition, as shown in Figure 3, the device that video image is carried out to self adaptation denoising of the present invention, comprising: video image read module, motion vector computation module, rim detection module, filtering mode are selected module and filtration module;
Described video image read module is for reading in a frame video image from buffering area;
Described motion vector computation module is used for calculating the motion vector in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position, and sends to described filtering mode to select module result of calculation;
Described rim detection module is used for respectively each pixel of described video image being carried out to rim detection, and sends to described filtering mode to select module testing result;
Described filtering mode selection module is the filtering mode that each pixel is selected respective strengths for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and which is sent to described filtration module;
Described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel.
Preferably, described motion vector computation module refers to for the motion vector that calculates in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position: described motion vector computation module is used for calculating the absolute value of the pixel value of each pixel and consecutive frame and the difference of the pixel value of the pixel of this pixel in same position.Described rim detection module refers to for respectively each pixel of described video image being carried out to rim detection: described rim detection module is used for utilizing Laplce's Gaussian transformation (LoG) detection operator respectively the edge details of each pixel to be detected, and obtains rim detection value.
Wherein, described filtering mode is selected to have preset 4 threshold values in module, comprising: the first threshold value, the second threshold value, the 3rd threshold value and the 4th threshold value;
Described filtering mode selection module is that each pixel selects the filtering mode of respective strengths to refer to for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and described filtering mode selects module for respectively each pixel being carried out to following operation:
A, the motion vector value that judges this pixel whether is less than described the first threshold value or whether rim detection value is greater than described the 4th threshold value, if so, selects time-domain filtering mode; Otherwise execution next step;
B, the motion vector value that judges this pixel whether is less than described the second threshold value or whether rim detection value is greater than described the 3rd threshold value, if so, selects weak filtering mode; Otherwise, select strong filtering mode.
Further, described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel referring to:
Described filtration module is for calculating the similarity value Gs between this pixel and neighborhood territory pixel; Also for the frame all pixels identical with current pixel neighborhood of a point position in adjacent front and back are carried out to similarity computing; Also for according to formula
calculate the pixel value after this pixel denoising;
Wherein, f (i, j, t) represents to be positioned in t frame the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and N is odd number, and works as
time be present frame; i
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel; While adopting different filtering modes, the value of n is different, and n
time-domain filtering< n
weak filtering< n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filter ripplethe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
Claims (11)
1. video image is carried out to a method for self adaptation denoising, comprising:
From buffering area, read in a frame video image, calculate respectively in this two field picture in each pixel and consecutive frame and the motion vector between the pixel of this pixel in same position and respectively each pixel in this two field picture is carried out to rim detection; Utilize respectively the result of calculation of motion vector of each pixel and edge detection results to select the filtering mode of respective strengths for each pixel, then utilize Gauss's similarity filter to carry out corresponding filtering to each pixel;
Wherein, in this pixel of described calculating and consecutive frame, the motion vector between the pixel in same position refers to: calculate in the pixel value of this pixel and consecutive frame the absolute value with the difference of the pixel value of the pixel of this pixel in same position;
Preset 4 threshold values, comprising: the first threshold value, the second threshold value, the 3rd threshold value and the 4th threshold value;
The result of calculation of the described motion vector that utilizes this pixel and edge detection results are that this pixel selects the filtering mode of respective strengths to refer to, this pixel is carried out to following operation:
A, the motion vector value that judges this pixel whether is less than described the first threshold value or whether rim detection value is greater than described the 4th threshold value, if so, selects time-domain filtering mode; Otherwise execution next step;
B, the motion vector value that judges this pixel whether is less than described the second threshold value or whether rim detection value is greater than described the 3rd threshold value, if so, selects weak filtering mode; Otherwise, select strong filtering mode.
2. the method for claim 1, is characterized in that, described method specifically comprises the following steps:
A, from buffering area, read in a frame video image;
B, from this two field picture, select the pixel that filtered ripple is processed;
C, calculate in this pixel and consecutive frame and the pixel of this pixel in same position between motion vector, and this pixel is carried out to rim detection;
D, the result of calculation of motion vector of utilizing this pixel and edge detection results are the filtering mode that this pixel is selected respective strengths, and the filtering mode of then selecting according to this utilizes Gauss's similarity filter to carry out filtering to this pixel;
E, judge in this two field picture whether all pixels are all processed after filtering, if not, perform step B.
3. method as claimed in claim 1 or 2, is characterized in that, described method also comprises:
Each pixel in this two field picture has been carried out, after filtering processing, the view data after denoising being sent into display buffer.
4. method as claimed in claim 2, is characterized in that,
Describedly this pixel is carried out to rim detection refer to: utilize Laplce's Gaussian transformation LoG to detect operator the edge details of this pixel is detected, obtain rim detection value.
5. method as claimed in claim 4, is characterized in that,
Described LoG detects operator
6. the method for claim 1, is characterized in that,
The span of described the first threshold value is 2～5, and the span of described the second threshold value is 10～15.
7. the method for claim 1, is characterized in that,
The described Gauss's of utilization similarity filter carries out corresponding filtering to this pixel and specifically comprises:
1) calculate the similarity value Gs between this pixel and neighborhood territory pixel;
2) all pixels identical with current pixel neighborhood of a point position in the frame of adjacent front and back are carried out to similarity computing;
3) according to formula
calculate the pixel value after this pixel denoising;
Wherein, f (i, j, t) represents to be positioned in t frame the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and N is odd number, and works as
time be present frame; i
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel; While adopting different filtering modes, the value of n is different, and n
time-domain filtering<n
weak filtering<n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filter ripplethe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
8. method as claimed in claim 7, is characterized in that,
N
time-domain filtering=1; n
weak filtering=5; n
strong filtering=7.
9. video image is carried out to a device for self adaptation denoising, comprising: video image read module, motion vector computation module, rim detection module, filtering mode are selected module and filtration module;
Described video image read module is for reading in a frame video image from buffering area;
Described motion vector computation module is used for calculating the motion vector in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position, and sends to described filtering mode to select module result of calculation;
Described rim detection module is used for respectively each pixel of described video image being carried out to rim detection, and sends to described filtering mode to select module testing result;
Described filtering mode selection module is the filtering mode that each pixel is selected respective strengths for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and which is sent to described filtration module;
Described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel;
Wherein, described motion vector computation module refers to for the motion vector that calculates in each pixel of described video image and consecutive frame and between the pixel of this pixel in same position: described motion vector computation module is used for calculating the absolute value of the pixel value of each pixel and consecutive frame and the difference of the pixel value of the pixel of this pixel in same position;
Described filtering mode is selected to have preset 4 threshold values in module, comprising: the first threshold value, the second threshold value, the 3rd threshold value and the 4th threshold value;
Described filtering mode selection module is that each pixel selects the filtering mode of respective strengths to refer to for utilizing result of calculation and the edge detection results of the motion vector of each pixel receiving, and described filtering mode selects module for respectively each pixel being carried out to following operation:
A, the motion vector value that judges this pixel whether is less than described the first threshold value or whether rim detection value is greater than described the 4th threshold value, if so, selects time-domain filtering mode; Otherwise execution next step;
B, the motion vector value that judges this pixel whether is less than described the second threshold value or whether rim detection value is greater than described the 3rd threshold value, if so, selects weak filtering mode; Otherwise, select strong filtering mode.
10. device as claimed in claim 9, is characterized in that,
Described rim detection module refers to for respectively each pixel of described video image being carried out to rim detection: described rim detection module is used for utilizing Laplce's Gaussian transformation (LoG) detection operator respectively the edge details of each pixel to be detected, and obtains rim detection value.
11. devices as claimed in claim 9, is characterized in that,
Described filtration module is used for utilizing Gauss's similarity filter to carry out corresponding filtering to each pixel referring to:
Described filtration module is for calculating the similarity value Gs between this pixel and neighborhood territory pixel; Also for the frame all pixels identical with current pixel neighborhood of a point position in adjacent front and back are carried out to similarity computing; Also for according to formula
calculate the pixel value after this pixel denoising;
Wherein, f (i, j, t) represents to be positioned in t frame the pixel value at (i, j) coordinate position place, and Gs represents corresponding similarity value, and N is odd number, and works as
time be present frame; i
max-i
min=j
max-j
min=n, and coordinate (i
max, j
min), (i
max, j
max), (i
min, j
min) and (i
min, j
max) be four apex coordinates of the neighborhood of current pixel; While adopting different filtering modes, the value of n is different, and n
time-domain filtering<n
weak filtering<n
strong filtering; Wherein, n
time-domain filteringthe value of n while representing to adopt time-domain filtering mode, n
weak filter ripplethe value of n while representing to adopt weak filtering mode, n
strong filteringthe value of n while representing to adopt strong filtering mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010197007.9A CN102281386B (en) | 2010-06-08 | 2010-06-08 | Method and device for performing adaptive denoising on video image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010197007.9A CN102281386B (en) | 2010-06-08 | 2010-06-08 | Method and device for performing adaptive denoising on video image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102281386A CN102281386A (en) | 2011-12-14 |
| CN102281386B true CN102281386B (en) | 2014-12-10 |
Family
ID=45106540
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010197007.9A Expired - Fee Related CN102281386B (en) | 2010-06-08 | 2010-06-08 | Method and device for performing adaptive denoising on video image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102281386B (en) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5399578B2 (en) * | 2012-05-16 | 2014-01-29 | シャープ株式会社 | Image processing apparatus, moving image processing apparatus, video processing apparatus, image processing method, video processing method, television receiver, program, and recording medium |
| CN103414845A (en) * | 2013-07-24 | 2013-11-27 | 中国航天科工集团第三研究院第八三五七研究所 | Self-adaptive video image noise reducing method and noise reducing system |
| CN104349080B (en) * | 2013-08-07 | 2018-12-14 | 联想(北京)有限公司 | A kind of image processing method and electronic equipment |
| CN103632352B (en) * | 2013-11-01 | 2017-04-26 | 华为技术有限公司 | Method for time domain noise reduction of noise image and related device |
| US10242432B2 (en) | 2015-04-16 | 2019-03-26 | Institute Of Automation Chinese Academy Of Sciences | Video denoising system based on noise correlation |
| US10025988B2 (en) * | 2015-05-22 | 2018-07-17 | Tektronix, Inc. | Anomalous pixel detection |
| CN105915761A (en) * | 2015-07-24 | 2016-08-31 | 乐视云计算有限公司 | Self-adaptive video space domain denoising method and device |
| US10559106B2 (en) | 2015-11-16 | 2020-02-11 | Huawei Technologies Co., Ltd. | Video smoothing method and apparatus |
| CN111915535B (en) * | 2015-12-30 | 2022-09-27 | 展讯通信(天津)有限公司 | Image denoising method and device |
| CN105872317B (en) * | 2016-03-31 | 2019-04-05 | 北京奇艺世纪科技有限公司 | A kind of video image filtering method and device based on time domain |
| TWI598584B (en) * | 2016-12-23 | 2017-09-11 | 財團法人工業技術研究院 | Gas leakage detection method |
| CN106998437B (en) * | 2017-03-31 | 2020-07-31 | 武汉斗鱼网络科技有限公司 | Method and device for reconstructing video image |
| CN107169982A (en) * | 2017-05-17 | 2017-09-15 | 重庆邮电大学 | A kind of quantum LoG edge detection methods |
| CN110858867B (en) * | 2018-08-07 | 2021-12-10 | 瑞昱半导体股份有限公司 | Image processing device and method thereof |
| CN111539895B (en) * | 2020-04-30 | 2023-04-18 | 广州市百果园信息技术有限公司 | Video denoising method and device, mobile terminal and storage medium |
| WO2021217643A1 (en) * | 2020-04-30 | 2021-11-04 | 深圳市大疆创新科技有限公司 | Method and device for infrared image processing, and movable platform |
| CN112826483B (en) * | 2021-01-08 | 2022-03-08 | 中国科学院自动化研究所 | Fingertip video-based heart rate detection method, system and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101031029A (en) * | 2007-04-04 | 2007-09-05 | 华为技术有限公司 | Method and apparatus for processing video image |
| CN101355648A (en) * | 2008-06-26 | 2009-01-28 | 天津市亚安科技电子有限公司 | Method for reducing image noise and enhancing image |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004318423A (en) * | 2003-04-15 | 2004-11-11 | Konica Minolta Photo Imaging Inc | Image processing method, image processor, and image processing program |
| CN100369478C (en) * | 2004-02-18 | 2008-02-13 | 华亚微电子（上海）有限公司 | Image edge smoothing system and method based on directional interpolation |
| TWI343743B (en) * | 2006-01-20 | 2011-06-11 | Realtek Semiconductor Corp | Image processing circuit and method |
-
2010
- 2010-06-08 CN CN201010197007.9A patent/CN102281386B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101031029A (en) * | 2007-04-04 | 2007-09-05 | 华为技术有限公司 | Method and apparatus for processing video image |
| CN101355648A (en) * | 2008-06-26 | 2009-01-28 | 天津市亚安科技电子有限公司 | Method for reducing image noise and enhancing image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102281386A (en) | 2011-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102281386B (en) | Method and device for performing adaptive denoising on video image | |
| CN103369209B (en) | Vedio noise reduction device and method | |
| US8768069B2 (en) | Image enhancement apparatus and method | |
| CN109584204B (en) | Image noise intensity estimation method, storage medium, processing and recognition device | |
| US7554611B2 (en) | Method and apparatus of bidirectional temporal noise reduction | |
| US20010019633A1 (en) | Noise reduction | |
| Yüksel | Edge detection in noisy images by neuro-fuzzy processing | |
| CN101996406A (en) | No-reference structural sharpness image quality evaluation method | |
| EP2836963B1 (en) | Noise reduction for image sequences | |
| CN104103041B (en) | Ultrasonoscopy mixed noise Adaptive Suppression method | |
| Zhang et al. | Decision-based non-local means filter for removing impulse noise from digital images | |
| CN101504769B (en) | Self-adaptive noise intensity estimation method based on encoder frame work | |
| US8908101B2 (en) | Method and apparatus for reducing noise of video | |
| CN103971345A (en) | Image denoising method based on improved bilateral filtering | |
| CN110351453A (en) | A kind of computer video data processing method | |
| Yang et al. | A design framework for hybrid approaches of image noise estimation and its application to noise reduction | |
| Ghazal et al. | Structure-oriented multidirectional wiener filter for denoising of image and video signals | |
| Li et al. | Research on motion blurred image restoration | |
| CN111417015A (en) | Method for synthesizing computer video | |
| Liu | New method for image denoising while keeping edge information | |
| Tan et al. | Multihypothesis recursive video denoising based on separation of motion state | |
| Patil et al. | Implementation of restoration of blurred image using blind deconvolution algorithm | |
| Gupta et al. | Blind restoration method for satellite images using memetic algorithm | |
| Laligant et al. | Edge enhancement by local deconvolution | |
| He et al. | Images denoising by improved non-local means algorithm |
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 |
Granted publication date: 20141210 Termination date: 20190608 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |