CN104202554A - Intra-field anti-aliasing and deinterlacing method - Google Patents
Intra-field anti-aliasing and deinterlacing method Download PDFInfo
- Publication number
- CN104202554A CN104202554A CN201410466905.8A CN201410466905A CN104202554A CN 104202554 A CN104202554 A CN 104202554A CN 201410466905 A CN201410466905 A CN 201410466905A CN 104202554 A CN104202554 A CN 104202554A
- Authority
- CN
- China
- Prior art keywords
- under
- dmax
- choose
- hunting zone
- model
- 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.)
- Pending
Links
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The invention discloses an intra-field anti-aliasing and deinterlacing method. The method is characterized in that an interpolation point P (x, y) is selected, with x and y being a row number and a column number of a to-be-interpolated point in an image; first, two rows of related pixels above and below the to-be-interpolated point are selected, and smoothing and filtering are performed to form vectors UPPER and UNDER; second, the vectors UPPER and UNDER are used as input of direction models, and direction relevancy values DR(i), i<[-5, 5], of eleven direction models are calculated; a maximum Dmax is found from the direction relevancy values DR(i), and screening is performed; third, a direction estimated value D'(x, y) of the to-be-interpolated point is calculated with direction values of two points on the left and right of the to-be-interpolated point; fourth, a search range is selected according to the direction estimated value D'(x, y) of the to-be-interpolated point; fifth, within the search range, a direction relevancy DR(k), k<(search range), of each direction model is calculated; a maximum is found from the direction relevancies DR (k), and the direction model corresponding to the maximum is an interpolation direction DI of the to-be-interpolated point.
Description
Technical field
The present invention relates to a kind of Computer Image Processing method, particularly an interior de-interlaced method of reverse sawtooth.
Background technology
In prior art, and the immediate scheme of the technology of the present invention: edge self-adaption interpolation method.1, to be inserted some position selected several detection sides to; 2, detection side, make progress, two corresponding pixel values are done difference computing, and result takes absolute value; 3, compare order of magnitude, find out that direction of absolute value minimum, as interpolation direction; 4, in interpolation direction, choose related pixel, add certain weight, calculate interpolation result.Adopt above-mentioned mode to have that row copies, the low side method such as average interpolation in the ranks, image border sawtooth effect is serious, and image vertical definition is not good; Edge self-adaption interpolation method, easily causes edge direction erroneous judgement, and then introduces noise; Interframe movement detects, and operand is large, and resource consumption is large, once and detect mistake, easily introduce the shortcoming of noise.
Summary of the invention
For above-mentioned the deficiencies in the prior art part, the invention provides a de-interlaced method of interior reverse sawtooth, effectively solved the problem that above-mentioned prior art exists.
To achieve these goals, the technical solution used in the present invention is: the de-interlaced method of reverse sawtooth in, it is P (x, y) that this method is established interpolation point, wherein x, y be respectively to be inserted in image line number, the columns at place;
The first step: choose to be inserted some lastrow related pixel: P (x-1, j-4), P (x-1, j-3), P (x-1, j-2), P (x-1, j-1), P (x-1, j), P (x-1, j+1), P (x-1, j+2), P (x-1, j+3), P (x-1, j+4), it is carried out to smothing filtering, obtain vectorial UPPER; Choose to be inserted some next line related pixel: P (x+1, j-4), P (x+1, j-3), P (x+1, j-2), P (x+1, j-1), P (x+1, j), P (x+1, j+1), P (x+1, j+2), P (x+1, j+3), P (x+1, j+4), it is carried out to smothing filtering, obtain vectorial UNDER.
Second step: by vectorial UPPER, UNDER, as the input of direction model, calculates the directional dependency value DR (i) of 11 direction models, i ∈ [5,5].Find out maximum Dmax in DR (i), then do following screening:
If DR (0)=Dmax, D (x, y)=0;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=1;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=-1;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=2;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=-2;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=3;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=-3;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=4;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=-4;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=5;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=-5;
Otherwise D=0;
Each point to be inserted, all calculates direction value D (x, y), wherein x, y be respectively to be inserted in image line number, the columns at place.
The 3rd step: with the direction value of to be inserted each two points of left and right, calculate the direction estimated value D ' (x, y) of point to be inserted, concrete formula is:
D’(x,y)=D(x,y-2)+D(x,y-1)+D(x,y+1)+D(x,y+2);
The 4th step: according to be inserted some direction estimated value D ' (x, y), choose hunting zone, concrete grammar is:
If D ' (x, y)≤-16, choose-5 ,-4 ,-3 ,-2 ,-1,0 direction model, as hunting zone;
D ' (x, y)≤-12, choose-4 ,-3 ,-2 else if, and-1,0 direction model, as hunting zone;
D ' (x, y) >=12, choose 4,3,2,1 else if, and 0 direction model, as hunting zone;
D ' (x, y) >=16, choose 5,4,3,2 else if, and 1,0 direction model, as hunting zone;
Otherwise, choose-3 ,-2 ,-1,0,1,2,3 direction models, as hunting zone.
As preferably, described calculating result to be inserted, concrete grammar is:
If, DI=0, P (x, y)=(UPPER (5)+UNDER (5))/2;
If, DI=-1 or-2, P (x, y)=(UPPER (4)+UPPER (5)+UNDER (5)+UNDER (6))/4;
If, DI=1 or 2, P (x, y)=(UPPER (5)+UPPER (6)+UNDER (4)+UNDER (5))/4;
If, DI=3, P (x, y)=(UPPER (6)+UNDER (4))/2;
If, DI=-3, P (x, y)=(UPPER (4)+UNDER (6))/2;
If, DI=4, P (x, y)=(UPPER (7)+UNDER (3))/2;
If, DI=-4, P (x, y)=(UPPER (3)+UNDER (7))/2;
If, DI=5, P (x, y)=(UPPER (8)+UNDER (2))/2;
If, DI=-5, P (x, y)=(UPPER (2)+UNDER (8))/2.
Compared with prior art, this beneficial effect of the invention: the present invention by the improvement in method make in field accurately, adaptive direction detects, and avoids interframe movement to detect a large amount of logical resources of required consumption.Effectively suppress the noise that edge sawtooth effect and direction erroneous judgement are introduced, obtain fine and smooth, soft image border.
Accompanying drawing explanation
Fig. 1 is the design diagram of direction model.
Embodiment
Below in conjunction with drawings and the specific embodiments, the present invention is described in further detail.
Referring to Fig. 1, the de-interlaced method of reverse sawtooth in, it is P (x, y) that this method is established interpolation point, wherein x, y be respectively to be inserted in image line number, the columns at place;
The first step: choose to be inserted some lastrow related pixel: P (x-1, j-4), P (x-1, j-3), P (x-1, j-2), P (x-1, j-1), P (x-1, j), P (x-1, j+1), P (x-1, j+2), P (x-1, j+3), P (x-1, j+4), it is carried out to smothing filtering, obtain vectorial UPPER; Choose to be inserted some next line related pixel: P (x+1, j-4), P (x+1, j-3), P (x+1, j-2), P (x+1, j-1), P (x+1, j), P (x+1, j+1), P (x+1, j+2), P (x+1, j+3), P (x+1, j+4), it is carried out to smothing filtering, obtain vectorial UNDER.Second step: by vectorial UPPER, UNDER (refers to and formulates 11 directions and give a numbering (5~5) as direction model, when a direction is detected, set a pixel extraction template, three pairs of specific pixel that extraction the party makes progress are carried out pixel interdependence calculating, wherein A0 and B0 are paired, A1 and B1 are paired, A2 and B2 are paired) input, the directional dependency that calculates 11 direction models (refers in direction model, three pairs of pixel interdependences and value, with DR, represent, be specially: DR=R (A0, B0)+R (A1, B1)+R (A2, B2) 11 directional dependency, be expressed as DR (i), i ∈ [5, 5]) value DR (i), i ∈ [5, 5].Find out maximum Dmax in DR (i), then do following screening:
If DR (0)=Dmax, D (x, y)=0;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=1;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=-1;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=2;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=-2;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=3;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=-3;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=4;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=-4;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=5;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=-5;
Otherwise D=0;
Each point to be inserted, all calculate direction value and (refer to the directional dependency that goes out 11 direction models to be inserted some position calculation, find out maximum wherein, the direction numbering that this direction model is corresponding, is the direction value of this point to be inserted, with D (x, y) represent, x, y are respectively to be inserted some place line number, columns) D (x, y), wherein x, y be respectively to be inserted in image line number, the columns at place.
The 3rd step: with the direction value of to be inserted each two points of left and right, calculate direction estimated value the D ' (x of point to be inserted, y) (refer to the direction value sum to be inserted left and right each two points in position, with D ' (x, y) represent, be specially: D ' (x, y)=D (x, y-2)+D (x, y-1)+D (x, y+1)+D (x, y+2) wherein x, y is respectively to be inserted some place line number, columns), concrete formula is: D ' (x, y)=D (x, y-2)+D (x, y-1)+D (x, y+1)+D (x, y+2);
The 4th step: according to be inserted some direction estimated value D ' (x, y), choose hunting zone, concrete grammar is:
If D ' (x, y)≤-16, choose-5 ,-4 ,-3 ,-2 ,-1,0 direction model, as hunting zone;
D ' (x, y)≤-12, choose-4 ,-3 ,-2 else if, and-1,0 direction model, as hunting zone;
D ' (x, y) >=12, choose 4,3,2,1 else if, and 0 direction model, as hunting zone;
D ' (x, y) >=16, choose 5,4,3,2 else if, and 1,0 direction model, as hunting zone;
Otherwise, choose-3 ,-2 ,-1,0,1,2,3 direction models, as hunting zone.
In the present embodiment, described calculating result to be inserted, concrete grammar is:
If, DI=0, P (x, y)=(UPPER (5)+UNDER (5))/2;
If, DI=-1 or-2, P (x, y)=(UPPER (4)+UPPER (5)+UNDER (5)+UNDER (6))/4;
If, DI=1 or 2, P (x, y)=(UPPER (5)+UPPER (6)+UNDER (4)+UNDER (5))/4;
If, DI=3, P (x, y)=(UPPER (6)+UNDER (4))/2;
If, DI=-3, P (x, y)=(UPPER (4)+UNDER (6))/2;
If, DI=4, P (x, y)=(UPPER (7)+UNDER (3))/2;
If, DI=-4, P (x, y)=(UPPER (3)+UNDER (7))/2;
If, DI=5, P (x, y)=(UPPER (8)+UNDER (2))/2;
If, DI=-5, P (x, y)=(UPPER (2)+UNDER (8))/2.
Claims (2)
1. a de-interlaced method of interior reverse sawtooth, is characterized in that: it is P (x, y) that this method is established interpolation point, wherein x, y be respectively to be inserted in image line number, the columns at place;
The first step: choose to be inserted some lastrow related pixel: P (x-1, j-4), P (x-1, j-3), P (x-1, j-2), P (x-1, j-1), P (x-1, j), P (x-1, j+1), P (x-1, j+2), P (x-1, j+3), P (x-1, j+4), it is carried out to smothing filtering, obtain vectorial UPPER; Choose to be inserted some next line related pixel: P (x+1, j-4), P (x+1, j-3), P (x+1, j-2), P (x+1, j-1), P (x+1, j), P (x+1, j+1), P (x+1, j+2), P (x+1, j+3), P (x+1, j+4), it is carried out to smothing filtering, obtain vectorial UNDER;
Second step: by vectorial UPPER, UNDER, as the input of direction model, calculates the directional dependency value DR (i) of 11 direction models, i ∈ [5,5].Find out maximum Dmax in DR (i), then do following screening:
If DR (0)=Dmax, D (x, y)=0;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=1;
Else if, DR (1)=Dmax and DR (1) >DR (1), D (x, y)=-1;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=2;
Else if, DR (2)=Dmax and DR (2) >DR (2), D (x, y)=-2;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=3;
Else if, DR (3)=Dmax and DR (3) >DR (3), D (x, y)=-3;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=4;
Else if, DR (4)=Dmax and DR (4) >DR (4), D (x, y)=-4;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=5;
Else if, DR (5)=Dmax and DR (5) >DR (5), D (x, y)=-5;
Otherwise D=0;
Each point to be inserted, all calculates direction value D (x, y), wherein x, y be respectively to be inserted in image line number, the columns at place;
The 3rd step: with the direction value of to be inserted each two points of left and right, calculate the direction estimated value D ' (x, y) of point to be inserted, concrete formula is:
D’(x,y)=D(x,y-2)+D(x,y-1)+D(x,y+1)+D(x,y+2);
The 4th step: according to be inserted some direction estimated value D ' (x, y), choose hunting zone, concrete grammar is:
If D ' (x, y)≤-16, choose-5 ,-4 ,-3 ,-2 ,-1,0 direction model, as hunting zone;
D ' (x, y)≤-12, choose-4 ,-3 ,-2 else if, and-1,0 direction model, as hunting zone;
D ' (x, y) >=12, choose 4,3,2,1 else if, and 0 direction model, as hunting zone;
D ' (x, y) >=16, choose 5,4,3,2 else if, and 1,0 direction model, as hunting zone;
Otherwise, choose-3 ,-2 ,-1,0,1,2,3 direction models, as hunting zone;
The 5th step: in hunting zone, calculate the directional dependency DR (k) of all directions model, k ∈ " hunting zone ".Find out maximum in DR (k), the direction model that it is corresponding, is the interpolation direction DI of point to be inserted.
2. the de-interlaced method of reverse sawtooth according to claim 1, is characterized in that: described calculating result to be inserted, and concrete grammar is:
If, DI=0, P (x, y)=(UPPER (5)+UNDER (5))/2;
If, DI=-1 or-2, P (x, y)=(UPPER (4)+UPPER (5)+UNDER (5)+UNDER (6))/4;
If, DI=1 or 2, P (x, y)=(UPPER (5)+UPPER (6)+UNDER (4)+UNDER (5))/4;
If, DI=3, P (x, y)=(UPPER (6)+UNDER (4))/2;
If, DI=-3, P (x, y)=(UPPER (4)+UNDER (6))/2;
If, DI=4, P (x, y)=(UPPER (7)+UNDER (3))/2;
If, DI=-4, P (x, y)=(UPPER (3)+UNDER (7))/2;
If, DI=5, P (x, y)=(UPPER (8)+UNDER (2))/2;
If, DI=-5, P (x, y)=(UPPER (2)+UNDER (8))/2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410466905.8A CN104202554A (en) | 2014-09-15 | 2014-09-15 | Intra-field anti-aliasing and deinterlacing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410466905.8A CN104202554A (en) | 2014-09-15 | 2014-09-15 | Intra-field anti-aliasing and deinterlacing method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104202554A true CN104202554A (en) | 2014-12-10 |
Family
ID=52087774
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410466905.8A Pending CN104202554A (en) | 2014-09-15 | 2014-09-15 | Intra-field anti-aliasing and deinterlacing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104202554A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104506791A (en) * | 2014-12-25 | 2015-04-08 | 珠海全志科技股份有限公司 | Deinterlacing method and deinterlacing device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101197995A (en) * | 2006-12-07 | 2008-06-11 | 深圳艾科创新微电子有限公司 | Edge self-adapting de-interlacing interpolation method |
CN101442648A (en) * | 2008-12-19 | 2009-05-27 | 四川虹微技术有限公司 | Field interpolation method |
US7944503B1 (en) * | 2006-01-27 | 2011-05-17 | Texas Instruments Incorporated | Interlaced-to-progressive video processing |
CN102868870A (en) * | 2012-09-28 | 2013-01-09 | 许丹 | Deinterlacing processing method |
CN103475838A (en) * | 2013-06-21 | 2013-12-25 | 青岛海信信芯科技有限公司 | Deinterlacing method based on edge self adaption |
-
2014
- 2014-09-15 CN CN201410466905.8A patent/CN104202554A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7944503B1 (en) * | 2006-01-27 | 2011-05-17 | Texas Instruments Incorporated | Interlaced-to-progressive video processing |
CN101197995A (en) * | 2006-12-07 | 2008-06-11 | 深圳艾科创新微电子有限公司 | Edge self-adapting de-interlacing interpolation method |
CN101442648A (en) * | 2008-12-19 | 2009-05-27 | 四川虹微技术有限公司 | Field interpolation method |
CN102868870A (en) * | 2012-09-28 | 2013-01-09 | 许丹 | Deinterlacing processing method |
CN103475838A (en) * | 2013-06-21 | 2013-12-25 | 青岛海信信芯科技有限公司 | Deinterlacing method based on edge self adaption |
Non-Patent Citations (2)
Title |
---|
刘然等: "一种用于DIBR的去隔行算法", 《计算机应用研究》 * |
马斌等: "加权边沿自适应的场内插值去隔行方法", 《计算机应用研究》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104506791A (en) * | 2014-12-25 | 2015-04-08 | 珠海全志科技股份有限公司 | Deinterlacing method and deinterlacing device |
CN104506791B (en) * | 2014-12-25 | 2017-09-22 | 珠海全志科技股份有限公司 | Interlace-removing method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104376319B (en) | A kind of method based on anisotropic Gaussian core extraction closed edge image outline | |
US20150131728A1 (en) | Method for motion vector estimation | |
CN104680483B (en) | The noise estimation method of image, video image denoising method and device | |
JP2012022654A5 (en) | ||
CN104011771A (en) | Method of and apparatus for scalable frame rate up-conversion | |
TW201342916A (en) | Complexity scalable frame rate up-conversion | |
TWI460681B (en) | Method for processing edges in an image and image processing apparatus | |
US20130342444A1 (en) | Method and Apparatus for Hand Gesture Trajectory Recognition | |
JP2013048375A (en) | Device and method for generating motion compensation frame | |
CN104202554A (en) | Intra-field anti-aliasing and deinterlacing method | |
US20150254815A1 (en) | Image downsampling apparatus and method | |
JP2013020605A (en) | Movement image area determination device or method for the same | |
CN103297659A (en) | Edge processing method of image and image processing device | |
WO2013031418A1 (en) | Device for detecting line segment and arc | |
GB2545649A (en) | Artefact detection | |
Silalahi et al. | Comparison of sensitivity analysis on linear optimization using optimal partition and optimal basis (in the simplex method) at some cases | |
CN109215046A (en) | A kind of Laplace operator edge detection method based on image interpolation arithmetic | |
JP2007221602A5 (en) | ||
JP2013228995A5 (en) | ||
CN110537202A (en) | Correlation arithmetic unit | |
CN111383183B (en) | Image edge enhancement method and device and computer storage medium | |
CN107124611B (en) | Method and device for converting video frame rate | |
CN108475339A (en) | For the method and system to the object classification in image | |
JP2008271492A (en) | Noise reducer and reducing method for television video signal | |
JP2017116985A5 (en) |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20141210 |