CN101699856A - De-interlacing method with self-adapting motion - Google Patents
De-interlacing method with self-adapting motion Download PDFInfo
- Publication number
- CN101699856A CN101699856A CN200910235968A CN200910235968A CN101699856A CN 101699856 A CN101699856 A CN 101699856A CN 200910235968 A CN200910235968 A CN 200910235968A CN 200910235968 A CN200910235968 A CN 200910235968A CN 101699856 A CN101699856 A CN 101699856A
- Authority
- CN
- China
- Prior art keywords
- point
- line
- odd
- judge whether
- image
- 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
- Television Systems (AREA)
Abstract
The invention relates to a de-interlacing method with self-adapting motion, which is a method for processing an electronic image. The method comprises the following steps: judging whether an image is an interlaced image or not; judging whether the frames of an odd number are generated or not; judging whether the rows of the odd number are generated or not; judging whether a point is a questionable rest point or not; judging whether the point is a true rest point or not; finally, obtaining a line-by-line video; and outputting a line-by-line video image. The invention divides a current field image into a motion area and a questionable rest area by two similarity degree threshold values, then further divides the questionable rest area into a rest area and a motion area by a third similarity degree threshold value, carries out de-interlacing processing on the rest area by a field reproduction algorithm, carries out de-interlacing processing on the motion area by an ELA algorithm and finally, obtains the line-by-line video.
Description
Technical field
The present invention relates to a kind of de-interlacing method of Motion Adaptive, is the method that a kind of electronic image is handled, and is the method that a kind of video image is handled, and is a kind of with the interleaved method of video image processing of lining by line scan that is treated to.
Background technology
Existing radio data system has two kinds of scan modes: a kind of is (Progressivescanning) mode of lining by line scan, and another kind is interlacing scan (Interlaced scanning) mode.Lining by line scan is in each Δ t time a complete image to be scanned, and it is called as a frame.Interlacing scan is to divide two to carry out interlacing scan in 2: 1 one two field picture.Wherein first is called odd field, only scans its odd-numbered line; And second be called even field, only scans its even number line.The advantage of lining by line scan is that image definition height, image detail are fine and smooth abundant, stable and do not have interline flicker etc.; Shortcoming be need be bigger transmission bandwidth.Interleaved advantage is that the data volume with half obtains higher refresh rate under fixed-bandwidth, thereby reduces flicker; Shortcoming is that image vertical direction resolution reduces half, has phenomenons such as interline flicker and edge sawtooth.
For the quality that promotes video image often needs interlaced video is converted into progressive, the technology that interlaced video is converted into progressive is called (De-interlacing) technology that deinterleaves again in a lot of practical applications.Difference according to interpolation method can be divided into the technology of deinterleaving two big classes: (intra-field) interpolation deinterleaves in, and (inter-field) interpolation deinterleaves between.
Field interpolation is to utilize same the interior picture element and the correlation of interpolation point to carry out interpolation, and it has, and amount of calculation is little, the realization characteristic of simple, is a kind of most effective technology that deinterleaves, thereby is used widely.Mainly comprise: the row iterative method (Line repetition), the row method of average (Line average), along edge interpolation method (Direction-dependent interpolation).But field interpolation deinterleaves and can not promote the vertical resolution of picture, but has caused high-frequency information, makes soft edge, also has phenomenons such as sawtooth, flicker simultaneously.
Interpolation is to utilize opposite field pixel or opposite field and the pixel in this to carry out interpolation between, and it needs to store a field picture information at least, and this has increased the complexity of algorithm usually, but the increase of the algorithm degree of freedom has promoted the quality of De-interlacing.Mainly comprise: the algorithm that deinterleaves of a replica method (Weave), a top-stitching method of average (Inter-field averaging), vertical time domain median filtering method (Vertical-temporal medianfiltering), Motion Adaptive.An its midfield replica method and a top-stitching method of average have the fine effect that deinterleaves to the stagnant zone in the image sequence, but then can produce pseudomorphism to the moving region.Though vertical time domain median filtering method has implied the self-adaptive processing to stagnant zone and moving region, and has improved the vertical resolution and the image quality of image to a certain extent, vertical detail runs off to some extent, also can introduce aliasing phenomenon.The algorithm that deinterleaves of Motion Adaptive is exactly by detecting the moving region in the video sequence, is stagnant zone and moving region with image division, then different interpolation algorithms is adopted in these two zones.Based on deinterleaving of motion compensation (Motion Compensation), be the algorithm that deinterleaves of a kind of advanced person's Motion Adaptive, can obtain very high-quality image on the basis of motion vector accurately calculating.Although the motion compensation technology of deinterleaving is optimum, if a given motion model, and motion vector can calculate accurately; The technology but motion compensation deinterleaves, design complexity are very high, operand is very big, responsive especially to the motion vector error, real-time is poor.The high-quality technology that deinterleaves generally adopts hardware to realize in order to promote efficient, so cost is than higher.In actual applications, in order to reduce development cost, promote efficient, need research high-quality, high efficiency, the more sane technology that deinterleaves.
Summary of the invention
Defective at prior art, the objective of the invention is to propose a kind of de-interlacing method of Motion Adaptive, the present invention makes the mode of difference combination between use interpolation value and field, do not carry out the calculating of motion vector, the motion of only carrying out between the field is compared, calculate greatly and simplify, can not use the expensive hardware that deinterleaves, use the software computing that is easy to realize also can obtain high-quality progressive scanning picture.
The object of the present invention is achieved like this: a kind of de-interlacing method of Motion Adaptive, described method comprises the steps:
Judge whether step for horizontally interlaced image: be used to judge whether the video image of input is horizontally interlaced image, if "No" then enter " progressive image output step " directly output video image, if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered frame: be used to judge whether that interlaced video by current input produces odd-numbered frame video line by line, if "No" then show generation even frame video line by line, promptly enter " each step that the dual numbers frame is handled ", if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered line: be used to judge whether current insertion point is the data of odd-numbered frame odd-numbered line, if "Yes" then directly fill and obtain the odd-numbered frame data by the corresponding pixel of current odd field, if "No" then carry out next step;
Judge whether it is the step of Dubious static point: be used to judge whether current insertion point is the Dubious static point, whether the difference of promptly judging the respective detection point of this point satisfies current similarity threshold and auxiliary similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain this pixel value, if "Yes" then enter next step;
Judge whether it is the step of real rest point: be used to judge whether current Dubious static point is real rest point, whether the difference of promptly judging the respective detection point of this point satisfies the correction similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain the pixel value of this point, if "Yes" then enter next step;
Finally obtain the step of progressive: be used for producing progressive by motor point and rest point, adopt the ELA algorithm at maintenance edge to obtain this pixel value to the motor point, adopt a replication strategy to obtain this pixel value for rest point, finally obtain odd-numbered frame video line by line;
For the production process of even frame line by line, in full accord with the process that produces odd-numbered frame line by line, the relevant position difference of respective detection point only;
The step of progressive image output: be used for and export through the complete progressive scanning picture that inserts row.
The beneficial effect that the present invention produces is: the present invention is a kind of high-quality, high efficiency, the technology that deinterleaves cheaply, the present invention at first adopts two similarity thresholds that current field picture is divided into moving region and Dubious static zone, then to the Dubious static zone, adopt the 3rd similarity threshold further it to be divided into stagnant zone and moving region, adopt a replication strategy to go interlacing to handle to stagnant zone at last, adopt the ELA algorithm to go interlacing to handle to the moving region, finally obtain progressive.Test shows, the present invention is a kind of high-quality, the high efficiency technology that deinterleaves, and has the characteristics simple, that cost is low that realize.This technology also can be used as a basic ancillary technique of other algorithm and technology, finally reaches the purpose that promotes efficient and quality.
Description of drawings
The invention will be further described below in conjunction with drawings and Examples.
Fig. 1 is the FB(flow block) of the embodiment of the invention one described method;
Fig. 2 is at (y, t) the continuous four field picture data projection schematic diagrames of an interlaced video on the coordinate;
Fig. 3 is odd field involved picture element and concern schematic diagram when being inserted into a some motion detection;
Fig. 4 is even field involved picture element and concern schematic diagram when being inserted into a some motion detection;
Fig. 5 is the field replication strategy schematic diagram of rest point in the odd field;
Fig. 6 is the ELA algorithm schematic diagram of 3 * 3 windows.
Embodiment
Embodiment one:
Present embodiment is a kind of de-interlacing method of Motion Adaptive, and described method comprises the steps:
Judge whether step for horizontally interlaced image: be used to judge whether the video image of input is horizontally interlaced image, if "No" then enter " progressive image output step " directly output video image, if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered frame: be used to judge whether that interlaced video by current input produces odd-numbered frame video line by line, if "No" then show generation even frame video line by line, promptly enter " each step that the dual numbers frame is handled ", if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered line: be used to judge whether current insertion point is the data of odd-numbered frame odd-numbered line, if "Yes" then directly fill and obtain the odd-numbered frame data by the corresponding pixel of current odd field, if "No" then carry out next step;
Judge whether it is the step of Dubious static point: be used to judge whether current insertion point is the Dubious static point, whether the difference of promptly judging the respective detection point of this point satisfies current similarity threshold and auxiliary similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain this pixel value, if "Yes" then enter next step;
Judge whether it is the step of real rest point: be used to judge whether current Dubious static point is real rest point, whether the difference of promptly judging the respective detection point of this point satisfies the correction similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain the pixel value of this point, if "Yes" then enter next step;
Finally obtain the step of progressive: be used for producing progressive by motor point and rest point, adopt the ELA algorithm at maintenance edge to obtain this pixel value to the motor point, adopt a replication strategy to obtain this pixel value for rest point, finally obtain odd-numbered frame video line by line; For the production process of even frame line by line, in full accord with the process that produces odd-numbered frame line by line, the relevant position difference of respective detection point only;
The step of progressive image output: be used for and export through the complete progressive scanning picture that inserts row.
The specific implementation of present embodiment is such: at first obtain 4 continuous video datas, be respectively former frame odd field, former frame even field, present frame odd field and present frame even field, and be designated as f respectively
N-2, f
N-1, f
n, f
N+1, f wherein
n, f
N+1Be respectively current odd field and current even field.This technology finally is will be with f
nAnd f
N+1Two interlaced video datas are separately converted to two frame progressive data, and this two frames progressive is called odd-numbered frame and even frame.For the odd-numbered frame or the even frame that generate, the corresponding row that corresponding odd field or Even Fields number can intactly be put in the frame is gone, and even number of lines certificate in the odd-numbered frame and the odd-numbered line data in the even frame, be to need to calculate to produce, the algorithm of its generation is realized by the present invention.
Below with s (x, y, t
i), i=e, o represent t even field or odd field constantly respectively, and wherein x represents horizontal direction, and y represents vertical direction, as shown in Figure 2.S when y gets odd number value (x, y, t
e)=0; S when y gets even number value (x, y, t
o)=0.Fig. 2 is illustrated in (y, t) continuous four projections of an interlaced video on the coordinate; Wherein, each circle is represented the cross section of a complete line of video; The circle that has hatching is represented actual available row; Empty circles is represented the row that needs insert to calculate, and index value capable in suppose every is since 1.
The technology that deinterleaves of Motion Adaptive, its characteristics are motion detection, this Technology Need is used three similarity thresholds and is carried out motion detection, these three threshold values are designated as threshold0 respectively, threshold1, threshold2. the first step of this detection method is, current strange and even is divided for the first time.To current odd field data, need to judge to be inserted into a s (x, y, t
o) (y is an even number) (point that has dotted line among Fig. 3) be the motion or static, adopt following criterion that it is carried out preliminary judgement here.
|s(x，y-1，t
o)-s(x，y-1，t
o-1)|≤threshold0(1)
|s(x，y+1，t
o)-s(x，y+1，t
o-1)|≤threshold0(2)
|s(x，y，t
e)-s(x，y，t
e-1)|≤threshold1(3)
If satisfy top three inequality simultaneously, then with s (x, y, t
o) be labeled as the point of Dubious static, otherwise it is labeled as the motor point.Inequality (1) and (2) are that corresponding pixel differences between two strange fields is carried out threshold decision, and inequality (3) is that corresponding pixel differences between two idol fields is carried out threshold decision.The threshold0 here is called current similarity threshold, and threshold1 is called auxiliary similarity threshold.Be not difficult to find out by three inequality, (current odd field is inserted into corresponding two pixel values of two row up and down a little need to use 6 pixel values in 4 video datas to the judgement of this point, before two pixel values of the corresponding and current odd field of frame odd field, the pixel value of current even field respective point, the pixel value of preceding frame even field respective point) and two similarity thresholds.By Fig. 3 this problem can be described easily.
To current Even Fields number, need to judge to be inserted into a s (x, y, t
e) (y is an odd number) (point that has dotted line among Fig. 4) be the motion or static, adopt following criterion that it is carried out preliminary judgement here.
|s(x，y-1，t
e)-s(x，y-1，t
e-1)|≤threshold0(4)
|s(x，y+1，t
e)-s(x，y+1，t
e-1)|≤threshold0(5)
|s(x，y，t
o)-s(x，y，t
o-1)|≤threshold1(6)
If satisfy top three inequality simultaneously, then with s (x, y, t
e) be labeled as the Dubious static point, otherwise it is labeled as the motor point.The judgement of this point is also needed to use 6 pixel values in 4 video datas, its position and relation, as shown in Figure 4.
Next step of motion detection is on the basis of previous step, and the Dubious static point is further proofreaied and correct division.In going interlacing, under the situation that is not having motion, optimum merge idol and strange when tactful, the field method of duplicating just, this method definition height, information lossless lose.But under the situation of motion, this technology will cause the motion virtual image.What therefore the accuracy of motion detection just showed is extremely important, and it directly has influence on the quality of final image.Accuracy for lifter motion detects adopts following criterion to judge to Dubious static point in the strange field.
|s(x，y，t
e-1)-s(x，y-1，t
o)|≤threshold2(7)
|s(x，y，t
e-1)-s(x，y+1，t
o)|≤threshold2(8)
If (7), have at least an inequality to set up in (8) formula, then will put s (x, y, t
o) be labeled as real rest point, otherwise will put s (x, y, t
o) be labeled as the motor point.
Dubious static point adopts following criterion to judge in the dual numbers field.
|s(x，y，t
e)-s(x，y-1，t
o)|≤threshold2(9)
|s(x，y，t
e)-s(x，y+1，t
o)|≤threshold2(10)
If (9), have at least an inequality to set up in (10) formula, then will put s (x, y, t
e) be labeled as real rest point, otherwise will put s (x, y, t
e) be labeled as the motor point.Above threshold2 in four inequality be called the rectification similarity threshold.
After the motion detection through top two steps, when the front court video finally has been divided into stagnant zone and moving region.To stagnant zone, adopt the field replication strategy of optimal policy, can be expressed as
S (x, y, t
o)=s (x, y, t
E-1) y is even number (11)
And
S (x, y, t
e)=s (x, y, t
o) y is odd number (12)
The field replication strategy can be illustrated by Fig. 5, puts s (x, y, t here
o) be the rest point in the odd field, then copy the sampled value (representing the place with horizontal arrow among Fig. 5) of relevant position in the front court.In the practical video image, often there is a large amount of stagnant zones, these stagnant zones are adopted a replication strategy, can obtain the optimum interlacing effect of going, have the highest efficient simultaneously.
For the sport interpolation point, present embodiment adopts and keeps the LEA algorithm at edge to obtain its pixel value.ELA (Edge-based Line Average) is promptly based on the line average algorithm at edge, this algorithm is a kind of field interpolation algorithm, it can go interlacing handle in reconstructed image side information more accurately, can eliminate simultaneously flicker and blooming, so this algorithm has application widely in going interlacing.The realization of this algorithm is described below, and is that 3 * 3 ELA algorithm is an example with window size here.
If (x, y) for be inserted into the position of a little doing in the field, its pixel value is by X=s (x, y) expression, s (x-1, y-1), s (x, y-1), s (x+1 is corresponding three pixel values of its lastrow y-1), respectively with A, and B, C represents, and s (x-1, y+1), s (x, y+1), (x+1 y+1) is corresponding three pixel values of its next line to s, respectively with D, E, F represents.And calculate following various:
a＝|A-F|???????????????????(13)
b＝|B-E|???????????????????(14)
c＝|C-D|???????????????????(15)
Then X can be obtained by following formula:
Fig. 6 is the ELA algorithm of 3 * 3 windows, and wherein X represents with empty circles for being inserted into a little, and solid circles is actual original sample point.From (16) formula as can be seen, the ELA algorithm is actually and carries out interpolation calculation along the edge of image, if therefore the edge calculates accurately, this algorithm can obtain effect preferably.As seen from Figure 6, the ELA algorithm of 3 * 3 windows has good effect to the hypotenuse of 45 ° and 135 °; But for the hypotenuse of low angle (less than 30 °), this algorithm will produce sawtooth and scintillation.Can (5 * 5,7 * 7,9 * 9,11 * 11) adopt the LEA algorithm to promote the quality of image in bigger window to this, but can increase amount of calculation like this, this algorithm synthesis be considered efficiency, has finally adopted the ELA algorithm of 3 * 3 windows.
Present embodiment is by motion detection and different interpolation strategies, can obtain the progressive of better quality, wherein the core is three threshold strategies of motion detection, the field replication strategy of stagnant zone, both can promote efficient greatly, have optimum effect again, the ELA algorithm of moving region is realized simple, and can keep the image side information.
Present embodiment is through detecting, and picture quality significantly improves than prior art, and testing environment and data are as follows:
The evaluation of algorithm performance of deinterleaving has subjective and objective two kinds of methods.Subjective assessment generally adopt human eye watch video image whether clear, whether phenomenons such as flicker, sawtooth arranged.Objective evaluation generally adopts calculating and compares Y-PSNR (PSNR) and mean square deviation (MSE), and the PSNR has here characterized the overall performance of picture quality, is the bigger the better, and MSE characterizes the mean difference of source and target, and is the smaller the better.For the performance of objective appraisal present embodiment, (Doubling, Average ELA) carry out PSNR and MSE relatively with other traditional algorithms with present embodiment.Concrete method of testing is as follows: at first test material line by line by standard and obtain the interlacing material, and to the processing that deinterleaves of this interlacing material, then calculate mean square deviation (MSE) and Y-PSNR (PSNR), MSE by analyzing other algorithms and present embodiment and PSNR finish the performance evaluation to present embodiment at last.Concrete test environment is: Dan Sihe Intel E5504 2.00GHz CPU, 2.00GB internal memory.The test cycle tests adopts general in the world 8 standard test sequences (Foreman, Akiyo, News, Paris, Mother-Daughter, Mobile, Miss, Stefan).(these sequences are general in the world standard test sequences, and its title has been represented the name of a cycle tests respectively)
Table one: the comparison of present embodiment and Foreman, Akiyo, News, Paris,
| Algorithm deinterleaves | Evaluation criterion | ??Foreman??(176×144) | ?Akiyo(176×144) | News(352×28?8) | ?Paris(355×288) |
| ??Doubling | ??MSE??PSNR(dBs) | ??118.37??27.41 | ??52.79??30.91 | ??88.77??28.66 | ??285.85??23.58 |
| ??Average | ??MSE??PSNR(dBs) | ??39.70??32.16 | ??18.00??35.59 | ??25.12??31.16 | ??141.22??26.64 |
| ??ELA | ??MSE??PSNR(dBs) | ??26.31??33.96 | ??23.87??34.37 | ??43.95??31.72 | ??188.93??25.37 |
| Present embodiment | ??MSE??PSNR(dBs) | ??20.90??35.07 | ??7.63??40.87 | ??11.96??37.40 | ??64.19??30.06 |
Table two: the comparison of present embodiment and Mother-Daughter, Mobile, Miss, Stefan,
| Algorithm deinterleaves | Evaluation criterion | ?Mother-Daughther(355×288) | Mobile(176×144) | ??Miss??(176×144) | ?Stefan(355×288) |
| ??Doubling | ??MSE??PSNR(dBs) | ??29.31??33.49 | ??530.86??20.89 | ??19.00??35.36 | ??301.31??23.35 |
| ??Average | ??MSE??PSNR(dBs) | ??8.02??39.14 | ??261.87??23.96 | ??5.85??40.48 | ??112.40??27.65 |
| Algorithm deinterleaves | Evaluation criterion | ?Mother-Daughther(355×288) | Mobile(176×144) | ??Miss??(176×144) | ?Stefan(355×288) |
| ??ELA | ??MSE??PSNR(dBs) | ??10.10??38.13 | ??314.50??23.16 | ??7.50??39.40 | ??154.23??26.28 |
| Present embodiment | ??MSE??PSNR(dBs) | ??3.64??42.76 | ??302.21??23.33 | ??4.75??41.48 | ??140.17??26.72 |
By the test data of table one, table two, be not difficult to find out most cycle tests present embodiments have been compared significant advantage with other algorithms: PSNR has significantly lifting, and MSE declines to a great extent.(Mobile, Stefan), the advantage of present embodiment is also not obvious, its PSNR even can be lower than the Average algorithm but for the violent especially cycle tests of motion.Advantage for such cycle tests present embodiment is embodied in the subjective assessment, and video image flicker, crenellated phenomena that the Average algorithm obtains are serious, and the video image that present embodiment obtains is clear, stable, sawtooth is few.So comprehensive, present embodiment is compared with traditional algorithm has significant advantage, can significantly promote the quality of video image.Concerning general HD video, most of zone of video image is a stagnant zone, adopt its efficient of this example to promote greatly, test shows is the HD video of 1920 * 1080i of YUVP422 to color format, and its efficient that deinterleaves can reach the super real-time processing of 30 frame/seconds.
Embodiment two:
Present embodiment is to implement one improvement, is that embodiment one is about the similarity threshold refinement.Present embodiment is described to judge whether it is to be 3-9 during the value of the similarity threshold in the step of Dubious static point.
Inequality among the embodiment one (1) and (2) are that corresponding pixel differences between two strange fields is carried out threshold decision, and inequality (4) and (5) are that corresponding pixel differences between two even fields is carried out threshold decision.The threshold value span of these four inequality is 3-9, wherein 6 best results.This threshold value is lower than in 3 images overwhelming majority point and is divided into the motor point, and the efficient of algorithm can reduce like this, and image definition decline simultaneously, sawtooth increase.This threshold value can be labelled unjustifiably the motor point in the image greater than 9 and be rest point, and image flicker, crenellated phenomena will highlight.
Embodiment three:
Present embodiment is to implement one improvement, is that embodiment one is about auxiliary similarity threshold refinement.Present embodiment is described to judge whether it is to be 5-11 during the value of the auxiliary similarity threshold in the step of Dubious static point.
Inequality among the embodiment one (3) is that corresponding pixel differences between two even fields is carried out threshold decision, inequality (6) is that corresponding pixel differences between two odd fields is carried out threshold decision, the threshold value span of these two inequality is 4-12, and wherein 8 is optimum.This threshold value is lower than 4 can carry out a step and correct and be the motor point implementing to be divided in two static point, and efficiency of algorithm can reduce like this, and the definition of image decline simultaneously, sawtooth increase.This threshold value can further be corrected the point that has been divided into motion in the enforcement two greater than 12 and is rest point, and image flicker, crenellated phenomena will increase.
Embodiment four:
Present embodiment is to implement one improvement, is that embodiment one is about proofreading and correct the similarity threshold refinement.Present embodiment is described to judge whether it is to be 10-30 during the value of the correction similarity threshold in the step of real rest point.
Inequality among the embodiment one (7) and (8) be to the insertion point pixel in the odd field and its up and down two pixel value differences carry out threshold decision.Inequality (9) and (10) be insertion point pixel in the dual field field and its up and down two pixel value differences carry out threshold decision.The threshold value span of these four inequality is 10-30, and wherein 20 is optimum.This threshold value is lower than 10 can be judged as the motor point with this some rectification, and efficiency of algorithm can reduce like this, and image definition descends simultaneously, and sawtooth increases.This threshold value no longer includes the rectification ability substantially greater than 30, and overwhelming majority point is handled according to rest point, and efficiency of algorithm has certain lifting like this, waits owing to adopt a replication strategy, may produce pseudomorphism.
These three threshold values are designated as threshold0 to ten inequality and three threshold values for the division of finishing image is shared among the embodiment one, threshold1, and threshold2 is called current similarity threshold, and auxiliary similarity threshold is corrected similarity threshold.Explain the implication of these inequality below respectively.
Inequality (1) and (2) are corresponding pixel differences between two strange fields to be carried out current similarity threshold judge, show that the pixel value variation between former and later two strange is very little if satisfy inequality (1) and (2), show as rest point or the very little point of motion from visually seeing.Etc. (3) are not that corresponding pixel differences between two even fields assist the similarity threshold decision, if satisfy this inequality then to show that pixel value between former and later two even fields changes very little, show as rest point or the very little point that moves from visually seeing.Comprehensively (1), (2), (3) formula can be judged current situation of change in four videos in front and back.If respective point changes very for a short time between former and later two odd fields, respective point changes very for a short time between two even fields simultaneously, and then this point is divided into the Dubious static point.Inequality (7) and (8) are to judge by the pixel differences of insertion point and respective point being corrected similarity threshold, show existence and the corresponding to picture element in insertion point in the relevant position of inserting the field if satisfy inequality (7) or (8), then this point finally is judged as rest point, otherwise is the motor point.The meaning of inequality (4), (5), (6), (9), (10) is identical with the realization function with the meaning that realizes function and inequality (1), (2), (3), (7), (8), the division to current odd field is finished in different is inequality (1), (2), (3), (7), (8), and the division to current even field is finished in inequality (4), (5), (6), (9), (10).
It should be noted that at last, below only unrestricted in order to technical scheme of the present invention to be described, although the present invention is had been described in detail with reference to the preferred arrangement scheme, those of ordinary skill in the art is to be understood that, can be to technical scheme of the present invention (such as increasing the rank of judging, or carry out motor point difference or the like with other modes) make amendment or be equal to replacement, and do not break away from the spirit and scope of technical solution of the present invention.
Claims (4)
1. the de-interlacing method of a Motion Adaptive, described method comprises the steps:
Judge whether step for horizontally interlaced image: be used to judge whether the video image of input is horizontally interlaced image, if "No" then enter " progressive image output step " directly output video image, if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered frame: be used to judge whether that interlaced video by current input produces odd-numbered frame video line by line, if "No" then show generation even frame video line by line, promptly enter " each step that the dual numbers frame is handled ", if "Yes" then enter next step;
Judge whether to produce the step of odd-numbered line: be used to judge whether current insertion point is the data of odd-numbered frame odd-numbered line, if "Yes" then directly fill and obtain the odd-numbered frame data by the corresponding pixel of current odd field, if "No" then carry out next step;
Judge whether it is the step of Dubious static point: be used to judge whether current insertion point is the Dubious static point, whether the difference of promptly judging the respective detection point of this point satisfies current similarity threshold and auxiliary similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain this pixel value, if "Yes" then enter next step;
Judge whether it is the step of real rest point: be used to judge whether current Dubious static point is real rest point, whether the difference of promptly judging the respective detection point of this point satisfies the correction similarity threshold, if "No" then show it is the motor point, directly adopt the ELA algorithm to obtain the pixel value of this point, if "Yes" then enter next step;
Finally obtain the step of progressive: be used for producing progressive by motor point and rest point, adopt the ELA algorithm at maintenance edge to obtain this pixel value to the motor point, adopt a replication strategy to obtain this pixel value for rest point, finally obtain odd-numbered frame video line by line;
For the production process of even frame line by line, in full accord with the process that produces odd-numbered frame line by line, the relevant position difference of respective detection point only;
The step of progressive image output: the progressive scanning picture that is used for obtaining after deinterleaving is exported.
2. the described de-interlacing method of claim 1 is characterized in that, described judging whether is to be 3-9 during the value of the similarity threshold in the step of Dubious static point.
3. the described de-interlacing method of claim 1 is characterized in that, described judging whether is to be 4-12 during the value of the auxiliary similarity threshold in the step of Dubious static point.
4. the described de-interlacing method of claim 1 is characterized in that, described judging whether is to be 10-30 during the value of the correction similarity threshold in the step of real rest point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102359681A CN101699856B (en) | 2009-10-30 | 2009-10-30 | De-interlacing method with self-adapting motion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102359681A CN101699856B (en) | 2009-10-30 | 2009-10-30 | De-interlacing method with self-adapting motion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101699856A true CN101699856A (en) | 2010-04-28 |
| CN101699856B CN101699856B (en) | 2012-01-18 |
Family
ID=42148296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009102359681A Active CN101699856B (en) | 2009-10-30 | 2009-10-30 | De-interlacing method with self-adapting motion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101699856B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102487443A (en) * | 2010-12-03 | 2012-06-06 | 奇扬网科股份有限公司 | Coding method of screen frame and electronic device applying the same |
| CN102611829A (en) * | 2012-02-03 | 2012-07-25 | 安徽云森物联网科技有限公司 | Image de-interleaving method based on block partitioning |
| CN103686097A (en) * | 2012-09-25 | 2014-03-26 | 深圳市赛格导航科技股份有限公司 | A video de-interleave method |
| CN106027943A (en) * | 2016-07-11 | 2016-10-12 | 北京大学 | Video de-interlacing method |
| WO2017101348A1 (en) * | 2015-12-14 | 2017-06-22 | 乐视控股（北京）有限公司 | Method and device for deinterlacing interlaced videos |
| WO2024001345A1 (en) * | 2022-06-30 | 2024-01-04 | 深圳市中兴微电子技术有限公司 | Image processing method, electronic device, and computer storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2742900B1 (en) * | 1995-12-22 | 1998-02-13 | Thomson Multimedia Sa | METHOD FOR INTERPOLATING PROGRESSIVE FRAMES |
| CN100464580C (en) * | 2006-06-12 | 2009-02-25 | 北京中星微电子有限公司 | Motion detecting and uninterlacing method for digital video frequency processing and apparatus thereof |
| CN101197997A (en) * | 2007-12-26 | 2008-06-11 | 上海广电集成电路有限公司 | De-interlacing method and system based on dynamic threshold value movement and edge self-adaption |
| CN101282430A (en) * | 2008-04-25 | 2008-10-08 | 伍刘 | Method for converting video signal with interlacing and non-interlaced mode |
-
2009
- 2009-10-30 CN CN2009102359681A patent/CN101699856B/en active Active
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102487443A (en) * | 2010-12-03 | 2012-06-06 | 奇扬网科股份有限公司 | Coding method of screen frame and electronic device applying the same |
| CN102487443B (en) * | 2010-12-03 | 2016-05-18 | 奇扬网科股份有限公司 | The coding method of screen frame and the electronic installation of application thereof |
| CN102611829A (en) * | 2012-02-03 | 2012-07-25 | 安徽云森物联网科技有限公司 | Image de-interleaving method based on block partitioning |
| CN102611829B (en) * | 2012-02-03 | 2014-09-24 | 安徽云森物联网科技有限公司 | Image de-interleaving method based on block partitioning |
| CN103686097A (en) * | 2012-09-25 | 2014-03-26 | 深圳市赛格导航科技股份有限公司 | A video de-interleave method |
| CN103686097B (en) * | 2012-09-25 | 2017-02-08 | 深圳市赛格导航科技股份有限公司 | A video de-interleave method |
| WO2017101348A1 (en) * | 2015-12-14 | 2017-06-22 | 乐视控股（北京）有限公司 | Method and device for deinterlacing interlaced videos |
| CN106027943A (en) * | 2016-07-11 | 2016-10-12 | 北京大学 | Video de-interlacing method |
| CN106027943B (en) * | 2016-07-11 | 2019-01-15 | 北京大学 | A kind of video interlace-removing method |
| WO2024001345A1 (en) * | 2022-06-30 | 2024-01-04 | 深圳市中兴微电子技术有限公司 | Image processing method, electronic device, and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101699856B (en) | 2012-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1265633C (en) | Interlacing-removing device and method | |
| CN1270526C (en) | Device and method for using adaptive moving compensation conversion frame and/or semi-frame speed | |
| US6473460B1 (en) | Method and apparatus for calculating motion vectors | |
| US6331874B1 (en) | Motion compensated de-interlacing | |
| CN101699856B (en) | De-interlacing method with self-adapting motion | |
| CN102025960B (en) | Motion compensation de-interlacing method based on adaptive interpolation | |
| CN103369208B (en) | Self adaptation interlace-removing method and device | |
| CN101088290B (en) | Spatio-temporal adaptive video de-interlacing method, device and system | |
| CN102946505B (en) | Self-adaptive motion detection method based on image block statistics | |
| CN102045530B (en) | Motion adaptive deinterleaving method based on edge detection | |
| Heinrich et al. | Optimization of hierarchical 3DRS motion estimators for picture rate conversion | |
| US20090219439A1 (en) | System and Method of Deinterlacing Interlaced Video Signals to Produce Progressive Video Signals | |
| CN101600062B (en) | In-field deinterlacing method based on enhanced edge adaptivity | |
| CN108259903B (en) | H.265 sample self-adaptive compensation method based on human eye region of interest | |
| Chen et al. | True motion-compensated de-interlacing algorithm | |
| CN106303338B (en) | A kind of in-field deinterlacing method based on the multi-direction interpolation of bilateral filtering | |
| CN102170549A (en) | Edge correlation image intra-field de-interlacing algorithm of edge pre-judgment | |
| CN101437137B (en) | Field interpolation method | |
| CN101309376A (en) | Method and device for eliminating alternate line | |
| US20030026498A1 (en) | Method for converting the scanning format of images, a system and computer program product therefor | |
| Tai et al. | A motion and edge adaptive deinterlacing algorithm | |
| CN102497523B (en) | Median-filtering de-interlacing method | |
| CN1529500A (en) | Three-dimensional video format conversion method based on motion adaption and marginal protection | |
| CN105611214A (en) | Method for de-interlacing through intra-field linear interpolation based on multidirectional detection | |
| CN110381324A (en) | A kind of frame interpolated value interlace-removing method based on edge detection |
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 |