CN101692691B - Method for detecting bands in video images - Google Patents
Method for detecting bands in video images Download PDFInfo
- Publication number
- CN101692691B CN101692691B CN2009101965913A CN200910196591A CN101692691B CN 101692691 B CN101692691 B CN 101692691B CN 2009101965913 A CN2009101965913 A CN 2009101965913A CN 200910196591 A CN200910196591 A CN 200910196591A CN 101692691 B CN101692691 B CN 101692691B
- Authority
- CN
- China
- Prior art keywords
- band
- image
- frame
- surveyed area
- horizontal
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting bands in video images. The method comprises the following steps: 1, allowing a user to choose the color and type of a to-be-repaired band according to band characteristics appearing in a video sequence and to determine a detection region according to a position where a band in a first frame appears; 2, judging the choice of the user and entering a step 3 for processing if a vertical band is chosen or entering a step 4 for processing if a horizontal band is chosen; 3, detecting the vertical band to obtain a band template drawing, generating the detection region of next frame and cyclically using the step to process follow-up frames; and 4, detecting the vertical band to obtain the band template drawing and cyclically using the step to process the follow-up frames. Compared with other band-detecting methods in the prior art, the method has the advantages of detecting various bands in the video images, improving reliability, allowing users to only need to perform interaction on the first frame having bands, completely automatically detecting the follow-up frames and improving detection efficiency.
Description
Technical field
The invention belongs to the digital video image process field, relate to the detection method of banded zone in a kind of video image, be used for detecting the band damage of video image.
Background technology
The birth of film has had the history in more than 100 year so far; Be accompanied by the birth and the development of motion picture technique; The human history accumulation the real historical image archives material of many records with witness the television data of culture and arts development, these data are the legacy of human history preciousness.Because historical technical reason, old video can produce damage when preserving and play, and need repair it, also its original appearance.The maturation of modern digital technology makes us can adopt digital technology that old video is repaired.In repair process, detect damage field automatically and will improve remediation efficiency, be the basis of realizing that old video image automation is repaired, be the prerequisite that improves old video image remediation efficiency.
Band is a kind of damage that is prevalent in the old video image, and it is physical abrasion that band produces reason.Normally in when projection, scraping institute causes on cinefilm or the tape motion direction being parallel to by hard particle.Reparation to this damage will remove the band object exactly from sequence of video images, need provide the exact position of band.Because number is huge, artificial demarcation is obviously unactual, need the band in the video image be detected automatically, obtains the precise region of band present position.For this reason, the detection of band research just seems and is even more important.
The method that detects about band the earliest is the model analysis method, like ANIL Kokaram, and Detection and removal of line scratches in degraded motion picture sequences; Signal Processing VIII, volume I, pages 5-8; September 1996 and Vittoria Bruni and Domenico Vitulano, A generalized model for scratch detection, IEEE Transaction in image processing; Vol.13; No.1, January 2004, and this method is set up model to band earlier; Through the random device corresponding numerical value that gets parms, the reasonability of analyzing numerical value is confirmed the existence of band.This modelling is that the hypothesis band continues up to last column from first trip in image; And fully linearly distribute, this hypothesis does not have generality, and actual band damage much is not linear; And only appear at the regional area of image, be not suitable for using this model description.
The locality that in image, occurs the zone to band; Produced the regional analysis method; Like Rong-Chi Chang; Louis H.Lin, Chia-Ton Tian, and Timothy K.Shih.Video Inpainting and Restoration Techniques.Proceedings of the 13th annual ACM international conference on Multimedia.Nov.2005 and Timothy K.Shih and Louis H.Lin; Wonjun Lee; Detection and removal of long scratch line in aged film.Multimedia and Expo, IEEE International Conference on July 2006Page (s): 477-480, this method is to be divided into several regions to the video image level; Test strip in each zone couples together adjacent band afterwards.Through the existence of analysis image at the brightness projection analysis band of column vector, the size of block has considerable influence for testing result to this method, especially when gradient is bigger in horizontal cut zone.Though this method has been considered band and the locality characteristic in zone occurred, still do not considered the gradient problem.
The 3rd quasi-nonlinear searching method; Like Lazhar Khriji; Mahmoud Meribout; Moncef Gabbouj.Detection and removal of video defects using rational-based techniques.Advances in Engineering Software 36 (2005) 487-495 are based on the analytic approach of pixel.This searching method is: by the pixel traversal, compare current pixel and next pixel in the horizontal direction, obtained two relevant pixels if luminance difference is then thought less than threshold value; Then detect next pixel; Reach a max-thresholds up to the related pixel number, think background, otherwise think possible band pixel less than threshold value; To the dimensionality reduction of on column direction, averaging by the image of above processing acquisition; For thinking band greater than the row of average threshold value, get back to former figure afterwards, delete the non-strip portion that this lists according to an image brightness threshold value.This method only is suitable for the even brightness band and detects, and for general band, brightness presents the cosine decay, and this method will lose efficacy.
The 4th type be based on wavelet theory and morphologic detection method as: the cinefilm cut based on wavelet decomposition damages digital recovery technique research, Chinese image graphics journal, Vol 11; No.11; Nov.2006, this method is earlier image to be carried out wavelet decomposition, thinks possible band to the little position of the big vertical dimension coefficients collection of horizontal dimension coefficients collection; Eliminate the small size zone through morphologic expansion erosion algorithm afterwards; The expansion that re-uses level and vertical direction at last is connected interrupted band, because size and bias size in the middle of interrupted all are not sure of, the selection of size of structure element is big for the result influence in the morphologic use.
Through the analysis of band in a large amount of video images, find that band can be divided into two kinds in the video image: vertical strip and horizontal band.The length of vertical strip is generally greater than 5% of video image width.Vertical strip can be divided into black vertical strip and white vertical band again.Black vertical strip gradient and vertical direction angle are less than 2 degree, and major part is similar to vertically, generally is linear distribution, and width is less than 5 pixels; White vertical band gradient and vertical direction angle are less than 8 degree, and width is less than 7 pixels.Horizontal bar is taken to toward being the standard straight-line kenel.Horizontal band also can be divided into white level band and black level band.Black level band brightness is obvious, with the surrounding brightness difference greater than 8, its strip width is less, is generally one to three pixel.Its length is generally greater than 5% of video image length.The white level strip width is bigger, generally all about ten pixels, and is interrupted transparence, and its brightness and surrounding brightness contrast are not obvious.
The existing method of strip analysis discovery has following deficiency in the restored video through treating:
(1), existing detection method is to detect to some typical particular vertical bands; Regional locality characteristics appear in gradient and the band of having ignored band; And think not have similar band object in the background, in actual repair, be not applicable to the detection method of various situation vertical strips.
(2), the research that detects for horizontal band seldom; Even mention; Think that also the detection method of horizontal band and vertical strip is the difference on direction, ignored the big and transparent characteristics of brightness of horizontal bar bandwidth fully, not can be used as simple direction difference and handle.
(3), mostly existing detection method be the single frames processing method, do not solve successive frame and handle problems.
Summary of the invention
The problem and shortage that exists of prior art in view of the above; The object of the present invention is to provide the detection method of band in a kind of video image; Can detect linearity, little gradient vertical strip and non-linear, big gradient vertical strip, also can detection level black, white ribbon, realize that the mutual subsequent frame of first frame detects automatically; Reach the flase drop when being reduced in background and having similar band object, improve detection efficiency and reliability.
For achieving the above object; The present invention adopts following technical solution to realize: at first by the user according to the band characteristic that occurs in the first frame video sequence; Select the color (white or black) and type of strip (level or vertical) of band to be repaired, and delimit surveyed area, in surveyed area, carry out automatic extraction based on the band feature object of spatial brightness information according to the appearance position of band in the first frame; And the surveying record object shapes, positional information; Then the band feature object that extracts is analyzed, merged, generate the two-value template of band; Again based on the band of present frame to the picture testing result, according to the band drift features, the zone that prediction next frame band occurs generates the surveyed area of next frame; According to new surveyed area next frame being carried out band at last detects.Its concrete steps are following:
Step 1, select the color (white or black) and type of strip (level or vertical) of band to be repaired according to the band in the first frame video, take existing position out of according to first frame discal patch and delimit surveyed area, calculate the length and the width of surveyed area by the user;
Step 2, user in the step 1 is judged the selection of type of strip, handle if vertical strip then carries out step 3.If then carrying out step 4, handles horizontal band;
Step 3, vertical strip is detected, obtain band template figure, generate the surveyed area of next frame afterwards and recycle this step subsequent frame is handled;
Step 4, the horizontal bar band is detected, obtain band template figure, and recycle this step subsequent frame is handled.
Detection described in the above-mentioned steps 3 to vertical strip, its concrete steps are following:
3-1, image preliminary treatment: the input original image is carried out format conversion, obtain the luminance component of conversion back image;
The detection of vertical strip in 3-2, the zone: the luminance picture that step 3-1 is obtained carries out the band detection.At first image is carried out Gauss's denoising; User in the step 1 is judged the selection of band color; For black stripe, select nine tap median filter to denoising after image carry out horizontal direction filtering, for white ribbon; Select 13 tap filters, image after the denoising is carried out horizontal direction filtering; Make the difference operation of medium filtering front and back afterwards, obtain error image, error image is adjudicated, obtain the bianry image that has comprised the band object according to a luminance threshold.Bianry image is carried out the connected region search of eight neighborhoods, obtain the connected region of eight connections, each connected region is as an object.For each object that searches, measure its object horizontal mean pixel width, length.Length and width to object are provided with threshold value respectively, remove the object that does not satisfy the band characteristic condition, obtain only to comprise the image of stripe information at last;
The reprocessing of 3-3, detection.The image that only comprises stripe information to step 3-2 obtains is analyzed, and detects whether to have the conllinear object, if do not exist, does not then handle, if exist, then judges whether two objects of merger according to certain judgment rule, fills up being interrupted between two objects.It is more complete that band is detected, and obtains the mask artwork of band at last;
On 3-4, the sequential to the tracking of banded zone.When detecting failure, explain that along with the increase that detects frame number the band surveyed area that possibly drift about out needs press the band drift characteristic and adjusts surveyed area, detection again.Afterwards the frame that detects failure is detected according to new surveyed area again,, think that present frame does not have band if re-detection does not still detect band; Abandon the detection of present frame; Output band template figure reads in the next frame view data, returns step 3 pair following frame again and detects;
Detection described in the above-mentioned steps 4 to horizontal band, its concrete steps are following:
4-1, image preliminary treatment: input picture is carried out format conversion, obtain the luminance component of conversion back image;
Horizontal band detects in 4-2, the zone: the luminance component of changing the back image is carried out horizontal band detect.Judge the band color,, select five tap, vertical median filters to carry out medium filtering, and image before and after the medium filtering is made difference operation, obtain error image if black stripe is carried out Gauss's denoising in the horizontal direction to luminance picture; If white ribbon adopts horizontal median filter to its filtering earlier, vertical median filter filtering is used in the back, and image before and after the vertical centering control value filtering is made difference operation, obtains error image; According to a luminance threshold error image is adjudicated afterwards, obtain bianry image; Bianry image is carried out the connected region search of eight neighborhoods, obtain the connected region of eight connections, each connected region is as an object.For each object that searches; Measure pixel peak excursion and horizontal length on its object vertical direction; Through the pixel peak excursion being set and the horizontal length threshold value is adjudicated, remove the object that satisfies the band characteristic, obtain only to comprise the image of stripe information at last;
The reprocessing of the detection of 4-3, horizontal band.Judge the band color again, if white ribbon then adopts morphological method again, the project organization operator carries out morphology to the image that only comprises stripe information that obtains among the step 4-2 to be handled, and makes testing result complete.If black stripe does not then deal with.Output detects template, reads in down frame image data, returns step 4 pair following frame again and detects.
The advantage that the detection method of band is compared with current other band detection method in the video image of the present invention is: this method has overcome current band and has only detected the limitation to the representative vertical band; Enlarged the scope of application, and proposed the horizontal bar band is carried out the method for fast detecting the detection method of vertical strip.This method offers user type and selects, and has improved the reliability that detects.In addition, the present invention has carried out integrated for the detection method of various bands, and the strategy that has adopted successive frame to handle makes the user in use, only needs to carry out interactive operation at first frame, and subsequent frame row fully automatically detects, and has improved detection efficiency.
Description of drawings
Fig. 1 is the flow chart of the detection method of band in the video image of the embodiment of the invention;
Fig. 2 is the surveyed area trace flow figure among Fig. 1;
Fig. 3 is the former figure that vertical white ribbon is arranged;
Fig. 4 among Fig. 3 according to the draw figure of surveyed area of white vertical band position;
Fig. 5 is the binary map that obtains for white vertical band difference;
The band object diagram of Fig. 6 for obtaining for white vertical band neighborhood search;
Fig. 7 connects the band template figure that the back obtains for the white vertical band;
Fig. 8 is the former figure that the black vertical strip is arranged;
Fig. 9 is according to the draw figure of surveyed area of black vertical strip position among Fig. 8;
Figure 10 is the binary map that obtains for black vertical strip difference;
Figure 11 is the band object diagram of obtaining for black vertical strip neighborhood search;
Figure 12 connects the band template figure that the back obtains for the black vertical strip;
Figure 13 is the former figure of the horizontal band of adularescent;
Figure 14 is according to the draw figure of surveyed area of white level band position among Figure 13;
Figure 15 is the binary map that white level band difference obtains;
Figure 16 is the band object diagram that white level band neighborhood search is obtained;
Figure 17 is the band template figure after the white level band is handled through morphology;
Figure 18 is the former figure that the black level band is arranged;
Figure 19 is according to the draw figure of surveyed area of black level band position among Figure 18;
Figure 20 is the binary map that black level band difference obtains;
Figure 21 is the band object template figure that black level band neighborhood search is obtained.
Embodiment
Below in conjunction with accompanying drawing embodiments of the invention are done further to specify.Present embodiment is to implement under the prerequisite with technical scheme of the present invention, provided detailed execution mode, but protection scope of the present invention is not limited to following embodiment.
As shown in Figure 1, the detection method of band in the above-mentioned video image, its concrete steps are following:
Step 1, select the color (white or black) and type of strip (level or vertical) of band to be repaired according to the band in the first frame video, take existing position out of according to first frame discal patch and delimit surveyed area by the user, like Fig. 4, Fig. 9, Figure 14, shown in Figure 19.Calculate the length DetW_H (on the line direction) and the width D etW_W (on the column direction) of surveyed area;
Step 2, user in the step 1 is judged the selection of type of strip, handle, handle if horizontal band then carries out step 4 to each frame circulation if vertical strip then carries out step 3 to each frame circulation;
Step 3, vertical strip is detected, obtain band template figure, generate the surveyed area of next frame afterwards and recycle this step subsequent frame is handled, its concrete steps are following:
3-1, image preliminary treatment
Read in original image, original image be designated as RGB (i, j), wherein, i, j be the row and column of presentation video respectively; To original image RGB (i j) carries out the conversion of color space, and conversion back image is designated as: Ycbcr (i, j), get conversion back image luminance component Y (i, j);
The detection of vertical strip in 3-2, the zone
(i j) handles to Y in the defined area in step 1.Carry out Gauss's denoising in vertical direction, and acquisition Y_G (i, j).User in the step 1 is judged the selection of band color,, select nine tap median filter Y_G (i for black stripe; J) carry out horizontal direction filtering,, select 13 tap filters for white ribbon; Obtain behind the medium filtering image Y_G_M (i, j).Make following difference operation afterwards:
Do difference operation for the black vertical strip:
Y_O(i，j)＝Y_G_M(i，j)-Y_G(i，j)(1)
Do difference operation for the white vertical band:
Y_O(i，j)＝Y_G(i，j)-Y_G_M(i，j)(2)
According to luminance threshold Th Y,, obtain bianry image information (, shown in Figure 10) like Fig. 5 by following rule judgement:
Threshold value Th_Y is based on that experience visually to band is provided with, and for eight luminance pictures, threshold value can be extracted the band position greater than 4 less than 10, does not introduce excessive interference again.Rule of thumb, the Th_Y value is set to 5.
(i j) carries out the connected region search of eight neighborhoods, obtains eight connected regions that are communicated with, and each connected region is as an object to bianry image Y_O_BW.For k the object Object_k that searches, measure its object horizontal mean pixel width w_k, length h_k.According to the band color, use following criterion to judge, extract black vertical strip, white vertical band:
For the black vertical strip:
For the white vertical band:
(4), the value of the th_h in (5) is provided with as follows:
Wherein W is the width of sequence frame.Through judgement, if Object_k 1 thinks that then this object is a band, if 0; Think that then this object is not a band, Y O_BW (i, j) removing judgement in the image is 0 object; Obtain only to comprise image Y_O_L BW (i, j) (, shown in Figure 11) of stripe information at last like Fig. 6;
The reprocessing that 3-3, vertical strip detect
To Y_O_L_BW (i; J) the band object Object_k1 in; (k1 ≠ k2) analyze under the nonoverlapping situation of two object row-coordinates, analyzes its slope and central series coordinate position Object_k2; Judge whether two objects of merger according to following judgment rule, fill up being interrupted between two objects.
Calculate:
Δcolumn＝|column_k1-column_k2|(7)
Δrow＝|row_k1-row_k2| (8)
Wherein, column_k1 and coulumn_k2 are the central series coordinate of two objects, and row_k1, row_k2 are the central row coordinate of two objects.
Carry out as judging:
If two conditions below satisfying:
(i)slope_k1*slope_k2＞＝0；
(ii)|Δcolumn-(Δrow*slope_k1+Δrow*slope_k2)/2|≤3，
Select merger.
Connect two bands up and down according to the mean breadth of band up and down, Y_O_L_BW (i, j) interrupted in the middle of filling up on the basis; If the length after connecting is not extended processing less than 1/3rd of picture traverse, otherwise, extend below on then carrying out according to the gradient of band, make it run through surveyed area all the time, obtain mask Y_BW (i, j) (, shown in Figure 12) of band at last like Fig. 7.
On 3-4, the sequential to the tracking in vertical strip zone
When detecting failure, the shows slice surveyed area that possibly drift about out needs the adjustment surveyed area to detect again.Through to the drift statistics of band on sequential, find that band did not wait to hundreds of frames at tens frames in the time that same area continues, do approximate cosine concussion drift, band drift maximum is no more than ten pixel columns between the consecutive frame.The row drift of non-consecutive frame generally also is no more than 20 pixel columns.For this reason, carry out the tracking of surveyed area by Fig. 2 process, follow the tracks of and obtain after the new surveyed area, output detects template, reads in down two field picture, returns step 3 again and in tracing area, following frame is detected; Concrete steps are following:
3-4-1, at first is provided with a frame counter, and its function continues to detect frame number for control, and initial value is the maximum frame number of continuous detecting, and according to band is continued the frame number statistics in same area, initial value is set to 30.Whenever handle a frame, counter subtracts 1.
3-4-2, current band testing result is judged, then changed over to step 3-4-3 and handle if detect band, otherwise; Operate as follows: at first, whether the judgment frame counter is 0, if be not 0; Keep surveyed area constant, the new surveyed area of generation is identical with the current detection zone; If frame counter is 0, then generating new surveyed area is that frame counter is set to 30 simultaneously, and new surveyed area generates by following rule:
Get the band central series position in the present frame, be designated as colcenter, the slope of band is designated as slope in the present frame, and the rightmost column coordinate (RColumn) of new surveyed area then changes into:
RColumn＝colcenter+10+|slope|*DetW_H
This coordinate figure is a boundary value with the border of image, and when a plurality of band, right margin serves as to calculate the basis with the rightmost band.The leftmost row coordinate of new surveyed area then changes (LColumn) into:
LColumn＝colcenter-(10+|slope|*DetW_H)
This coordinate figure is a boundary value with the border of image, and when a plurality of band, left margin serves as to calculate the basis with the Far Left band.The surveyed area up-and-down boundary is owing to the existence of Connection Step can remain unchanged.
3-4-3, adjustment surveyed area, left and right sides surveyed area is expanded ten pixel wide respectively, and in spreading range, carries out band again and detect; Be repeating step 3-2, the 3-3 operation is judged testing result afterwards; If do not detect band, think that then present frame does not have band, the new surveyed area of generation is the zone after the expansion; Otherwise, new surveyed area set by step among the 3-4-2 new surveyed area create-rule generate, be re-set as 30 to frame counter;
Step 4, the horizontal bar band is detected, obtain band template figure, and recycle this step subsequent frame is handled, its concrete steps are following:
4-1, image preliminary treatment
Read in original image, original image be designated as RGB (i, j), wherein, i, j be the row and column of presentation video respectively; To original image RGB (i j) carries out the conversion of color space, and conversion back image is designated as: Ycbcr (i, j), get conversion back image luminance component Y (i, j);
Horizontal band detects in 4-2, the zone
(i j) handles to Y in the defined area in step 1.Judge the band color, if black stripe, to Y (i j) carries out Gauss's denoising in the horizontal direction, obtain Y_G (i, j).Select five tap, vertical median filters, acquisition Y_G_M (i, j), make following difference operation:
Y_O(i，j)＝Y_G_M(i，j)-Y_G(i，j) (9)
If white ribbon, adopt width be 9 horizontal median filter to its filtering, obtain band brightness continuous images Y_HM (i, j); To adopt width be 21 vertical median filter to Y_HM (i j) removes band brightness, obtain Y_VM (i, j), make following difference operation:
Y_O(i，j)＝Y_HM(i，j)-Y_VG(i，j) (10)
According to luminance threshold Th_Y,, obtain bianry image (, shown in Figure 20) afterwards like Figure 15 by following rule judgement:
For eight luminance pictures, rule of thumb, threshold value is set to 3,
(i j) carries out the connected region search of eight neighborhoods, obtains eight connected regions that are communicated with, and each connected region is as an object to bianry image Y_O_BW.For k the object Object_k that searches, measure the pixel peak excursion v_div on its object vertical direction, horizontal length h_k.According to prior information, following criterion is judged, extracts horizontal band object:
For the black level band:
According to the situation that band occurs, th_h is set to 1/8th of picture traverse.
For the white level band:
Wherein, choosing th_h is 1/20th of picture traverse.
Through judgement, if Object_k 1 thinks that then this object is a band, if 0; Think that then this object is not a band, Y_O_BW (i, j) removing judgement in the image is 0 object; Obtain only to comprise image Y_O_L_BW (i, j) (, shown in Figure 21) of stripe information at last like Figure 16;
The reprocessing of the detection of 4-3, horizontal band
Judge the band color,, adopt morphological method white ribbon; (i j) expands, and obtains band mask artwork (shown in figure 17) to Y_O_L_BW to set width and be 9 horizontal structure operator; Need not handle black stripe, (i j) is last mask artwork to Y_O_L_BW.
Claims (1)
1. the detection method of band in the video image; Before it is characterized in that detecting, at first by the user according to the band characteristic that occurs in the first frame video sequence, select the color and the type of strip of band to be repaired; And delimit surveyed area according to the appearance position of vertical strip in the first frame; In surveyed area, carry out automatic extraction based on the band of spatial brightness information, and the surveying record object shapes, positional information; Then the band that extracts is analyzed, merged, generate the two-value template figure of band; According to the band testing result to present frame, in conjunction with the band drift features, predict the zone that the next frame band occurs again, generate the surveyed area of next frame, according to new surveyed area next frame is carried out band and detect, its concrete steps are following:
Step 1, select the color and the type of strip of band to be repaired according to the band in the video, take existing position out of according to first frame discal patch and delimit surveyed area, calculate the length and the width of surveyed area by the user;
Step 2, user in the step 1 is judged the selection of type of strip, handle, handle if horizontal band then carries out step 4 if vertical strip then carries out step 3;
Step 3, vertical strip is detected, obtain to detect template figure, generate the surveyed area of next frame afterwards and recycle this step subsequent frame is handled, concrete steps are following:
3-1, image preliminary treatment are carried out format conversion to the input original image, obtain the luminance component of conversion back image;
3-2, the luminance picture that step 3-1 is obtained carry out the band detection; At first image is carried out Gauss's denoising, user in the step 1 is judged the selection of band color, for black stripe; Image carries out horizontal direction filtering after selecting nine tap median filter to denoising; For white ribbon, select 13 tap filters, image after the denoising is carried out horizontal direction filtering; Make the difference operation of medium filtering front and back afterwards, obtain error image, error image is adjudicated according to a luminance threshold; Obtain the bianry image that has comprised the band object, bianry image is carried out the connected region search of eight neighborhoods, obtain the connected region of eight connections; Each connected region for each object that searches, is measured its object horizontal mean pixel width as an object; Length; Length and width to object are provided with threshold value respectively, remove the object that does not satisfy the band characteristic condition, obtain only to comprise the image of stripe information at last;
3-3, detection reprocessing: the image that only comprises stripe information to step 3-2 obtains is analyzed, and detects whether to have the conllinear object, if do not exist; Then do not handle; If exist, then, fill up being interrupted between two objects according to two objects of certain judgment rule merger; It is more complete that band is detected, and obtains the mask artwork of band at last;
On 3-4, the sequential to the tracking of banded zone: when detecting failure, explain that the band surveyed area that possibly drift about out needs press the band drift characteristic and adjusts surveyed area, detection again along with the increase that detects frame number; Afterwards the frame that detects failure is detected according to new surveyed area again,, think that present frame does not have band if re-detection does not still detect band; Abandon the detection of present frame; Output band template figure reads in the next frame view data, returns step 3 pair following frame again and detects;
Step 4, the horizontal bar band is detected, obtain band template figure, and recycle this step subsequent frame is handled, concrete steps are following:
4-1, image preliminary treatment: input picture is carried out format conversion, obtain the luminance component of conversion back image;
Horizontal band detects in 4-2, the zone: the luminance component of changing the back image is carried out horizontal band detect; Judge the band color; If black stripe is carried out Gauss's denoising in the horizontal direction to luminance picture, select five tap, vertical median filters to carry out medium filtering; And image before and after the medium filtering made difference operation, obtain error image; If white ribbon adopts horizontal median filter to its filtering earlier, vertical median filter filtering is used in the back, and image before and after the vertical centering control value filtering is made difference operation, obtains error image; According to a luminance threshold error image of white or black stripe is adjudicated afterwards, obtain bianry image; Bianry image is carried out the connected region search of eight neighborhoods, obtain the connected region of eight connections, each connected region is as an object; For each object that searches; Measure pixel peak excursion and horizontal length on its object vertical direction; Through the pixel peak excursion being set and the horizontal length threshold value is adjudicated, remove the object that satisfies the band characteristic, obtain only to comprise the image of stripe information at last;
The detection reprocessing of 4-3, horizontal band: judge the band color again, if white ribbon then adopts morphological method; The project organization operator carries out morphology to the image that only comprises stripe information that obtains among the step 4-2 to be handled, and makes testing result complete, if black stripe; Then do not deal with; Output detects template, reads in down frame image data, returns step 4 pair following frame again and detects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101965913A CN101692691B (en) | 2009-09-27 | 2009-09-27 | Method for detecting bands in video images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101965913A CN101692691B (en) | 2009-09-27 | 2009-09-27 | Method for detecting bands in video images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101692691A CN101692691A (en) | 2010-04-07 |
| CN101692691B true CN101692691B (en) | 2012-07-04 |
Family
ID=42081350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009101965913A Expired - Fee Related CN101692691B (en) | 2009-09-27 | 2009-09-27 | Method for detecting bands in video images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101692691B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102231790A (en) * | 2011-06-23 | 2011-11-02 | 上海大学 | Method for restoring scratches of video sequence |
| CN104820972B (en) * | 2015-05-07 | 2017-06-16 | 北京空间机电研究所 | A kind of infrared image ME noise remove methods based on in-orbit statistic of classification |
| US10795456B2 (en) * | 2016-03-22 | 2020-10-06 | Guangdong Virtual Reality Technology Co., Ltd. | Method, device and terminal for determining effectiveness of stripe set |
| CN108759678A (en) * | 2018-07-19 | 2018-11-06 | 广州富唯电子科技有限公司 | Automatic measuring equipment and its measurement method in heat sink sizes and flatness line |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1567372A (en) * | 2003-07-02 | 2005-01-19 | 德鑫科技股份有限公司 | Method for treating scratch of digital image |
-
2009
- 2009-09-27 CN CN2009101965913A patent/CN101692691B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1567372A (en) * | 2003-07-02 | 2005-01-19 | 德鑫科技股份有限公司 | Method for treating scratch of digital image |
Non-Patent Citations (1)
| Title |
|---|
| 韩军 等.视频图像修复算法的研究.《电视技术》.2007,第31卷(第7期),72-74,82. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101692691A (en) | 2010-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102132323B (en) | System and method for automatic image straightening | |
| US20070286499A1 (en) | Method for Classifying Digital Image Data | |
| CN105488758A (en) | Image scaling method based on content awareness | |
| CN101453575B (en) | Video subtitle information extracting method | |
| Vanhamel et al. | Multiscale gradient watersheds of color images | |
| US20080232715A1 (en) | Image processing apparatus | |
| CN105205488B (en) | Word area detection method based on Harris angle points and stroke width | |
| CN108447021B (en) | Video scaling method based on block division and frame-by-frame optimization | |
| CN102208023A (en) | Method for recognizing and designing video captions based on edge information and distribution entropy | |
| CN106548160A (en) | A kind of face smile detection method | |
| Wang et al. | A novel video caption detection approach using multi-frame integration | |
| Drew et al. | Video dissolve and wipe detection via spatio-temporal images of chromatic histogram differences | |
| CN104166983A (en) | Motion object real time extraction method of Vibe improvement algorithm based on combination of graph cut | |
| CN101692691B (en) | Method for detecting bands in video images | |
| CN101510304B (en) | Method, device and pick-up head for dividing and obtaining foreground image | |
| CN111553851A (en) | Video rain removing method based on time domain rain line decomposition and spatial structure guidance | |
| Hu et al. | Hybrid shift map for video retargeting | |
| CN107886518A (en) | Picture detection method, device, electronic equipment and read/write memory medium | |
| CN105701515A (en) | Face super-resolution processing method and system based on double-layer manifold constraint | |
| CN105335930A (en) | Edge data driven robustness-based face super-resolution processing method and system | |
| CN109448010B (en) | Automatic four-side continuous pattern generation method based on content features | |
| CN107481253B (en) | Edge-based spot detection method | |
| Kumar et al. | An efficient algorithm for text localization and extraction in complex video text images | |
| Tehsin et al. | Survey of region-based text extraction techniques for efficient indexing of image/video retrieval | |
| Gllavata et al. | Finding text in images via local thresholding |
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: 20120704 Termination date: 20150927 |
|
| EXPY | Termination of patent right or utility model |