CN102831611B - Extracting method and the device of angle advertisement is hung in a kind of TV programme - Google Patents

Abstract

The embodiment of the invention discloses the extracting method and device of hanging angle advertisement in a kind of TV programme, belong to image processing field.Method comprises: analyze each key frame detected in video, according to the stabilized zone in key-frame extraction video; Described stabilized zone is followed the tracks of, detects the situation of change in the sequential of described stabilized zone; Extract from described stabilized zone according to described situation of change and hang angle advertisement.The present invention by analyzing key frame respectively respectively, following the tracks of in the scene belonging to each camera lens, not only be not subject to the impact of picture noise, and fully adapted to the diversity feature of hanging the position of angle advertisement in television image, shape, color, brightness etc., thus the extension angle advertisement in television image is effectively extracted, and then angle advertising area filler pixels will be hung, make adless video resource.

Description

Extracting method and the device of angle advertisement is hung in a kind of TV programme

Technical field

The present invention relates to image processing field, particularly relate to the extracting method and device of hanging angle advertisement in a kind of TV programme.

Background technology

In recent years, TV spreads all over huge numbers of families, sees that TV becomes the daily pastime of people, the requisite mode of amusement.Along with developing rapidly of tv broadcasting technology, there is explosive growth in the quantity of television channel and corresponding TV programme.While broadcasting TV programme, television image periphery break for commercialsy carry out the method that publicizes also adopt by a lot of television channel.

In the commercial detection method of existing television image, because extension angle advertisement and station symbol are all picture element, therefore normal employing extracts extension angle advertisement to platform object detection method.Prior art adopts the recognition methods based on global registration to identify station symbol usually, although the recognition effect of this recognition methods to most of station symbol is all right, but the information such as the position of angle advertisement in television image, shape, color, brightness of hanging are different, therefore may not be suitable for extension angle advertisement.

Concrete, now conventional station caption detection method has frame difference method and model statistics method.

Frame difference method carries out purposes of commercial detection by analysis chart picture frame, because the color of adjacent two frames in station symbol region is relatively constant, the color of background area changes greatly relatively, therefore less in station symbol region difference, non-station symbol region difference is large, utilizes threshold value station symbol and non-station symbol region segmentation can be come.But the method is better to opaque station symbol Detection results, and poor to the station symbol Detection results hanging the types such as the translucent of angle advertisement and so on or animation, and easily by the impact of picture noise.

Model statistics method is also be based upon on the basis of frame difference method, due to station symbol region, this is more fixing, non-station symbol regional change is comparatively large, generally adds up the statistical model of each pixel according to time variations by frame-to-frame differences, and judges whether new pixel belongs to station symbol according to statistical model.But the pattern hanging angle advertisement is various, cannot determine do not have unified template to new samples.

Therefore, it is comparatively large that difficulty is extracted in the extension angle advertisement in television image, video carry out secondary utilize time, often effectively just cannot hang angle advertising area filler pixels, form adless video resource.

Summary of the invention

The embodiment provides the extracting method and device of hanging angle advertisement in a kind of TV programme, solving traditional extracting mode to hanging the poor problem of angle advertising results.

For achieving the above object, adopt following technical scheme:

Hang an extracting method for angle advertisement in video, comprising:

Analyze each key frame detected in video, according to the stabilized zone in key-frame extraction video;

Described stabilized zone is followed the tracks of, detects the situation of change in the sequential of described stabilized zone;

Extract from described stabilized zone according to described situation of change and hang angle advertisement.

Preferably, when described analysis detects each key frame in video,

Analyzing video based on video content, is each camera lens by Video segmentation;

Respectively each picture frame in each camera lens is extracted, obtain corresponding each key frame.

It is preferably, described when video being analyzed based on video content,

Extract the frame-to-frame differences between each picture frame in video, thus the transform boundary in extracting in video;

According to described transform boundary, video is split, obtain each camera lens.

It is preferably, described when each picture frame in each camera lens is extracted,

Read in each picture frame in camera lens by sequence, each picture frame is converted to HSV space respectively by RGB color space;

Calculate the similarity in adjacent two picture frame hsv color spaces respectively;

When the similarity in hsv color space is greater than similar threshold value, using the picture frame of correspondence as key frame corresponding to described camera lens.

Preferably, during described stabilized zone according in key-frame extraction video,

Each key frame is separately converted to luminance picture, adds up each pixel maximum brightness value and minimum luminance value in sequential in each luminance picture, the region of statistics brightness vary stable is as the stabilized zone in video.

Preferably, described each key frame is separately converted to luminance picture before, remove the black surround at each key frame edge according to given threshold.

Preferably, during situation of change in the sequential of described detection described stabilized zone, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone.

The invention also discloses the extraction element hanging angle advertisement in a kind of video, comprising:

Analytic unit, for analyzing each key frame detected in video, according to the stabilized zone in key-frame extraction video;

Tracking cell, for following the tracks of described stabilized zone, detects the situation of change in the sequential of described stabilized zone;

Extraction unit, hangs angle advertisement for extracting from described stabilized zone according to described situation of change.

Preferably, when described analytic unit analysis detects each key frame in video,

Analyzing video based on video content, is each camera lens by Video segmentation;

Respectively each picture frame in each camera lens is extracted, obtain corresponding each key frame.

Preferably, when described tracking cell detects the situation of change in the sequential of described stabilized zone, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone.

Hanging extracting method and the device of angle advertisement in a kind of TV programme that the embodiment of the present invention provides, by extracting each key frame respectively according to each camera lens in video, and the stabilized zone in key frame being followed the tracks of, thus extract extension angle advertisement.The present invention analyzes key frame respectively respectively, follows the tracks of in the scene belonging to each camera lens, not only be not subject to the impact of picture noise, and fully adapted to the diversity feature of hanging the position of angle advertisement in television image, shape, color, brightness etc., thus the extension angle advertisement in television image is effectively extracted, and then angle advertising area filler pixels will be hung, make adless video resource.

Accompanying drawing explanation

The process flow diagram of the extracting method of angle advertisement is hung in a kind of TV programme that Fig. 1 provides for the embodiment of the present invention one;

Process flow diagram when analyzing each key frame detected in video is hung in the extracting method of angle advertisement in a kind of TV programme that Fig. 2 provides for the embodiment of the present invention one;

Process flow diagram when video being analyzed in the extracting method of angle advertisement is hung in a kind of TV programme that Fig. 3 provides for the embodiment of the present invention one;

Process flow diagram when each picture frame in each camera lens being extracted in the extracting method of angle advertisement is hung in a kind of TV programme that Fig. 4 provides for the embodiment of the present invention one;

The function structure chart of the extraction element of angle advertisement is hung in a kind of TV programme that Fig. 5 provides for the embodiment of the present invention two.

Embodiment

Below in conjunction with accompanying drawing, the extracting method and device of hanging angle advertisement in a kind of TV programme of the embodiment of the present invention are described in detail.

In the past twenty years in the time, television advertising development is very fast, and occurred advertisement miscellaneous, some advertisements are loved by all, but some advertisements cause and denouncing, and wherein hang angle advertisement and just belong to the class caused and denouncing.

This aspect mainly solves in television image the problem of hanging angle advertisement and extracting.Hang angle advertiser will exist in the TV programme of China's Mainland, relatively less discovery in external TV programme.Filler pixels after station symbol or Iogo often extract by existing television image, makes automatically to eliminate station symbol in television image.Will be positioned at the corner location of television image owing to hanging angle advertiser, its position size is relatively fixing, and has certain duration, and extension angle advertisement is generally bright in luster, and major part is opaque Iogo.The present invention is about to hang angle advertisement as a kind of special Iogo, realizes the extraction of hanging angle advertisement.

The invention discloses the extracting method hanging angle advertisement in a kind of video, as shown in Figure 1, comprising:

Step 10, each key frame analyzed in detection video, according to the stabilized zone in key-frame extraction video;

In video analysis field, camera lens is the elementary cell of composition video and the basis of carrying out video sequence analysis.The quality of shot segmentation directly has influence on the extraction result of final key frame.Such as, video is usually across some camera lenses, and in same arrangement of mirrors head, picture material change is very little.Therefore by shot segmentation, the lasting camera lens number of one group of advertisement and the information of key frame can be provided.

Preferably, when described analysis detects each key frame in video, as shown in Figure 2, comprise the steps:

Step 101, analyzing video based on video content, is each camera lens by Video segmentation;

It is preferably, as shown in Figure 3, described when video being analyzed based on video content,

Frame-to-frame differences in step 101a, extraction video between each picture frame, thus the transform boundary in extracting in video;

In the present embodiment, moving window is adopted to carry out image retrieval to the consecutive frame in video, according to the frame-to-frame differences between consecutive frame image change and otherness statistics consecutive frame.Color Local Accumulation Histogram corresponding for consecutive frame is combined with the texture of picture frame, scans in consecutive frame respectively, thus count the frame-to-frame differences of consecutive frame.

From the frame-to-frame differences between each picture frame by Grads threshold count frame-to-frame differences larger time corresponding picture frame, then the transform boundary in the corresponding video of described picture frame.

Step 101b, according to described transform boundary, video to be split, obtain each camera lens.

Split by video by transform boundary, video content corresponding to adjacent two transform boundaries is camera lens.

Step 102, respectively each picture frame in each camera lens to be extracted, obtain corresponding each key frame.

It is preferably, as shown in Figure 4, described when each picture frame in each camera lens is extracted,

Step 102a, each picture frame read in by sequence in camera lens, be converted to HSV space by each picture frame respectively by RGB color space;

Step 102b, calculate the similarity in adjacent two picture frame hsv color spaces respectively;

Step 102c, when the similarity in hsv color space is greater than similar threshold value, using the picture frame of correspondence as key frame corresponding to camera lens.

In the present embodiment, when each picture frame in each camera lens is extracted, a two field picture is extracted as key frame to each camera lens, for the initialization condition that stabilized zone judges.Adopt and carry out shot segmentation based on the method for color histogram, namely from the 1st frame key frame, adopt cyclic process to complete following steps:

1. read in the i-th two field picture frame p _k, by the i-th two field picture frame p _kbe hsv color space by RGB color space conversion;

2., in hsv color space, according to H in color space HSV, channel S calculates HS spatial color histogram respectively

3. calculate current image frame histogram with a upper picture frame histogram similarity Sim (k, k-1):

$\mathrm{Sim}(k,k-1)=\underset{\mathrm{hs}}{\mathrm{\Σ}}\left|\right|{\mathrm{hist}}_{\mathrm{hs}}^k-{\mathrm{hist}}_{\mathrm{hs}}^{k-1}\left|\right|$

4. judge whether similarity Sim (k, k-1) is greater than similar threshold value, when being greater than given threshold, by current image frame p _kas the key frame that camera lens is corresponding, and bring next step calculating into.

Before the stabilized zone in key-frame extraction video, need video, particularly television image carries out pre-service.In practical situations both, there are many videos, particularly television image always often to display by screens some schedule informations etc. in lower edge, cause the lower edge of television image often to there is black surround.Therefore, the method for threshold value is adopted to remove Video Edge black surround.

In general, the extension angle advertisement in video, is usually located at 4 angles of video, and for opaque shape or be similar to certain symbol sticking on the picture of video, its size position is fixed.For this reason, the present invention uses and detects extension angle advertisement based on detecting the method stablizing zone of opacity in video.Hang that angle advertisement is often stable to be present in the key frame of a continuous videos.When video generation consecutive variations, hang angle advertising area and be unanimously in steady state (SS).Therefore utilize the stability of hanging angle advertisement as the characteristic detected.

In the present embodiment, after removing the black surround at each key frame edge, during according to stabilized zone in key-frame extraction video, each key frame is separately converted to luminance picture, add up each pixel maximum brightness value and minimum luminance value in sequential in each luminance picture, the region of statistics brightness vary stable is as the stabilized zone in video.

In the present embodiment, adopt (MLVR) method (minimallumainacevaricanceregion) graphical analysis, the maximum brightness value of each pixel in sequential and minimum luminance value in record luminance picture, when brightness keeps the most stable region to be the candidate region hanging angle advertisement, after trying to achieve each candidate region on time orientation, filter out stabilized zone with based on threshold value.Concrete steps are as follows:

1. read in current frame of video P _i, first judge present frame P _iwhether be key frame, if so, proceed to 2, no other frame of video continuing to read in video;

2. couple key frame P _i, eliminate black surround;

3. obtain the very big brightness frame value maxp of each pixel in each key frame _i=max (p _i, maxp _i-1), minimum brightness frame value minp _i=min (p _i, minp _i-1), and then obtain candidate region SR _i=|| maxp _i-minp _i||≤thresh, calculates by pixel;

4. screen based on threshold value, judge candidate region SR _iwhether be stabilized zone, be, enter the tracking that next step carries out stabilized zone, otherwise proceed to 1;

For the candidate region SR judging current key frame _iwhen whether being stabilized zone, the candidate region SR of Main Basis and previous keyframe _i-1difference, if change excessive, can instability be thought, this candidate region SR _iit not stabilized zone.

Step 11, described stabilized zone to be followed the tracks of, detect the situation of change in the sequential of described stabilized zone;

In video display process, the initial sum final position hanging angle advertisement is unexpected.Need to judge whether the stabilized zone of previous keyframe and the stabilized zone of current key frame are same extension angle advertisement.Therefore when carrying out monitoring of the advertisement, the stabilized zone to detecting is needed to follow the tracks of, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone, finally to obtain hanging angle advertisement.

In the video playback of reality, there is following several situation:

1. hang angle advertisement from without hanging angle advertisement to appearance;

2. hang angle advertisement from appearance to disappear to extension angle advertisement;

3. hang angle advertisement from one and hang angle advertisement to another.

Therefore, for these three kinds of situations, need in video analysis, judge the situation of change that current described stabilized zone belongs to, and then carry out following the tracks of.

For situation 1: carry out in the process of video sequence analysis to extension angle advertisement, video key frame when there is not extension angle advertisement is changeable, and the brightness maxima in stabilized zone and brightness minimum value are also along with image texture ceaselessly converts.In video sequence is play chronologically, brightness maxima maxp _ican become increasing, brightness minimum value minp _ibecome and more come less, cause SR _imiddle stabilized zone will more be come less.Therefore, by setting pixel threshold T _sr, when the number of pixels in stabilized zone is less than T _srtime, by brightness maxima maxp _iwith brightness minimum value minp _iupgrade, then readjust the size of stabilized zone.

For situation 2: video is by occurring hanging angle advertisement to when hanging angle advertisement disappearance, stabilized zone SR _ifront and back change obvious, in like manner can according to stabilized zone SR _isituation of change judge.

For situation 3: this kind of situation is very similar with situation 2, hang angle advertisement to when disappearing or occur that another hangs angle advertisement in existence, image texture can occur comparatively significantly to change, and in like manner application 2 can process.

Step 12, according to described situation of change extract from described stabilized zone hang angle advertisement.

Hang angle advertisement general persistence longer, the band of position is generally positioned at video 4 angles, extracts and hangs angle advertisement, thus the extension angle advertisement in television image effectively extracted according to the situation of change of stabilized zone, and then angle advertising area filler pixels will be hung, make adless video resource.

In other embodiments of the present invention, also disclose the extraction element hanging angle advertisement in a kind of video, as shown in Figure 5, comprising:

Analytic unit 501, for analyzing each key frame detected in video, according to the stabilized zone in key-frame extraction video;

Tracking cell 502, for following the tracks of described stabilized zone, detects the situation of change in the sequential of described stabilized zone;

Extraction unit 503, hangs angle advertisement for extracting from described stabilized zone according to described situation of change.

Preferably, when described analytic unit analysis detects each key frame in video,

Analyzing video based on video content, is each camera lens by Video segmentation;

Respectively each picture frame in each camera lens is extracted, obtain corresponding each key frame.

Preferably, when described tracking cell detects the situation of change in the sequential of described stabilized zone, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone.

Hanging extracting method and the device of angle advertisement in a kind of TV programme that the embodiment of the present invention provides, by extracting each key frame respectively according to each camera lens in video, and the stabilized zone in key frame being followed the tracks of, thus extract extension angle advertisement.The present invention analyzes key frame respectively respectively, follows the tracks of in the scene belonging to each camera lens, not only be not subject to the impact of picture noise, and fully adapted to the diversity feature of hanging the position of angle advertisement in television image, shape, color, brightness etc., thus the extension angle advertisement in television image is effectively extracted, and then angle advertising area filler pixels will be hung, make adless video resource.

The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited thereto, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; change can be expected easily or replace, all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should described be as the criterion with the protection domain of claim.

Claims (7)

1. hang an extracting method for angle advertisement in video, it is characterized in that, comprising:

Analyze each key frame detected in video, according to the stabilized zone in key-frame extraction video, during described stabilized zone according in key-frame extraction video, each key frame is separately converted to luminance picture, add up each pixel maximum brightness value and minimum luminance value in sequential in each luminance picture, the region of statistics brightness vary stable is as the stabilized zone in video;

Described stabilized zone is followed the tracks of, detect the situation of change in the sequential of described stabilized zone, during situation of change in the sequential of described detection described stabilized zone, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone, namely by setting pixel threshold T _sr, when the number of pixels in stabilized zone is less than T _srtime, by brightness maxima maxp _iwith brightness minimum value minp _iupgrade, then readjust the size of stabilized zone;

Extract from described stabilized zone according to described situation of change and hang angle advertisement.

2. method according to claim 1, is characterized in that: when described analysis detects each key frame in video, analyzing video based on video content, is each camera lens by Video segmentation; Respectively each picture frame in each camera lens is extracted, obtain corresponding each key frame.

3. method according to claim 2, is characterized in that: described when analyzing video based on video content,

Extract the frame-to-frame differences between each picture frame in video, thus extract the transform boundary in video;

According to described transform boundary, video is split, obtain each camera lens.

4. method according to claim 2, is characterized in that: described when extracting each picture frame in each camera lens,

Read in each picture frame in camera lens by sequence, each picture frame is converted to HSV space respectively by RGB color space;

Calculate the similarity in adjacent two picture frame hsv color spaces respectively;

When the similarity in hsv color space is greater than similar threshold value, using the picture frame of correspondence as key frame corresponding to described camera lens.

5. method according to claim 1, is characterized in that: described each key frame is separately converted to luminance picture before, remove the black surround at each key frame edge according to given threshold.

6. hang an extraction element for angle advertisement in video, it is characterized in that, comprising:

Analytic unit, for analyzing each key frame detected in video, according to the stabilized zone in key-frame extraction video,

During described stabilized zone according in key-frame extraction video, each key frame is separately converted to luminance picture, add up each pixel maximum brightness value and minimum luminance value in sequential in each luminance picture, the region of statistics brightness vary stable is as the stabilized zone in video;

Tracking cell, for following the tracks of described stabilized zone, detects the situation of change in the sequential of described stabilized zone,

During situation of change in the sequential of described detection described stabilized zone, according to the size of the change threshold adjustment stabilized zone in brightness value sequential in stabilized zone, namely by setting pixel threshold T _sr, when the number of pixels in stabilized zone is less than T _srtime, by brightness maxima maxp _iwith brightness minimum value minp _iupgrade, then readjust the size of stabilized zone;

Extraction unit, hangs angle advertisement for extracting from described stabilized zone according to described situation of change.

7. device according to claim 6, is characterized in that: when described analytic unit analysis detects each key frame in video,

Analyzing video based on video content, is each camera lens by Video segmentation;

Respectively each picture frame in each camera lens is extracted, obtain corresponding each key frame.