CN105049824A - Method for automatically detecting three-dimensional video format - Google Patents

Abstract

The invention discloses a method for automatically detecting three-dimensional video format. The method includes conducting necessary calculation /detection for each new program for the fact that a same video program has only one format. In specific detection method, whether the image information is abundant is determined by characteristic histogram variance, and similarity between image areas is calculated with the distance of the characteristic histogram to determine the possible video format of the video. Multi-frame voting determination is carried on the time domain again, the format of the current video program is finally output, and the display apparatus is controlled to play the video program according to the detected video format. High detection accuracy is ensured, and the calculating complexity for detecting the video format is substantially reduced.

Description

The method of automatic detection 3 D video form

Technical field

The invention belongs to 3D video detection technology field, relate to a kind of 3 D video format detection method, particularly relate to a kind of method of automatic detection 3 D video form.

Background technology

Along with the film of three-dimensional (3D) video the starting broadcasting of more and more universal and 3D digital television program in video website, 3D number of videos will become explosive growth.In these 3D videos, main form has left and right 3D form and upper and lower 3D form.Meanwhile, long-term (and always) exists by 2D video content.How automatically detecting video frequency program form is 2D or 3D, if 3D video format, belong to again 3D form which kind of, be indispensable for correct display.

Simultaneously according to the convention of existing video program production, the form of each program is determined and unique.Be 2D video, or be that 3D video adopts left-right format, or be that 3D video adopts top-down format.Do not exist inside a program and comprise various video form.

Publication number be the first computed image entropy of Chinese patent of CN101980545A to carry out the Effective judgement of picture frame, the linearly dependent coefficient then calculated between regional also compares with threshold value, and multiframe differentiates carrys out video format.It is high to there is computation complexity in it, and is directly proportional to video resolution, and this also means that the raising along with video resolution improves by amount of calculation.

Publication number is multiple features in the Chinese patent computed image region of CN104185023A, comprise block gradient modular character, block histogram feature, frame histogram feature, projection properties, mid line boundary characteristic, the form being carried out spatial domain by Fusion Features is just differentiated, multiframe differentiation is carried out in binding time territory simultaneously, and then formation finally differentiates result.This method is multiple feature because needs calculate, and equally also there is the king-sized problem of amount of calculation.

In addition, above two patents all do not utilize each video frequency program form to be this unique condition, and then need every frame all carry out calculating/detect/judge, be therefore only applicable to non-moving type display device (such as 3D TV) and adopt hardware asics to realize.

In view of this, nowadays in the urgent need to designing a kind of 3D video detection mode, to overcome the above-mentioned defect of existing detection mode.

Summary of the invention

Technical problem to be solved by this invention is: a kind of method providing automatic detection 3 D video form, can automatically detect 3 D video form quickly and accurately.

For solving the problems of the technologies described above, the present invention adopts following technical scheme:

A method for automatic detection 3 D video form, described method comprises the steps:

Step S1. new video program starts, to the first counter, the second counter O reset;

Step S2. obtains a two field picture, carries out view division to input picture; Left subregion, right subregion is divided into by left and right 3D format; Sub-zones, lower subregion is divided into by upper and lower 3D format;

Each regional luminance histogram of step S3. statistical picture the variance of whole frame brightness histogram Yhist and histogram Yhist; Wherein Yhist calculates according to one of following two equatioies;

$Yhist={\mathrm{Yhist}}_{lr}^L+{\mathrm{Yhist}}_{lr}^R;$

$Yhist={\mathrm{Yhist}}_{tb}^T+{\mathrm{Yhist}}_{tb}^B;$

Step S4. judges whether the variance of histogram Yhist is greater than the first predetermined threshold value, if then think valid frame, performs step S5, otherwise performs step S2;

Step S5., according to following formula, calculates the normalized cumulant between two histograms, obtains D _lrand D _tb;

$D=F({\mathrm{Yhist}}_a,{\mathrm{Yhist}}_b)=\underset{g=0}{\overset{M}{\Σ}}|h_a\left(g\right)-h_b\left(g\right)|/(2*N);$

Wherein, N is pixel sum in image, h _a(g) and h _bg () is respectively Yhist _aand Yhist _bmiddle brightness value is the number of pixels of g; Wherein D ∈ [0,1], the value of D is larger, shows that two histogram difference are larger;

Step S6. gets D _lrand D _tbin minimum value D _min, judge whether to be less than the second predetermined threshold value.If be less than the second predetermined threshold value and D _min=D _lr, then the first counter adds 1; If be less than the second predetermined threshold value and D _min=D _tb, then the second counter adds 1;

Step S7. repeats step S2 to step S6, until step S6 discriminating step has been performed m time;

If the value of step S8. first counter is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is left and right 3D video format;

If the value of the second counter is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is upper and lower 3D video format;

If the value of above two counters does not satisfy condition, then the current video program detected is 2D video format.

A method for automatic detection 3 D video form, described method comprises:

First utilize feature histogram variance to judge whether this image information is enriched, then the similitude adopting the distance of feature histogram to come between computed image region, to determine which kind of video format this image may be more; In time-domain, again carry out multiframe ballot differentiate, finally export the form of current video program, control display unit and play according to the video format detected.

As a preferred embodiment of the present invention, described method comprises the steps:

Step S1. is to each counter O reset;

Step S2. obtains a two field picture, carries out view division to input picture; Respectively input picture is divided into the first subregion, the second subregion by the different video format of setting; A corresponding counter of video format;

The feature histogram of all subregion of step S3. statistical picture, whole frame feature histogram, and the variance of whole frame feature histogram;

Step S4. judges the first subregion under same video format, whether the variance of whole frame feature histogram that the second subregion is corresponding is greater than the first predetermined threshold value, if then think valid frame, performs step S5, otherwise performs step S2;

Step S5. calculates the distance between the feature histogram of two sub regions corresponding to each video format;

Step S6. gets the minimum value of each video format respective distances in step S5, judges whether to be less than the second predetermined threshold value; If be less than the second predetermined threshold value, and distance minimum value is the feature histogram distance of certain form correspondence first subregion, the second subregion, then the counter that this form is corresponding adds 1;

Step S7. repeats step S2 to step S6, until step S6 discriminating step has been performed m time of setting;

If the value of the corresponding counter of certain video format of step S8. is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is this 3D video format;

If the value of all counters is all not more than k*m, and k ∈ (0.5,1], then the current video program detected is 2D video format.

As a preferred embodiment of the present invention, in step S2, be divided into left subregion, right subregion by left and right 3D format; Sub-zones, lower subregion is divided into by upper and lower 3D format;

In step S5, the distance between the feature histogram calculating each sub regions, comprises the feature histogram distance between left subregion, right subregion, the feature histogram distance between sub-zones, lower subregion.

At least one that as a preferred embodiment of the present invention, in described step S2, video format comprises left and right 3D form, 3D form, line interlacing 3D form, the staggered 3D form of row, grid interlock in 3D form, grid 3D form arranged side by side.

As a preferred embodiment of the present invention, the feature histogram that described method adopts is brightness histogram.

As a preferred embodiment of the present invention, described method is simultaneously in conjunction with chroma histogram, histogram of gradients, by the Histogram distance formula in method, similitude in computed image between zones of different, and multiple similarity measurement information is merged, obtain final similarity indices thus complete single frames form judging.

Beneficial effect of the present invention is: the method for the automatic detection 3 D video form that the present invention proposes, and can automatically detect 3 D video form quickly and accurately.

First, present invention utilizes same video frequency program form this true condition unique, just do necessary calculating/detection when each new program.Secondly, in concrete detection method, first utilize feature histogram variance to judge whether this image information is enriched, then the similitude adopting the distance of feature histogram to come between computed image region, to determine which kind of video format this image may be more; In time-domain, again carry out multiframe ballot differentiate, finally export the form of current video program, control display unit and play according to the video format detected.Use the method under the prerequisite ensureing high Detection accuracy, greatly reduce the computation complexity that video format detects.

Owing to make use of this unique true condition of same video frequency program form, calculating/the detection of necessity is only occurred over just the beginning of each new program, simultaneously in feature selecting design simple/fully, make automatic detection video frequency program format approach of the present invention compared to existing technology, be specially adapted to that mobile device (panel computer, notebook computer) is this kind of requires very harsh system to power consumption, manufacturing cost.

Accompanying drawing explanation

Fig. 1 is the flow chart that the present invention detects 3 D video format approach automatically.

Fig. 2 is the schematic diagram formaing molecular domains by left and right 3D.

Fig. 3 is the schematic diagram formaing molecular domains by upper and lower 3D.

Embodiment

The preferred embodiments of the present invention are described in detail below in conjunction with accompanying drawing.

Embodiment one

No matter which kind of 3D video format (up and down, left and right), main contents in the view of left and right are from Same Scene, namely two have the sampling of video camera to Same Scene of horizontal parallax, are therefore closely similar in terms of content, and this and 2D video have fundamental difference.For the feature of 3D video image, the present invention proposes a kind of method automatically detecting 3 D video form quickly and accurately, namely judges according to the similarity of the feature histogram of left and right view.

Refer to Fig. 1, present invention is disclosed a kind of method of automatic detection 3 D video form, described method comprises the steps:

S1. new video program starts, to the first counter, the second counter O reset.

S2. obtain a two field picture, view division is carried out to input picture.Be formatted as Fig. 2 by left and right 3D, wherein left view 11 on the left side, right view 12 on the right; Be formatted as Fig. 3 by upper and lower 3D, wherein left view 21 on top, and right view 22 is below.

S3. each regional luminance histogram of statistical picture the variance of whole frame brightness histogram Yhist and histogram Yhist.Attention: wherein Yhist can calculate according to one of following two equatioies.

$Yhist={\mathrm{Yhist}}_{lr}^L+{\mathrm{Yhist}}_{lr}^R$

$Yhist={\mathrm{Yhist}}_{tb}^T+{\mathrm{Yhist}}_{tb}^B$

S4. judge whether the variance of histogram Yhist is greater than the first predetermined threshold value, if then think valid frame, perform step S5, otherwise perform step S2;

S5. according to following formula, calculate the normalized cumulant between two histograms, obtain D _lrand D _tb

$D=F({\mathrm{Yhist}}_a,{\mathrm{Yhist}}_b)=\underset{g=0}{\overset{M}{\Σ}}|h_a\left(g\right)-h_b\left(g\right)|/(2*N)$

Wherein N is pixel sum in image, h _a(g) and h _bg () is respectively Yhist _aand Yhist _bmiddle brightness value is the number of pixels of g.Wherein D ∈ [0,1], the value of D is larger, shows that two histogram difference are larger.

S6. D is got _lrand D _tbin minimum value D _min, judge whether to be less than the second predetermined threshold value.If be less than the second predetermined threshold value and D _min=D _lr, then the first counter adds 1; If be less than the second predetermined threshold value and D _min=D _tb, then the second counter adds 1.

S7. S2 ~ ~ S6 is repeated, until S6 discriminating step has been performed m time.

If S8. the value of the first counter is more than or equal to k*m, (k ∈ (0.5,1]), then the current video program form detected is left and right 3D video format.

If S9. the value of the second counter is more than or equal to k*m, (k ∈ (0.5,1]), then the current video program form detected is upper and lower 3D video format.

If S10. the value of above two counters does not satisfy condition, then the current video program detected is 2D video format.

The present invention, according to same video frequency program form this true condition unique, only does necessary calculating/detection when each new program.Adopt the distance of feature histogram as the standard passing judgment on similitude between image-region, while ensureing high Detection accuracy, the resolution of computation complexity and image is had nothing to do, greatly reduce the amount of calculation that video format detects.Utilize same video frequency program form this true condition unique equally, time-domain adopts the voting mechanism of multiframe complete Final Format and differentiate, fully ensure the high-accuracy detected.

Embodiment two

Refer to Fig. 1, the present invention discloses a kind of method of automatic detection 3 D video form, and described method comprises the steps:

Step S1. is to each counter O reset;

Step S2. obtains a two field picture, carries out view division to input picture; Respectively input picture is divided into the first subregion, the second subregion by the different video format of setting; A corresponding counter of video format;

The feature histogram of all subregion of step S3. statistical picture, whole frame feature histogram, and the variance of whole frame feature histogram;

Step S4. judges the first subregion under same video format, whether the variance of whole frame feature histogram that the second subregion is corresponding is greater than the first predetermined threshold value, if then think valid frame, performs step S5, otherwise performs step S2;

Step S5. calculates the distance between the feature histogram of two sub regions corresponding to each video format;

Step S6. gets the minimum value of each video format respective distances in step S5, judges whether to be less than the second predetermined threshold value; If be less than the second predetermined threshold value, and distance minimum value is the feature histogram distance of certain form correspondence first subregion, the second subregion, then the counter that this form is corresponding adds 1;

Step S7. repeats step S2 to step S6, until step S6 discriminating step has been performed m time of setting;

If the value of the corresponding counter of certain video format of step S8. is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is this 3D video format; If the value of all counters is all not more than k*m, and k ∈ (0.5,1], then the current video program detected is 2D video format.

Embodiment three

The detection of left and right 3D form, up and down the 3D video format that 3D form these two kinds is common is only discuss in embodiment one.But method is not confined to this two kinds of 3D video formats, only need to adopt different sub-zone dividing methods in S2 and S3 step, method of the present invention can be directly applied for line interlacing 3D form, the staggered 3D form of row, grid-staggered 3D form, grid-3D form (publication number is the Chinese patent of CN102742283A) side by side.

Particularly, such as: line interlacing 3D form, only need odd-numbered line/even number line to be divided into two groups, as two regions statistic histogram respectively.In like manner row are staggered is that odd column/even column is divided into two groups, and all the other are also similar.

Embodiment four

The difference of the present embodiment and embodiment one is, in the present embodiment, the feature histogram of employing is brightness histogram.Method is not confined to this, also can simultaneously in conjunction with other information such as chroma histogram, histogram of gradients, by the Histogram distance formula in method, similitude in computed image between zones of different, and multiple similarity measurement information is merged, obtain final similarity indices thus complete single frames form judging.

In sum, the method for the automatic detection 3 D video form that the present invention proposes, can detect 3 D video form quickly and accurately automatically.

First, present invention utilizes same video frequency program form this true condition unique, just do necessary calculating/detection when each new program.Secondly, in concrete detection method, first utilize feature histogram variance to judge whether this image information is enriched, then the similitude adopting the distance of feature histogram to come between computed image region, to determine which kind of video format this image may be more; In time-domain, again carry out multiframe ballot differentiate, finally export the form of current video program, control display unit and play according to the video format detected.Use the method under the prerequisite ensureing high Detection accuracy, greatly reduce the computation complexity that video format detects.

Owing to make use of this unique true condition of same video frequency program form, calculating/the detection of necessity is only occurred over just the beginning of each new program, simultaneously in feature selecting design simple/fully, make automatic detection video frequency program format approach of the present invention compared to existing technology, be specially adapted to that mobile device (panel computer, notebook computer) is this kind of requires very harsh system to power consumption, manufacturing cost.

Here description of the invention and application is illustrative, not wants by scope restriction of the present invention in the above-described embodiments.Distortion and the change of embodiment disclosed are here possible, are known for the replacement of embodiment those those of ordinary skill in the art and the various parts of equivalence.Those skilled in the art are noted that when not departing from spirit of the present invention or substantive characteristics, the present invention can in other forms, structure, layout, ratio, and to realize with other assembly, material and parts.When not departing from the scope of the invention and spirit, can other distortion be carried out here to disclosed embodiment and change.

Claims (7)

1. automatically detect a method for 3 D video form, it is characterized in that, described method comprises the steps:

Step S1. new video program starts, to the first counter, the second counter O reset;

Step S2. obtains a two field picture, carries out view division to input picture; Left subregion, right subregion is divided into by left and right 3D format; Sub-zones, lower subregion is divided into by upper and lower 3D format;

Each regional luminance histogram of step S3. statistical picture the variance of whole frame brightness histogram Yhist and histogram Yhist; Wherein Yhist calculates according to one of following two equatioies;

$Yhist={\mathrm{Yhist}}_{lr}^L+{\mathrm{Yhist}}_{lr}^R;$

$Yhist={\mathrm{Yhist}}_{tb}^T+{\mathrm{Yhist}}_{tb}^B;$

Step S4. judges whether the variance of histogram Yhist is greater than the first predetermined threshold value, if then think valid frame, performs step S5, otherwise performs step S2;

Step S5., according to following formula, calculates the normalized cumulant between two histograms, obtains D _lrand D _tb;

$D=F({\mathrm{Yhist}}_a,{\mathrm{Yhist}}_b)=\underset{g=0}{\overset{M}{\Σ}}|h_a\left(g\right)-h_b\left(g\right)|/(2*N);$

Wherein, N is pixel sum in image, h _a(g) and h _bg () is respectively Yhist _aand Yhist _bmiddle brightness value is the number of pixels of g; Wherein D ∈ [0,1], the value of D is larger, shows that two histogram difference are larger;

Step S6. gets D _lrand D _tbin minimum value D _min, judge whether to be less than the second predetermined threshold value.If be less than the second predetermined threshold value and D _min=D _lr, then the first counter adds 1; If be less than the second predetermined threshold value and D _min=D _tb, then the second counter adds 1;

Step S7. repeats step S2 to step S6, until step S6 discriminating step has been performed m time;

If the value of step S8. first counter is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is left and right 3D video format;

If the value of the second counter is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is upper and lower 3D video format;

If the value of above two counters does not satisfy condition, then the current video program detected is 2D video format.

2. automatically detect a method for 3 D video form, it is characterized in that, described method comprises:

Utilize feature histogram variance to judge whether image information is enriched;

The distance of employing feature histogram carrys out the similitude between computed image region, to determine which kind of video format this image may be more;

Time-domain is carried out multiframe ballot differentiate, finally export the form of current video program, control display unit and play according to the video format detected.

3. the method for automatic detection 3 D video form according to claim 2, is characterized in that:

Described method comprises the steps:

Step S1. is to each counter O reset;

Step S2. obtains a two field picture, carries out view division to input picture; Respectively input picture is divided into the first subregion, the second subregion by the different video format of setting; A corresponding counter of video format;

The feature histogram of all subregion of step S3. statistical picture, whole frame feature histogram, and the variance of whole frame feature histogram;

Step S4. judges the first subregion under same video format, whether the variance of whole frame feature histogram that the second subregion is corresponding is greater than the first predetermined threshold value, if then think valid frame, performs step S5, otherwise performs step S2;

Step S5. calculates the distance between the feature histogram of two sub regions corresponding to each video format;

Step S6. gets the minimum value of each video format respective distances in step S5, judges whether to be less than the second predetermined threshold value; If be less than the second predetermined threshold value, and distance minimum value is the feature histogram distance of certain form correspondence first subregion, the second subregion, then the counter that this form is corresponding adds 1;

Step S7. repeats step S2 to step S6, until step S6 discriminating step has been performed m time of setting;

If the value of the corresponding counter of certain video format of step S8. is more than or equal to k*m, and k ∈ (0.5,1], then the current video program form detected is this 3D video format;

If the value of all counters is all not more than k*m, and k ∈ (0.5,1], then the current video program detected is 2D video format.

4. the method for automatic detection 3 D video form according to claim 2, is characterized in that:

In step S2, be divided into left subregion, right subregion by left and right 3D format; Sub-zones, lower subregion is divided into by upper and lower 3D format;

In step S5, the distance between the feature histogram calculating each sub regions, comprises the feature histogram distance between left subregion, right subregion, the feature histogram distance between sub-zones, lower subregion.

5. the method for automatic detection 3 D video form according to claim 2, is characterized in that:

At least one that in described step S2, video format comprises left and right 3D form, 3D form, line interlacing 3D form, the staggered 3D form of row, grid interlock in 3D form, grid 3D form arranged side by side.

6. the method for automatic detection 3 D video form according to claim 2, is characterized in that:

The feature histogram that described method adopts is brightness histogram.

7. the method for automatic detection 3 D video form according to claim 2, is characterized in that:

Described method adopts at least one in brightness histogram, chroma histogram, histogram of gradients, by calculating setpoint distance, similitude in computed image between zones of different, and multiple similarity measurement information is merged, obtain final similarity indices thus complete single frames form judging.