CN101853286B - Intelligent selection method of video thumbnails - Google Patents

Abstract

Present invention is disclosed a kind of intelligent selection method of video thumbnails, comprise the following steps: 1) obtain the N hypertonic sketch map of one section of video, wherein, 2≤N≤100, and N is constant；2) every hypertonic sketch map is calculated its front face position and size, face position and size, whole body position of human body and size, upper body position and size and lower part of the body position and size, grey level histogram Distribution Value, grey level histogram variance Distribution Value, colored RGB histogram variances Distribution Value, color RGB image become the discrete distance value between gray level image difference value, the variance yields of gray level image pixel, the variance yields of color RGB image pixel, colored RGB block etc. feature；3) every hypertonic sketch map is tried to achieve by ambiguity function respectively the score value of above-mentioned each feature；4) the feature score value of every hypertonic sketch map is obtained a final mark by weighted evaluation；5) according to the size of mark, N hypertonic sketch map is ranked up；6) thumbnail that selection mark is the highest is as the thumbnail representing this video.

Description

Intelligent selection method of video thumbnails

Technical field

The present invention relates to picture choosing method, particularly to a kind of video thumbnails carried out by Computer Image Processing Intelligent selecting method.

Background technology

In the daily use of video website, there is an important technology to extract a two field picture exactly from video file and make For the representative picture of this video file, our this picture is called the thumbnail of video file.Our way is from video at present File extracts eight hypertonic sketch maps, and the 4th thumbnail by default is set.This employing fixed position method is given tacit consent to And the thumbnail obtained is frequently not the most representational thumbnail in eight hypertonic sketch maps.It is permissible that we have employed certain methods Solve this problem, such as (1) user and can determine any pictures thumbnail by default by self-defining method, (2) right that website staff selects which pictures thumbnail by default is given.Although said method can solve this Problem, but reality still cannot meet requirement, this is because (1) is by data analysis, it is understood that user generally will not or Little self-defined thumbnail (2) is the most numerous due to the video file quantity uploading to video website every day, is worked by website The workload that personnel define thumbnail is the heaviest.Based on above-mentioned two reason, we invent one can pass through computer The method of image procossing selects the method for representative thumbnail thumbnail by default automatically.

P.Viola and M.Jones et al. is in the computer vision and pattern recognition international conference of calendar year 2001 What (Proceedings ofComputer Vision and Pattern Recognition, 2001) proposed " uses simple special The Stacked type lifting method levied carries out quick object detection " (Rapid Object Detection Using a Boosted Cascade of SimpleFeatures.) method carries out detecting.

In view of this, those skilled in the art are for the problems referred to above, it is provided that one can pass through Computer Image Processing Method selects the intelligent selection method of video thumbnails of representative thumbnail thumbnail by default automatically.

Summary of the invention

The invention provides a kind of intelligent selection method of video thumbnails, overcome the difficulty of prior art, to reach logical Cross Computer Image Processing method and select the purpose of representative thumbnail thumbnail by default automatically.

The present invention adopts the following technical scheme that

The invention provides a kind of intelligent selection method of video thumbnails, should for selecting a representative from one section of video The thumbnail of video, comprises the following steps:

(1) obtain the N hypertonic sketch map of one section of video, wherein, 2≤N≤100, and N is constant；

(2) every hypertonic sketch map is calculated position and size, the position of side face and size, the whole body people of its front face The position of body and size, the position of upper body and size and the position of the lower part of the body and size, grey level histogram Distribution Value, grey level histogram Variance Distribution Value, colored RGB histogram variances Distribution Value, color RGB image become the difference value of gray level image, gray level image picture Discrete distance value between the variance yields of vegetarian refreshments, the variance yields of color RGB image pixel, colored RGB block etc. feature；

(3) every hypertonic sketch map is tried to achieve by ambiguity function respectively the score value of above-mentioned each feature；

(4) the feature score value of every hypertonic sketch map is obtained a final mark by weighted evaluation；

(5) according to the size of mark, N hypertonic sketch map is ranked up；

(6) thumbnail that selection mark is the highest is as the thumbnail representing this video.

Preferably, in described step (1), use extracted at equal intervals method, according to the video duration being previously obtained, it is put down Being divided into N equal portions, the most each equal portions extract a frame as thumbnail, and N=8.

Preferably, described step (2) comprises the following steps:

(21) input thumbnail, detects face characteristic, including the position of front face and size, the position of side face and Size, and calculate fractional value；

(22) input thumbnail, detects characteristics of human body, including position and size, the position of upper body and the size of whole body human body With position and the size of the lower part of the body, and calculate fractional value；

(23) input thumbnail, calculates grey level histogram Distribution Value；

(24) input thumbnail, calculates grey level histogram variance Distribution Value；

(25) input thumbnail, calculates colored RGB histogram variances Distribution Value；

(26) input thumbnail, calculates color RGB image and becomes the difference value of gray level image；

(27) input thumbnail, calculates the variance yields of gray level image pixel；

(28) input thumbnail, calculates the variance yields of color RGB image pixel；

(29) input thumbnail, calculates the discrete distance value between colored RGB block.

Preferably, in described step (21):

The then score value of front face

Wherein w_iFor the width of front face, n is face number, if W is the width of thumbnail；

The score value of side face

Wherein w_iFor the width of side face, n is face number, if W is the width of thumbnail.

Preferably, described step (22) sets (x_i, y_i, w_i, h_i), i=1 ..., n, wherein x_i, y_iRepresent face location, w_iFor the width of face, n is face number, if W is the width of thumbnail,

The then score value of whole body

Wherein w_iFor the width of whole body human body, n is number, if W is the width of thumbnail；

The score value of upper body

Wherein w_iFor the width of whole body human upper limb, n is number, if W is the width of thumbnail；

The score value of the lower part of the body

Wherein w_iFor the width of whole body human body lower limbs, n is number, if W is the width of thumbnail.

Preferably, input thumbnail in described step (23), if its grey level histogram is gray_hist=(gh_1, gh_ 2 .., gh_n), n is histogrammic number, and t1, t2 are integer thresholds, T1 and T2 is the threshold value of rectangular histogram sum, base_score On the basis of mark, then the computing formula of grey level histogram Distribution Value score_hist_distrib is:

Wherein, n=10, t1=2, t2=8, T₁=0.7, T₁=0.8, base_score=1.0.

Preferably, input thumbnail in described step (24), if its grey level histogram is gray_hist=(gh_1, gh_ 2 .., gh_n), n is histogrammic number,Max_num=max (gh_1, gh_2 .., gh_n), base_ Mark on the basis of score, then the score_gray_hist_std computing formula calculating grey level histogram variance Distribution Value is:

Wherein, n=64, base_score=1.0.

Preferably, histogram variances Distribution Value in R plane, G plane, B plane respectively in described step (25), then ask All it is worth to colored RGB histogram variances Distribution Value score_RGB_hist_std；

$\mathrm{score}\_\mathrm{RGB}\_\mathrm{hist}\_\mathrm{distrib}=\frac{\mathrm{score}\_R\_\mathrm{hist}\_\mathrm{distrib}+\mathrm{score}\_G\_\mathrm{hist}\_\mathrm{distrib}+\mathrm{score}\_B\_\mathrm{hist}\_\mathrm{distrib}}{3}$

；Wherein, n=64, base_score=1.0.

Preferably, input thumbnail in described step (26), if the pixel value of gray level image is x_ij, i=1 ..., W, j= 1 ..., H, the pixel value (r of color RGB image_ij, g_ij, b_ij), i=1 ..., W, j=1 ..., H, wherein W, H is respectively contracting The width of sketch map and height, mark on the basis of base_score1 and base_score2, t1 is threshold value, if then calculating color RGB image The computing formula of difference value score_rgb2gray_diff becoming gray level image is:

Wherein,Base_score1=1.0, base_ Score2=2.0, t₁=-0.9.

Preferably, input thumbnail in described step (27), if the pixel value of gray level image is x_ij, i=1 ..., W, j= 1 ..., H, wherein W, H is respectively width and the height of thumbnail, then calculates the score_gray_ of the variance yields of gray level image pixel Std computing formula is:

$\mathrm{score}\_\mathrm{gray}\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}}.$

Preferably, respectively at R plane, G plane, the variance yields of B plane epigraph pixel in described step (28), then Average and obtain the variance yields score_RGB_std of color RGB image pixel；

$\mathrm{score}\_R\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_G\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_B\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_\mathrm{RGB}\_\mathrm{std}=\frac{\mathrm{score}\_R\_\mathrm{std}+\mathrm{score}\_G\_\mathrm{std}+\mathrm{score}\_B\_\mathrm{std}}{3}.$

Preferably, input thumbnail in described step (29), if the pixel value (r of color RGB image_i, g_i, b_i), i= 1 ..., W × H, wherein W, H is respectively width and the height of thumbnail, mark on the basis of base_score, then calculate colored RGB block it Between the computing formula of discrete distance value score_rgb_dist be:

Wherein, base_score=0.0.

Preferably, in described step (4), final score value final_score is calculated by following equation:

Final_score=w₁×score_face_frontal+w₂×score_face_profile+w₃×score_ full_body+w₄×score_upper_body+w₅×score_lower_body+w₆×score_hist_distrib+w₇ ×score_gray_hist_std+w₈×score_RGB_hist_std+w₉×score_rgb2gray_diff+w₁₀× score_gray_std+w₁₁×score_RGB_std；

Wherein, w₁=1.0, w₂=0.5, w₃=0.5, w₄=0.5, w₅=0.5, w₆=0.5, w₇=0.5, w₈=1.2, W₉ =1.2, w₁₀=0.5, w₁₁=1.2.

Owing to have employed above-mentioned technology, compared with prior art, present invention have the advantage that the present invention can be by meter Calculation machine image processing method selects representative thumbnail thumbnail by default automatically.

The present invention is further illustrated below in conjunction with drawings and Examples.

Accompanying drawing explanation

Fig. 1 is the flow chart of the intelligent selection method of video thumbnails of the present invention.

Detailed description of the invention

A kind of specific embodiment of the present invention is introduced below by Fig. 1.

As it is shown in figure 1, a kind of intelligent selection method of video thumbnails, comprise the following steps:

(1) obtaining the N hypertonic sketch map of a video, the acquisition methods of N hypertonic sketch map is to use extracted at equal intervals method, i.e. According to the video duration being previously obtained, it being divided into N equal portions, the most each equal portions extract a frame as thumbnail.Here Taking N=8, but be not limited to N=8, N can be between 2 to 100.

(2) every hypertonic sketch map is calculated position and size, the position of side face and size, the whole body people of its front face The position of body and size, the position of upper body and size and the position of the lower part of the body and size, grey level histogram Distribution Value, histogram variances Distribution Value, colored RGB histogram variances Distribution Value, color RGB image become the difference value of gray level image, gray level image pixel Variance yields, the variance yields of color RGB image pixel, discrete distance value between colored RGB block etc. feature.

(3) score value of features described above is tried to achieve respectively by ambiguity function.

(4) a final mark is obtained by weighted evaluation.

(5) according to the size of mark, N hypertonic sketch map is ranked up.

(6) thumbnail that score value is the highest thumbnail by default.

Above-mentioned steps (2), is carried out as follows:

(21) input thumbnail, detects face characteristic, includes position and size, the position of side face of front face And size；Calculate fractional value；

(22) input thumbnail, detects characteristics of human body, includes the position of whole body human body and size, the position of upper body and big The position of the little and lower part of the body and size；Calculate fractional value；

(23) input thumbnail, calculates grey level histogram Distribution Value；

(24) input thumbnail, calculates grey level histogram variance Distribution Value；

(25) input thumbnail, calculates colored RGB histogram variances Distribution Value；

(26) input thumbnail, calculates color RGB image and becomes the difference value of gray level image；

(27) input thumbnail, calculates the variance yields of gray level image pixel；

(28) input thumbnail, calculates the variance yields of color RGB image pixel；

(29) input thumbnail, calculates the discrete distance value between colored RGB block.

Embodiments of the present invention are as follows:

With continued reference to Fig. 1, a kind of intelligent selection method of video thumbnails of the present invention, for selecting one from one section of video Zhang represents the thumbnail of this video, comprises the following steps:

Step (1) obtains the N hypertonic sketch map of one section of video, wherein, 2≤N≤100, and N is constant, uses and takes out at equal intervals Access method, according to the video duration being previously obtained, is divided into N equal portions by it, and the most each equal portions extract a frame as breviary Figure, and N=8.

Step (2) calculates the position of its front face and size, the position of side face and size, complete to every hypertonic sketch map The position of body human body and size, the position of upper body and size and the position of the lower part of the body and size, grey level histogram Distribution Value, gray scale are straight Side's figure variance Distribution Value, colored RGB histogram variances Distribution Value, color RGB image become the difference value of gray level image, gray-scale map As the discrete distance value between the variance yields of pixel, the variance yields of color RGB image pixel, colored RGB block etc. feature；

Described step comprises the following steps in (2):

Step (21) input thumbnail, detects face characteristic, including position and size, the position of side face of front face Put and size, and calculate fractional value；

Step (22) input thumbnail, detects characteristics of human body, including the position of whole body human body and size, the position of upper body and Size and the position of the lower part of the body and size, and calculate fractional value；

Step (23) input thumbnail, calculates grey level histogram Distribution Value；

Step (24) input thumbnail, calculates grey level histogram variance Distribution Value；

Step (25) input thumbnail, calculates colored RGB histogram variances Distribution Value；

Step (26) input thumbnail, calculates color RGB image and becomes the difference value of gray level image；

Step (27) input thumbnail, calculates the variance yields of gray level image pixel；

Step (28) input thumbnail, calculates the variance yields of color RGB image pixel；

Step (29) input thumbnail, calculates the discrete distance value between colored RGB block.

In described step (21): the then score value of front face Wherein w_iFor the width of front face, n is face number, if W is the width of thumbnail；The score value of side face

Wherein w_iFor the width of side face, n is face number, if W is the width of thumbnail.

Described step sets (x in (22)_i, y_i, w_i, h_i), i=1 ..., n, wherein x_i, y_iRepresent face location, w_iFor face Width, n is face number, if W is the width of thumbnail,

The then score value of whole body

Wherein w_iFor the width of whole body human body, n is number, if W is the width of thumbnail；

The score value of upper body

Wherein w_iFor the width of whole body human upper limb, n is number, if W is the width of thumbnail；

The score value of the lower part of the body

Wherein w_iFor the width of whole body human body lower limbs, n is number, if W is the width of thumbnail.

Input thumbnail in described step (23), if its grey level histogram is gray_hist=(gh_1, gh_2 .., gh_ N), n is histogrammic number, and t1, t2 are integer thresholds, T1 and T2 is the threshold value of rectangular histogram sum, divides on the basis of base_score Number, then the computing formula of grey level histogram Distribution Value score_hist_distrib is:

Wherein, n=10, t1=2, t2=8, T₁=0.7, T₁=0.8, base_score=1.0.

Input thumbnail in described step (24), if its grey level histogram is gray_hist=(gh_1, gh_2 .., gh_ N), n is histogrammic number,Max_num=max (gh_1, gh_2 .., gh_n), base_score are base Quasi-mark, then the score_gray_hist_std computing formula calculating grey level histogram variance Distribution Value is:

Wherein, n=64, base_score=1.0.

Histogram variances Distribution Value in R plane, G plane, B plane respectively in described step (25), then averages To colored RGB histogram variances Distribution Value score_RGB_hist_std；

$\mathrm{score}\_\mathrm{RGB}\_\mathrm{hist}\_\mathrm{distrib}=\frac{\mathrm{score}\_R\_\mathrm{hist}\_\mathrm{distrib}+\mathrm{score}\_G\_\mathrm{hist}\_\mathrm{distrib}+\mathrm{score}\_B\_\mathrm{hist}\_\mathrm{distrib}}{3}$

；Wherein, n=64, base_score=1.0.

Input thumbnail in described step (26), if the pixel value of gray level image is x_ij, i=1 ..., W, j=1 ..., H, the pixel value (r of color RGB image_ij, g_ij, b_ij), i=1 ..., W, j=1 ..., H, wherein W, H is respectively thumbnail Wide and high, mark on the basis of base_score1 and base_score2, t1 is threshold value, if then calculating color RGB image to become ash The computing formula of difference value score_rgb2gray_diff of degree image is:

Wherein,Base_score1=1.0, base_ Score2=2.0, t₁=-0.9.

Input thumbnail in described step (27), if the pixel value of gray level image is x_ij, i=1 ..., W, j=1 ..., H, wherein W, H is respectively width and the height of thumbnail, then calculate the score_gray_std meter of the variance yields of gray level image pixel Calculation formula is:

$\mathrm{score}\_\mathrm{gray}\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}}.$

Respectively at R plane, G plane, the variance yields of B plane epigraph pixel in described step (28), then average Obtain the variance yields score_RGB_std of color RGB image pixel；

$\mathrm{score}\_R\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_G\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_B\_\mathrm{std}=-\mathrm{base}\_\mathrm{score}+\sqrt{\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-\stackrel{\‾}{x})}^2}/255,$ Wherein $\stackrel{\‾}{x}=\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\underset{j=1}{\overset{H}{\mathrm{\Σ}}}{x}_{\mathrm{ij}};$

$\mathrm{score}\_\mathrm{RGB}\_\mathrm{std}=\frac{\mathrm{score}\_R\_\mathrm{std}+\mathrm{score}\_G\_\mathrm{std}+\mathrm{score}\_B\_\mathrm{std}}{3}.$

Input thumbnail in described step (29), if the pixel value (r of color RGB image_i, g_i, b_i), i=1 ..., W × H, wherein W, H is respectively width and the height of thumbnail, mark on the basis of base_score, then calculate between colored RGB block is discrete The computing formula of distance value score_rgb_dist is:

$\mathrm{score}\_\mathrm{rgb}\_\mathrm{dist}=-\mathrm{base}\_\mathrm{score}+\underset{i=1}{\overset{W\×H}{\mathrm{\Σ}}}\underset{j=1,j\≠i}{\overset{W\×H}{\mathrm{\Σ}}}\left(\right|r_i-r_j|+|g_i-g_j|+|b_i-b_j\left|\right)/(3\×W\×H\×255),$ Wherein, base_score=0.0.

The score value of above-mentioned each feature tried to achieve by every hypertonic sketch map by step (3) respectively by ambiguity function；

Step (4) obtains a final mark to the feature score value of every hypertonic sketch map by weighted evaluation, and final divides Value final_score is calculated by following equation:

Final_score=w₁×score_face_frontal+w₂×score_face_profile+w₃×score_ full_body+w₄×score_upper_body+w₅×score_lower_body+w₆×score_hist_distrib+w₇ ×score_gray_hist_std+w₈×score_RGB_hist_std+w₉×score_rgb2gray_diff+w₁₀× score_gray_std+w₁₁×score_RGB_std；

Wherein, w₁=1.0, w₂=0.5, w₃=0.5, w₄=0.5, w₅=0.5, w₆=0.5, w₇=0.5, w₈=1.2, W₉ =1.2, w₁₀=0.5, w₁₁=1.2.

N hypertonic sketch map is ranked up by step (5) according to the size of mark；

Step (6) selects the highest thumbnail of mark as the thumbnail representing this video.

In summary, owing to have employed above-mentioned technology, present invention have the advantage that the present invention can pass through computer graphic Representative thumbnail thumbnail by default is selected automatically as processing method.

Embodiment described above is merely to illustrate technological thought and the feature of the present invention, in its object is to make this area Technical staff will appreciate that present disclosure and implement according to this, it is impossible to only limit the patent model of the present invention with the present embodiment Enclose, the most all equal changes made according to disclosed spirit or modification, still fall in the scope of the claims of the present invention.

Claims (12)

1. an intelligent selection method of video thumbnails, for selecting a thumbnail representing this video from one section of video, It is characterized in that: comprise the following steps:

(1) obtain the N hypertonic sketch map of one section of video, wherein, 2≤N≤100, and N is constant；

(2) every hypertonic sketch map is calculated the position of its front face and size, the position of side face and size, whole body human body Position and size, the position of upper body and size and the position of the lower part of the body and size, grey level histogram Distribution Value, grey level histogram variance Distribution Value, colored RGB histogram variances Distribution Value, color RGB image become the difference value of gray level image, gray level image pixel Variance yields, the variance yields of color RGB image pixel, discrete distance value tag between colored RGB block；

(3) every hypertonic sketch map is tried to achieve by ambiguity function respectively the score value of above-mentioned each feature；

(4) the feature score value of every hypertonic sketch map is obtained a final mark by weighted evaluation；

(5) according to the size of mark, N hypertonic sketch map is ranked up；

(6) thumbnail that selection mark is the highest is as the thumbnail representing this video；

Wherein, described step (2) comprises the following steps:

(21) input thumbnail, detects face characteristic, including the position of front face and size, the position of side face and big Little, and calculate fractional value；

(22) input thumbnail, detects characteristics of human body, including the position of whole body human body and size, the position of upper body and size and under The position of body and size, and calculate fractional value；

(23) input thumbnail, calculates grey level histogram Distribution Value；

(24) input thumbnail, calculates grey level histogram variance Distribution Value；

(25) input thumbnail, calculates colored RGB histogram variances Distribution Value；

(26) input thumbnail, calculates color RGB image and becomes the difference value of gray level image；

(27) input thumbnail, calculates the variance yields of gray level image pixel；

(28) input thumbnail, calculates the variance yields of color RGB image pixel；

(29) input thumbnail, calculates the discrete distance value between colored RGB block.

2. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: in described step (1), employing etc. Interval abstracting method, according to the video duration being previously obtained, is divided into N equal portions by it, and the most each equal portions extract a frame and make For thumbnail, and N=8.

3. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: in described step (21):

The score value of front face

Wherein w_iFor the width of front face, n is face number, if W is the width of thumbnail；

The score value of side face

Wherein w_iFor the width of side face, n is face number, if W is the width of thumbnail.

4. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: described step sets (x in (22)_i, y_i, w_i, h_i), i=1 ..., n, wherein x_i, y_iRepresent face location, w_iFor the width of face, n is face number, if W is breviary The width of figure,

The then score value of whole body

Wherein w_iFor the width of whole body human body, n is number, if W is the width of thumbnail；

The score value of upper body

Wherein w_iFor the width of whole body human upper limb, n is number, if W is the width of thumbnail；

The score value of the lower part of the body

Wherein w_iFor the width of whole body human body lower limbs, n is number, if W is the width of thumbnail.

5. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: input contracting in described step (23) Sketch map, if its grey level histogram is gray_hist=(gh₁, gh₂.., gh_n), gh_iFor i-th grey level histogram, n is Nogata The number of figure, t1, t2 are integer thresholds, T₁And T₂For the threshold value of rectangular histogram sum, mark on the basis of base_score, then gray scale is straight The computing formula of side figure Distribution Value score_hist_distrib is:

Wherein, n=10, t1=2, t2=8, T₁=0.7, T₁=0.8, base_score=1.0.

6. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: input contracting in described step (24) Sketch map, if its grey level histogram is gray_hist=(gh₁, gh₂.., gh_n), n is histogrammic number, Max_num=max (gh₁, gh₂.., gh_n), mark on the basis of base_score, then calculate grey level histogram variance Distribution Value Score_gray_hist_std computing formula is:

When max_num ≠ 0

Wherein, n=64, base_score=1.0.

7. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: described step exists in (25) respectively Histogram variances Distribution Value in R plane, G plane, B plane, then averages and obtains colored RGB histogram variances Distribution Value score_RGB_hist_std；

Wherein, max_num=max (gh₁, gh₂.., gh_n),Forgh_iFor i-th grey level histogram, n= 64, base_score=1.0.

8. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: input contracting in described step (26) Sketch map, if the pixel value of gray level image is x_ij, i=1 ..., W, j=1 ..., H, the pixel value (r of color RGB image_ij, g_ij, b_ij), i=1 ..., W, j=1 ..., H, wherein W, H is respectively width and the height of thumbnail, base_score1 and base_ Mark on the basis of score2, t₁For threshold value, if then calculating color RGB image to become difference value score_ of gray level image The computing formula of rgb2gray_diff is:

Wherein,Base_score1=1.0, base_score2= 2.0, t₁=-0.9.

9. intelligent selection method of video thumbnails AA as claimed in claim 1, it is characterised in that: input in described step (27) Thumbnail, if the pixel value of gray level image is x_ij, i=1 ..., W, j=1 ..., H, wherein W, H be respectively thumbnail width and Height, mark on the basis of base_score, then calculate the score_gray_std computing formula of the variance yields of gray level image pixel For:

Wherein

10. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: in described step (28) respectively At R plane, G plane, the variance yields of B plane epigraph pixel, then average and obtain the variance of color RGB image pixel Value score_RGB_std；Wherein mark on the basis of base_score, x_ijPixel value for gray level image

Wherein

Wherein

Wherein

11. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: input in described step (29) Thumbnail, if the pixel value (r of color RGB image_i, g_i, b_i), i=1 ..., W × H, wherein W, H be respectively thumbnail width and Height, mark on the basis of base_score, then the calculating of discrete distance value score_rgb_dist between the colored RGB block of calculating is public Formula is:

Wherein, base_score=0.0.

12. intelligent selection method of video thumbnails as claimed in claim 1, it is characterised in that: final in described step (4) Score value final_score is calculated by following equation:

Final_score=w₁×score_face_frontal+w₂×score_face_profile+w₃×score_full_ body+w₄×score_upper_body+w₅×score_lower_body+w₆×score_hist_distrib+w₇× score_gray_hist_std+w₈×score_RGB_hist_std+w₉×score_rgb2gray_diff+w₁₀×score_ gray_std+w₁₁×score_RGB_std；

Wherein, w₁=1.0, w₂=0.5, w₃=0.5, w₄=0.5, w₅=0.5, w₆=0.5, w₇=0.5, w₈=1.2, W₉= 1.2, w₁₀=0.5, w₁₁=1.2, score_face_frontal are the score value of front face, and score_face_profile is The score value of side face, score_full_body is whole body score value, and score_upper_body is the score value of upper body, score_ Lower_body is the score value of the lower part of the body, and score_hist_distrib is grey level histogram Distribution Value, score_gray_hist_ Std is grey level histogram variance Distribution Value, and score_RGB_hist_std is colored RGB histogram variances Distribution Value, score_ Rgb2gray_diff is the difference value that color RGB image becomes gray level image, and score_gray_std is gray level image pixel Variance yields, score_RGB_std is the variance yields of color RGB image pixel.