CN101369281A - Retrieval method based on video abstract metadata - Google Patents
Retrieval method based on video abstract metadata Download PDFInfo
- Publication number
- CN101369281A CN101369281A CNA2008101702190A CN200810170219A CN101369281A CN 101369281 A CN101369281 A CN 101369281A CN A2008101702190 A CNA2008101702190 A CN A2008101702190A CN 200810170219 A CN200810170219 A CN 200810170219A CN 101369281 A CN101369281 A CN 101369281A
- Authority
- CN
- China
- Prior art keywords
- video
- metadata
- abstract
- image
- probability density
- 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.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the video information processing field, particularly to a fast and novel searching method based on video summary metadata aiming at video search, comprising: extracting a key frame in the video data, building a video summary; respectively solving gray scale distribution and probability density function for each frame of image in the video summary; building the metadata through gray scale distribution and probability density function, storing in a database; comparing the metadata of the original data for searching and the metadata of each video summary in the database, using the video image with highest semblance as a searching result. The invention generates the metadata of the video summary according to gray scale distribution and probability density function of the key frame in the video image, obtains the video search result through comparing the similarity of the metadata, fast and correctly searches out the required video content for the user compared with the traditional method in the video search system based on the content.
Description
Technical field
The present invention relates to the video information process field, be specifically related to propose a kind of new fast based on the search method of video abstract metadata at video frequency searching.
Background technology
Video frequency searching is exactly to retrieve the video clips that oneself needs from the video data of magnanimity.Video frequency abstract promptly is the brief summary to a long section video content.Clearer and more definite more, video frequency abstract is exactly the image of a succession of static or motion, and their have represented the content of former video with the mode of simplifying, and have kept the main points of raw content.
Along with the increase of the accumulation of TV station's video frequency program, online digital video, and a large amount of multimedia application such as digital library, video request program, remote teaching, how in the magnanimity video, to retrieve needed data fast and seem most important.Because the video data volume is big, has time and space structure, its expression, storage, transmission, tissue have bigger difficulty, have only and reasonably organize these video datas, could browse, retrieve these information effectively.
Relatively more classical video retrieval technology is based on the video frequency searching of content, this is that a kind of more specific vision content is described (the more approaching mankind's perception), can give full play to the exchange method of human perception, the user not only can be by the text query method, also can browse and utilize search engine to enter index with vision query by example method to retrieve by vision, this technology is directly to carry out the inquiry of image information according to picture material, promptly extract the feature of entire image, compare with the feature of want retrieving images, judge its similarity according to certain similarity matching criterior.It is based on the comparison of the whole information of two width of cloth pictures, and user's request is a part in the width of cloth picture, and other irrelevant information has just influenced calculated amount and judged the accuracy of picture correlativity so.A kind of like this video retrieval technology do not generate yet an effective video structure come for the user browse, retrieve video information.
Summary of the invention
The present invention has overcome the defective of above-mentioned existing video frequency searching algorithm, provide a kind of efficiently, fast based on the search method of video abstract metadata.
The present invention solves the technical scheme that its technical matters takes: a kind of search method based on video abstract metadata comprises the steps:
Extract the key frame in the video data, set up video frequency abstract;
To the every two field picture in the video frequency abstract, obtain intensity profile and probability density function respectively;
Utilize intensity profile and probability density function to set up metadata, deposit database in;
According to comparing as the metadata of the original image of retrieving and the metadata of each video frequency abstract in the database, getting the highest video image of similarity is result for retrieval.
The step of the key frame in described extraction video data comprises that further video data is carried out camera lens to be sorted out, and chooses designated frame in the current camera lens respectively the camera lens after sorting out separately as key frame.
Every two field picture in the video frequency abstract is divided, the each several part after dividing is obtained intensity profile and probability density function respectively.
Described every two field picture in the video frequency abstract is divided, comprised that with image division be scene part and theme part.
The classification of described camera lens can be sorted out according to preset threshold according to the transinformation content that utilizes the adjacent image frame.
The comparison of described similarity is specially, and calculates the metadata in the described original image and the transinformation content of the metadata in described each video frequency abstract one by one, and getting transinformation content the maximum is the highest video image of similarity.
According to the step that the metadata of metadata that is used as the original image of retrieving and each video frequency abstract in the database is compared, further be that the metadata of dividing the back each several part in the metadata of original image and each video frequency abstract in the database is compared.
The metadata that the present invention produces video frequency abstract according to the intensity profile and the probability density function of key frame in the video image, utilization is to the comparison of the similarity of metadata, obtain the video frequency searching result, compare the method in traditional Content-based Video Retrieval system, can retrieve the needed video content of user more fast and accurately.
Description of drawings
Fig. 1 is the process flow diagram that generates video metadata among the present invention;
Fig. 2 is transinformation content experimental data figure as a result between the image of news video fragment among the present invention;
Fig. 3 utilizes metadata to judge the process flow diagram of the similarity of two sections video place scenes among the present invention.
Embodiment
The invention provides a kind of search method based on video abstract metadata, mainly is that the metadata structure according to video information carries out video frequency searching, by structurized metasearch and determine the needed video content of user.Its specific implementation process is as follows:
1. passing threshold θ carries out the camera lens classification and extracts key frame, and sets up video frequency abstract with these key frames.
By repeatedly experiment, calculate two adjacent image transinformation contents, reasonable setting threshold θ, when the transinformation content of two adjacent images surpasses this threshold value θ, think that then the picture in the camera lens changes, camera lens is sorted out, and chosen designated frame in the current camera lens the camera lens after sorting out, for example the first frame in the set of shots, tail frame or intermediate frame as key frame.
2. each two field picture in the video frequency abstract is divided, and obtained intensity profile and the probability density function of dividing the back each several part respectively.
As shown in fig. 1, from video frequency abstract, generate the process flow diagram of video metadata, read each frame in the video frequency abstract frame sequence at first one by one, each two field picture is divided, for example, different according to scene and theme picture material, K is divided into scene and two parts of theme in proportion, for the independent image of a width of cloth, calculate the intensity profile and the probability density of scene part or theme part, preserve the result who obtains.
Can think its separately the gray-scale value of pixel be sample independently, then this width of cloth gray distribution of image is P={P
0, P
1, P
2..., P
L-1, P wherein
iIt for gray-scale value the ratio of the total picture element of number and the image of picture element of i.If two image A, B being compared have identical grey level (as: being 256 grades of gray level images), the gray level of establishing image is L, makes P
A(a) and P
B(b) probability density function of presentation video A, B respectively, probability density function can be easily obtained divided by the total pixel number of image by the histogram of image, make P
AB(it is on the basis of the associating grey level histogram of the image A of obtaining, B for a, the b) joint probability density of presentation video A, B, and the pixel number total divided by image obtains.
In lens detection method,, calculate the transinformation content of its three RGB components respectively independently for adjacent two continuous frames.Adjacent image frame t, the transinformation content of t+1 on the R component can be expressed as
In like manner can obtain adjacent image frame t, the transinformation content of t+1 on G, B component, image t then, transinformation content total between the t+1 can be expressed as
I
(t，t+1)＝I
R (t，t+1)+I
G (t，t+1)+I
B (t，t+1)
Fig. 2 shows between each two field picture of one section news video transinformation content experimental data figure as a result, as can be seen, for a two field picture, its intensity profile and probability density are two important parameters of expressing picture material, and transinformation content has well been described the correlativity between the image.
3. set up the video information metadata structure, at first, generate video metadata, promptly by the scene of each two field picture of statistics in second step or the intensity profile of theme part, calculating probability density function P (a) obtains a plurality of metadata.Secondly, the video metadata that generates being carried out structuring handles; Structuring video metadata, its organization definition are sequence number, video file name, scene name, subject and metadata.Wherein metadata is made up of the probability density function of gradation of image distribution and image.
Described metadata is the structured message with description, explanation, locating information resource function, had metadata can so that obtain, use and management information is more prone to, popular, metadata is exactly the data of data of description, the perhaps information of descriptor, we use the metadata relevant with video to describe the main contents of video, and information extraction generates metadata from video, when follow-up the retrieval, compare then with metadata.
The present invention generates the metadata of video and carries out video frequency searching, and the probability density function of distribution of definition gradation of image and image is formed the metadata as this picture as the metadata of this picture, preserves the formal definition of metadata, and dominant term is as follows:
<pictureMeta>
" gray-scale value is the number and the total picture element of image of the picture element of i to＜intensity profile means=
Ratio "
＜pixel 〉
＜gray scale 〉
<value>i<\value>
＜gray scale
＜number 〉
<value>number<\value>
＜number 〉
＜pixel
＜intensity profile
＜probability density function means=" the ratio of total number of pixels of image histogram and image
″>
＜pixel 〉
＜gray scale 〉
<value>i<\value>
＜gray scale
＜number 〉
<value>number<\value>
＜number 〉
＜pixel
＜probability density function
<\pictureMeta>
That is, form as the video metadata structure in the following table:
4. last, the use of video metadata utilizes metadata to judge the similarity of scene in original retrieving image and the picture database or theme.
As shown in Figure 3, for utilizing metadata to judge the process flow diagram of the similarity of two sections video place scenes among the present invention.At first to the input as the retrieval original image A, calculate this gray distribution of image S and probability density P (a) according to method in above-mentioned 2,3 steps, take out the video metadata that writes down in the database then one by one, calculate both transinformation contents, thereby, get the similarity soprano and obtain result for retrieval by judging the similarity of the two.
By in step 2 to the division of picture frame, be convenient to retrieval to specific factor, for example retrieve the video image that carries out of particular topic, or the video image of retrieval special scenes etc., and, can carry out other forms of division according to actual needs, for example, divide etc. according to the different colours in the picture frame.
More than the search method based on video abstract metadata provided by the present invention is described in detail, used specific case herein principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.
Claims (7)
1. the search method based on video abstract metadata is characterized in that, comprises the steps:
Extract the key frame in the video data, set up video frequency abstract;
To the every two field picture in the video frequency abstract, obtain intensity profile and probability density function respectively;
Utilize intensity profile and probability density function to set up metadata, deposit database in;
According to comparing as the metadata of the original image of retrieving and the metadata of each video frequency abstract in the database, getting the highest video image of similarity is result for retrieval.
2. the search method based on video abstract metadata according to claim 1, it is characterized in that: the step of the key frame in described extraction video data, comprise that further video data is carried out camera lens to be sorted out, choose designated frame in the current camera lens respectively the camera lens after sorting out separately as key frame.
3. the search method based on video abstract metadata according to claim 2 is characterized in that: the classification of described camera lens, can sort out according to preset threshold according to the transinformation content that utilizes the adjacent image frame.
4. the search method based on video abstract metadata according to claim 1 is characterized in that: the every two field picture in the video frequency abstract is divided, the each several part after dividing is obtained intensity profile and probability density function respectively.
5. the search method based on video abstract metadata according to claim 4 is characterized in that: described every two field picture in the video frequency abstract is divided, comprised that with image division be scene part and theme part.
6. according to each described search method in the claim 1～5 based on video abstract metadata, it is characterized in that: the comparison of described similarity is specially, calculate the metadata in the described original image and the transinformation content of the metadata in described each video frequency abstract one by one, getting transinformation content the maximum is the highest video image of similarity.
7. according to claim 4 or 5 described search methods based on video abstract metadata, it is characterized in that:, further be that the metadata of dividing the back each several part in the metadata of original image and each video frequency abstract in the database is compared according to the step that the metadata of metadata that is used as the original image of retrieving and each video frequency abstract in the database is compared.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2008101702190A CN101369281A (en) | 2008-10-09 | 2008-10-09 | Retrieval method based on video abstract metadata |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2008101702190A CN101369281A (en) | 2008-10-09 | 2008-10-09 | Retrieval method based on video abstract metadata |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101369281A true CN101369281A (en) | 2009-02-18 |
Family
ID=40413096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2008101702190A Pending CN101369281A (en) | 2008-10-09 | 2008-10-09 | Retrieval method based on video abstract metadata |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101369281A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102103686A (en) * | 2009-12-21 | 2011-06-22 | 韩国电子通信研究院 | Video identification system using symmetric information of graded image block and method thereof |
| CN102207966A (en) * | 2011-06-01 | 2011-10-05 | 华南理工大学 | Video content quick retrieving method based on object tag |
| CN102819528A (en) * | 2011-06-10 | 2012-12-12 | 中国电信股份有限公司 | Method and device for generating video abstraction |
| WO2014082288A1 (en) * | 2012-11-30 | 2014-06-05 | Thomson Licensing | Method and apparatus for video retrieval |
| CN104185089A (en) * | 2013-05-23 | 2014-12-03 | 三星电子（中国）研发中心 | Video summary generation method, server and client-terminal |
| CN104636330A (en) * | 2013-11-06 | 2015-05-20 | 北京航天长峰科技工业集团有限公司 | Related video rapid searching method based on structural data |
| CN105323657A (en) * | 2014-07-07 | 2016-02-10 | 韩华泰科株式会社 | Imaging apparatus and method of providing video summary |
| CN105589974A (en) * | 2016-02-04 | 2016-05-18 | 通号通信信息集团有限公司 | Surveillance video retrieval method and system based on Hadoop platform |
| CN105718597A (en) * | 2016-03-04 | 2016-06-29 | 北京邮电大学 | Data retrieving method and system thereof |
| CN105843930A (en) * | 2016-03-29 | 2016-08-10 | 乐视控股（北京）有限公司 | Video search method and device |
| CN106126590A (en) * | 2016-06-20 | 2016-11-16 | 北京航空航天大学 | A kind of unmanned plane based on key message scouts video retrieval method |
| US9799376B2 (en) | 2014-09-17 | 2017-10-24 | Xiaomi Inc. | Method and device for video browsing based on keyframe |
| CN108234883A (en) * | 2011-05-18 | 2018-06-29 | 高智83基金会有限责任公司 | Video frequency abstract including particular person |
| CN110012349A (en) * | 2019-06-04 | 2019-07-12 | 成都索贝数码科技股份有限公司 | A kind of news program structural method and its structuring frame system end to end |
| CN110019951A (en) * | 2017-09-29 | 2019-07-16 | 华为软件技术有限公司 | A kind of method and apparatus generating video thumbnails |
| CN111602411A (en) * | 2018-02-27 | 2020-08-28 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and program |
| CN112486923A (en) * | 2020-12-11 | 2021-03-12 | 北京梧桐车联科技有限责任公司 | File updating method, device and system |
| CN114297441A (en) * | 2021-12-08 | 2022-04-08 | 北京格灵深瞳信息技术股份有限公司 | Video object retrieval method and device, computer equipment and readable storage medium |
-
2008
- 2008-10-09 CN CNA2008101702190A patent/CN101369281A/en active Pending
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102103686B (en) * | 2009-12-21 | 2014-10-01 | 韩国电子通信研究院 | Video recognition system and method using symmetric information of hierarchical image blocks |
| CN102103686A (en) * | 2009-12-21 | 2011-06-22 | 韩国电子通信研究院 | Video identification system using symmetric information of graded image block and method thereof |
| CN108234883A (en) * | 2011-05-18 | 2018-06-29 | 高智83基金会有限责任公司 | Video frequency abstract including particular person |
| CN102207966A (en) * | 2011-06-01 | 2011-10-05 | 华南理工大学 | Video content quick retrieving method based on object tag |
| CN102207966B (en) * | 2011-06-01 | 2013-07-10 | 华南理工大学 | Video content quick retrieving method based on object tag |
| CN102819528B (en) * | 2011-06-10 | 2016-06-29 | 中国电信股份有限公司 | The method and apparatus generating video frequency abstract |
| CN102819528A (en) * | 2011-06-10 | 2012-12-12 | 中国电信股份有限公司 | Method and device for generating video abstraction |
| WO2014082288A1 (en) * | 2012-11-30 | 2014-06-05 | Thomson Licensing | Method and apparatus for video retrieval |
| CN104185089A (en) * | 2013-05-23 | 2014-12-03 | 三星电子（中国）研发中心 | Video summary generation method, server and client-terminal |
| CN104185089B (en) * | 2013-05-23 | 2018-02-16 | 三星电子（中国）研发中心 | Video summary generation method and server, client |
| CN104636330A (en) * | 2013-11-06 | 2015-05-20 | 北京航天长峰科技工业集团有限公司 | Related video rapid searching method based on structural data |
| CN105323657A (en) * | 2014-07-07 | 2016-02-10 | 韩华泰科株式会社 | Imaging apparatus and method of providing video summary |
| CN105323657B (en) * | 2014-07-07 | 2021-03-16 | 韩华泰科株式会社 | Imaging apparatus and method for providing video summary |
| US9799376B2 (en) | 2014-09-17 | 2017-10-24 | Xiaomi Inc. | Method and device for video browsing based on keyframe |
| CN105589974A (en) * | 2016-02-04 | 2016-05-18 | 通号通信信息集团有限公司 | Surveillance video retrieval method and system based on Hadoop platform |
| CN105589974B (en) * | 2016-02-04 | 2019-05-17 | 通号通信信息集团有限公司 | Monitor video searching method and system based on Hadoop platform |
| CN105718597A (en) * | 2016-03-04 | 2016-06-29 | 北京邮电大学 | Data retrieving method and system thereof |
| CN105843930A (en) * | 2016-03-29 | 2016-08-10 | 乐视控股（北京）有限公司 | Video search method and device |
| CN106126590A (en) * | 2016-06-20 | 2016-11-16 | 北京航空航天大学 | A kind of unmanned plane based on key message scouts video retrieval method |
| CN110019951A (en) * | 2017-09-29 | 2019-07-16 | 华为软件技术有限公司 | A kind of method and apparatus generating video thumbnails |
| CN110019951B (en) * | 2017-09-29 | 2021-06-04 | 华为软件技术有限公司 | Method and equipment for generating video thumbnail |
| CN111602411A (en) * | 2018-02-27 | 2020-08-28 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and program |
| CN111602411B (en) * | 2018-02-27 | 2022-07-22 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and storage medium |
| US11422914B2 (en) | 2018-02-27 | 2022-08-23 | Omron Corporation | Metadata generation apparatus, metadata generation method, and program |
| CN110012349A (en) * | 2019-06-04 | 2019-07-12 | 成都索贝数码科技股份有限公司 | A kind of news program structural method and its structuring frame system end to end |
| CN112486923A (en) * | 2020-12-11 | 2021-03-12 | 北京梧桐车联科技有限责任公司 | File updating method, device and system |
| CN114297441A (en) * | 2021-12-08 | 2022-04-08 | 北京格灵深瞳信息技术股份有限公司 | Video object retrieval method and device, computer equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101369281A (en) | Retrieval method based on video abstract metadata | |
| WO2021223567A1 (en) | Content processing method and apparatus, computer device, and storage medium | |
| RU2628192C2 (en) | Device for semantic classification and search in archives of digitized film materials | |
| CN112052387B (en) | Content recommendation method, device and computer readable storage medium | |
| CN113395578B (en) | Method, device, equipment and storage medium for extracting video theme text | |
| WO2022184117A1 (en) | Deep learning-based video clipping method, related device, and storage medium | |
| CN107832724A (en) | The method and device of personage's key frame is extracted from video file | |
| CN102902826B (en) | A kind of image method for quickly retrieving based on reference picture index | |
| CN104021140B (en) | A kind of processing method and processing device of Internet video | |
| CN109408672B (en) | Article generation method, article generation device, server and storage medium | |
| CN116975615A (en) | Task prediction method and device based on video multi-mode information | |
| CN109308324A (en) | An image retrieval method and system based on hand-drawn style recommendation | |
| CN116935170B (en) | Processing method and device of video processing model, computer equipment and storage medium | |
| CN114550070A (en) | Video clip identification method, device, equipment and storage medium | |
| US20190082236A1 (en) | Determining Representative Content to be Used in Representing a Video | |
| CN102236714A (en) | Extensible markup language (XML)-based interactive application multimedia information retrieval method | |
| CN113254696B (en) | Cover image acquisition method and device | |
| CN103984778A (en) | Video retrieval method and video retrieval system | |
| CN116010711A (en) | KGCN model movie recommendation method integrating user information and interest change | |
| Fei et al. | Learning user interest with improved triplet deep ranking and web-image priors for topic-related video summarization | |
| WO2025015804A1 (en) | Method and apparatus for recommending media data, and computer device and storage medium | |
| CN115171014B (en) | Video processing method, video processing device, electronic equipment and computer readable storage medium | |
| CN116977992A (en) | Text information identification method, apparatus, computer device and storage medium | |
| Hu et al. | Video summarization via exploring the global and local importance | |
| CN113821676A (en) | Video retrieval method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20090218 |