JP2004159107A - Method for index generation, program, and storage medium with program stored therein - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate indexes for temporally associating a scenario with an edited video with small workloads after completion of the video for effectively utilizing a scenario document at the time of planing in association with the edited video. <P>SOLUTION: An actual time length of the whole video is acquired from the edited video, predicted time lengths at each scene, and the whole video from the scenario are calculated, a ratio between the actual time length of the whole edited video and the predictive time length of the whole video calculated from the scenario is acquired, the predictive time lengths at each scene calculated from the scenario are corrected by using the ratio, and the indexes describing prediction start times, prediction end times and the predictive time lengths are generated at each scene. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an index generation method, a program, and a storage medium storing an index generation program, and more particularly, to an index generation method, a program, and an index generation program for generating an index of a video by associating an edited video with a scenario. And a storage medium storing the information.
[0002]
[Prior art]
Considering the workflow of video production, usually a scenario is first created in which the scenes are combined to create a video, and the final video (edited video) is created by performing shooting, editing, etc. based on the scenario. It becomes the flow of making.
[0003]
Here, a scenario is a document that describes the planning composition of a video for each scene. For example, information such as the length of a scene, title, sketch, narration, telop, BGM, composition, camera work, performers, and shooting location Is described. It is not necessary that all of this information be described, and the format to be described is not particularly specified. Some are handwritten, others are electronic files created with a word processor or special software, and some are printed.
[0004]
In general, the edited video has a vague temporal correspondence with the original scenario. In some cases, there is no description about the time in the scenario, and in other cases, the time length of each scene is described.However, the time length of the scenario and the time length of the edited video may be changed during the editing process. They do not always match. In this situation, the temporal correspondence between the scenario information and the edited video is ambiguous, and it is not clear which part of the video the metadata described in the scenario corresponds to. It is difficult to add as metadata.
[0005]
As one means for solving this, there is a method of temporally associating a scenario with an edited video by giving an identifier indicating a position in a structure at the time of video production (for example, see Patent Document 1). According to this method, the scenario and the edited video can be temporally associated, and the scenario information can be used as metadata.
[0006]
As another approach, there is a method in which information is extracted from edited video by telop recognition (for example, see Patent Literature 2) or voice recognition (for example, see Patent Literature 3), and is added as metadata.
[0007]
[Patent Document 1]
Japanese Patent Application Laid-Open No. 2000-92419 "Program information management / editing system and hierarchical program information storage / management device used therein"
[Patent Document 2]
Japanese Patent Application Laid-Open No. 2002-279433 "Method and apparatus for searching characters in video"
[Patent Document 3]
JP 2002-175304 A "Video Search Apparatus and Method"
[0008]
[Problems to be solved by the invention]
However, in order to associate the edited video with the scenario using the above “program information management and editing system and the hierarchical program information storage and management device used for the same”, an identifier must be used each time a shot is taken by a camera during video production. Therefore, even if it is necessary to perform the association after the video is completed, there is a problem that the scenario information cannot be associated unless the identifier is assigned at the shooting stage. In addition, at all stages of planning, photographing, and editing, there is a problem that equipment for giving and recording an identifier is required, and the introduction cost is high.
[0009]
Of course, it is conceivable to manually associate the edited video with the scenario, but manually associating all enormous scenes included in the video has a problem that the time cost is high.
[0010]
Furthermore, in the method of extracting information from the edited video by telop recognition or voice recognition and adding it as metadata, the information that can be extracted is only a part of the information included in the scenario, and the recognition accuracy is 100%. Absent.
[0011]
The present invention has been made in view of the above points, and in order to effectively utilize a scenario document at the time of planning in association with an edited video, the scenario and the edited video can be temporally reduced with a small amount of work after the video is completed. An object of the present invention is to provide an index generation method and a program for associating, and a storage medium storing the index generation program.
[0012]
Here, the index describes metadata such as a predicted start time, a predicted end time, a predicted time length, and a scene title for each scene in the scenario description.
[0013]
[Means for Solving the Problems]
FIG. 1 is a diagram for explaining the principle of the present invention.
[0014]
The present invention provides an index generation method for generating an index for a video that associates a video and a scenario based on an edited video and a scenario document,
A real-time length measuring process for obtaining the real-time length of the entire video from the edited video (step 1)
A prediction time calculation process (step 2) for calculating a prediction time length for each scene and a prediction time length of the entire video from the scenario;
A time length comparing step (step 3) for obtaining a ratio between the actual time length of the entire edited video and the predicted time length of the entire video calculated from the scenario;
A predicted time length correcting step (step 4) for correcting the predicted time length for each scene calculated from the scenario using the ratio;
An index generation step (step 5) for generating an index describing the predicted start time, predicted end time, and predicted time length for each scene is performed.
[0015]
Further, in the process of calculating the predicted time of the present invention,
Count all or part of the narration, comment, and dialogue for each scene described in the scenario,
Find the ratio of the number of characters for each scene,
The process of calculating the predicted time length for each scene using the ratio is further performed.
[0016]
In addition, the present invention, after the end of the index generation process,
Detect the cut point and the time of the cut point from the edited video,
By comparing the time of the detected cut point to be a correction candidate and the predicted start time of each scene calculated in the predicted time calculation process, any time of the cut point is set as the corrected predicted start time,
A process of generating an index describing time information obtained by correcting the predicted start time and the predicted time length for each scene calculated from the scenario is further performed.
[0017]
In addition, the present invention, after the end of the index generation process,
Extract metadata for each scene from the scenario,
A process of generating an index describing the extracted metadata for each scene is further performed.
[0018]
The present invention is an index generation program for generating an index for a video that associates a video with a scenario based on an edited video and a scenario document,
A real time length measuring step for obtaining the real time length of the entire video from the edited video,
Calculate the predicted time length of each scene and the predicted time length of the entire video from the scenario, and if there is no description about time in the scenario, narration or comment for each scene described in the scenario, all or dialogue, A predicted time calculation step of counting a number of characters and calculating a ratio of the number of characters for each scene;
A predicted time calculation step of calculating a predicted time length for each scene using the ratio,
A time length comparing step of calculating a ratio of a real time length of the entire edited video to a predicted time length of the entire video calculated from the scenario;
A predicted time length correction step of correcting the predicted time length for each scene calculated from the scenario using the ratio,
An index generation step of generating an index describing a prediction start time, a prediction end time, and a prediction time length for each scene;
A cut point detection step of detecting a cut point and the time of the cut point from the edited video;
A correction prediction start that compares the time of the detected cut point that is a correction candidate with the prediction start time of each scene calculated in the prediction time calculation step and sets any of the cut points as the corrected prediction start time. A time setting step;
A modified index generation step of generating an index describing time information in which the predicted start time and the predicted time length of each scene calculated from the scenario are corrected,
A metadata extraction step of extracting metadata for each scene from the scenario,
Generating an index in which the extracted metadata is described for each scene to generate an index.
[0019]
The present invention is a storage medium storing an index generation program for generating an index for a video that associates a video with a scenario based on a edited video and a scenario document,
A real time length measuring step for obtaining the real time length of the entire video from the edited video,
Calculate the predicted time length of each scene and the predicted time length of the entire video from the scenario, and if there is no description about time in the scenario, narration or comment for each scene described in the scenario, all or dialogue, A predicted time calculation step of counting a number of characters and calculating a ratio of the number of characters for each scene;
A predicted time calculation step of calculating a predicted time length for each scene using the ratio,
A time length comparing step of calculating a ratio of a real time length of the entire edited video to a predicted time length of the entire video calculated from the scenario;
A predicted time length correction step of correcting the predicted time length for each scene calculated from the scenario using the ratio,
An index generation step of generating an index describing a prediction start time, a prediction end time, and a prediction time length for each scene;
A cut point detection step of detecting a cut point and the time of the cut point from the edited video;
A correction prediction start that compares the time of the detected cut point that is a correction candidate with the prediction start time of each scene calculated in the prediction time calculation step and sets any of the cut points as the corrected prediction start time. A time setting step;
A modified index generation step of generating an index describing time information in which the predicted start time and the predicted time length of each scene calculated from the scenario are corrected,
A metadata extraction step of extracting metadata for each scene from the scenario,
And a metadata generation index generation step of generating an index in which the extracted metadata is described for each scene.
[0020]
As described above, in the present invention, when the operator plays the edited video and labels the start point of each scene through the entire video, the start point exists by the index generated by the present invention. Since it is possible to narrow down the sections, it is not necessary to look through all of the images, and labor can be saved.
[0021]
Further, according to the present invention, an index can be generated even when the scheduled start time of a scenario scene is not described.
[0022]
Furthermore, by using the edited cut point at the time of index generation, it becomes possible to correct the scheduled start time of the scene in the index to more accurate information.
[0023]
Also, by describing not only the information on the start time of the scene but also the information on the scenario as metadata in the index, it is possible to perform a scene search using a keyword.
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0024]
[First Embodiment]
FIG. 2 is a flowchart of index generation according to the first embodiment of the present invention.
[0025]
The flowchart shown in the figure includes a real time length measuring step (step 101), a predicted time length calculating step (step 102), a time length comparing step (step 103), a predicted time length correcting step (step 104), an index generation step ( Step 105).
[0026]
Hereinafter, the index creation operation will be described in detail with reference to FIG.
[0027]
In the present invention, since there is a difference between the time length of the video to be completed described in the scenario and the time length of the actual edited video, an object is to generate an index roughly correcting the difference. .
[0028]
First, the input of the real time length measurement step (step 101) is an already edited video. The genre, content, editing method, and recording medium of the video are not particularly limited.
[0029]
In step 101, a real time length L1 of the entire video is obtained from the input edited video. For example, if the edited video is recorded on a tape such as a VHS, the time length is obtained from the difference between the start value and the end value of the counter using a playback deck or the like. If the video is recorded in an electronic format such as MPEG, the time length of the video can be obtained from the properties of the file. Here, it is assumed that the real time length L1 = 200 (s) of the edited video is obtained as the output of step 101, for example.
[0030]
Next, the input of step 102 is the scenario used when producing the edited video. When implemented as a computer program, it is assumed that the scenarios are in a machine-readable format. For example, assume that a scenario in which a scene time length 201, a title 202, a sketch 203, a narration 204, a telop 205, and a BGM 206 are described as shown in FIG.
[0031]
From this scenario, the predicted time lengths S1 to Sn of the scenes (1 to n) are calculated. When the time length 201 is described in advance in each scene as in the scenario of FIG. 3, the value is directly used as the predicted time length S1 to Sn of the edited video. For example, if the time length of scene i is described as “0:20” in the scenario, the predicted time length Si of scene i is set to “Si = 20 (s)”. By the way, in some scenarios, a start time for each scene is described instead of describing a time length for each scene. In this case, the predicted time lengths S1 to Sn of the scene i are
Si = Ti + 1-Ti
Ask by Here, Ti is the start time of scene i.
[0032]
Further, the entire predicted time length L2 is
L2 = S1 + S2 +... + Sn
Is calculated by
[0033]
In the case of the scenario of FIG. 3, the predicted time length of each scene is
S1 = 90 (s),
S2 = 50 (s),
S3 = 40 (s)
And the overall prediction time length is
L2 = S1 + S2 + S3 = 180 (s)
It becomes.
[0034]
The input in step 103 is the real time length L1 of the edited video obtained in step 101 and the predicted time length L2 obtained in step 102.
[0035]
In step 103, a ratio R = L1 / L2 of the entire actual time length L1 calculated from the edited video and the entire predicted time length L2 calculated from the scenario is obtained.
[0036]
In case of the scenario in Fig. 3
R = L1 / L2 = 200/180 = 10/9
It becomes.
[0037]
In step 104, the predicted time lengths S1 to Sn of each scene obtained in step 102 are corrected using the ratio R obtained in step 103. The corrected predicted time lengths S′1 to S′n are
S′1 = S′1 * R, S′2 = S′2 * R,..., S′n = S′n * R
It is obtained by the calculation formula.
[0038]
In the case of the scenario shown in FIG. 3, the estimated time length of each corrected scene is
S′1 = S1 * R = 100 (s),
S′2 = S2 * R ≒ 56 (s),
S′3 = S3 * R ≒ 44 (s)
It becomes.
[0039]
In step 105, each scene number 301, prediction start time is used as shown in FIG. 4 using the corrected predicted time lengths S ′ 1 to S′n for each scene and the real time length L 1 of the video obtained in step 104. An index describing 302, a predicted end time 303, a predicted time length 304, a start time 305, an end time 306, and a time length 307 of the entire video is generated.
[0040]
Here, the prediction start time 302 of the scene 1 and the start time 305 of the entire video are always set to “0:00”, and the corrected prediction time length 304 is added. The prediction start time 302 of the scene is sequentially calculated.
[0041]
The index in FIG. 4 indicates that scene 1 exists at 0:00 to 1:40, scene 2 exists at 1:40 to 2:36, and scene 3 exists at 2:36 to 3:20.
[0042]
Although the predicted start time and predicted time length of the scene of the index generated in step 105 are not always accurate, the scene start time and actual scene start time of the edited video are compared with the start time and time length of the scene described in the scenario. It is likely that the time length is approaching.
[0043]
The predicted start time of the index generated here usually has an error compared to the start time of the edited video. Therefore, taking this error into account, an index may be generated that describes a range that is a measure of the actual start time of the scene. For example, assuming that there is an error of about the difference between before and after the correction of the prediction time length, the index of FIG. For example, assuming that the difference between the predicted time length of the scene i before and after the correction and the predicted time length after the correction is di, and the prediction end time is Ti, the predicted end time of the scene i in the range expression is (Ti-di) to (Ti + di) It can be expressed as. Then, the prediction time length and the prediction start time of the next scene are also corrected according to the range of the prediction end time. At this time, the range of the prediction end time is set as the range of the prediction start time in the next scene, and the minimum and maximum values in the range of the prediction start time and the prediction end time are used as the range of the prediction time length. However, since the predicted start time of the first scene and the predicted end time of the last scene are accurate values calculated from the edited video, there is no need to perform range display.
[0044]
For example, the predicted time length of scene 1 before correction is 90 (s), and the predicted time length after correction is 100 (s), so that the difference is 10 (s). Then, since the predicted end time of scene 1 is 1:40, ± 10 (s) is calculated from this value, and 1:30 to 1:50 is a standard range of the predicted end time. At this time, since the start time of scene 1 is 0:00, when the minimum value and the maximum value of this section are obtained, the predicted time length of scene 1 is 90 to 110 s. By repeating this procedure, the index shown in FIG. 5 is created.
[0045]
It can be expected that the start time information of the scene described in the index generated according to the present embodiment has a value close to the start time of each scene of the edited video. Therefore, even if the operator does not look at the entire video in order to obtain the correct scene start time from the edited video, the scene start time is described before or after or described in the index generated according to the present embodiment. It is sufficient to search around the range, and labor can be saved.
[0046]
Also, no special device or procedure is required for video shooting for index generation.
[0047]
[Second embodiment]
In the present embodiment, an index generation method in a case where there is no description about time in a scenario will be described.
[0048]
FIG. 6 is an example of a scenario according to the second embodiment of the present invention, and shows an example in which there is no description about time in the scenario. FIG. 7 is a flowchart of index generation according to the second embodiment of the present invention.
[0049]
Steps 201 and 205 in FIG. 7 are the same as steps 101 and 105 of the above-described first embodiment, and a description thereof will be omitted.
[0050]
For example, a scenario having no description about time as shown in FIG. In step 202, the number of characters described for each scene in this scenario is counted. Here, the number of characters in scene i is ri.
[0051]
In the scenario of FIG. 6, it is assumed that the narration described in scene 1 has 856 characters and the narration described in scene 2 has 632 characters. In the present embodiment, of the narration, comments, dialogues, etc. described in the scenario, the number of characters of the narration is counted and defined as ri. May be the number of characters of ri. If there is a scene in which no characters such as narration, comments, dialogue, etc. are described in the scenario, it is assumed that the scene has a specified number of characters or time length, and a predetermined appropriate value a is defined as ri. I do. For example, for a, the average number of characters for each scene is used.
[0052]
In step 203, the ratio r1: r2: ...: rn for each scene is calculated based on the number of characters obtained in step 202. In the example of the scenario of FIG. 6, r1: r2 = 856: 632.
[0053]
In step 204, the predicted time lengths S1 to Sn of the respective scenes are calculated from the real time length L1 of the edited video obtained in step 201 and the ratios r1: r2: ... rn obtained in step 203. The predicted time length of each scene is as follows.
[0054]
S1 = L1 * r1 / (r1 + r2 +... + Rn)
S2 = L1 * r2 / (r1 + r2 + ... + rn)
…
Sn = L1 * rn / (r1 + r2 + ... + rn)
When the real time length of the edited video obtained in step 201 is 300 (s), in the example of FIG. 6, the predicted time lengths S1 and S2 of scene 1 and scene 2 are:
S1 = 300 * 856 / (856 + 632) ≒ 173 (s),
S2 = 300 * 632 / (856 + 632) ≒ 127 (s),
Is calculated.
[0055]
In the present embodiment, even when there is no description of time information in the scenario, it is possible to generate an index describing a rough time, and the same effect as in the first embodiment can be obtained.
[0056]
Further, in the present embodiment, the number of characters is counted, but the number of cuts included in each scene can be used.
[0057]
[Third Embodiment]
In the present embodiment, a method for correcting the predicted start time of a scene using the time information of the cut point of the edited video will be described.
[0058]
FIG. 8 is a flowchart of a method for correcting the predicted time of a scene according to the third embodiment of the present invention.
[0059]
Here, the cut point is a joint between shots (video sections continuously captured by a camera). These cut points are candidates for scene and scene breaks. However, not all cut points are scene breaks. In FIG. 8, steps 301 to 305 are the same as steps 101 to 105 of the above-described first embodiment, and a description thereof will not be repeated. This embodiment can also be realized by the procedure of steps 201 to 205 of the second embodiment instead of steps 301 to 305.
[0060]
The input in step 306 is an edited video. In step 306, a cut point is obtained from the edited video based on the feature amount of the video. As a method therefor, a cut point is detected along with the time using an existing technique such as “Japanese Patent Application Laid-Open No. 2002-218376“ Recording medium storing a program of a cut detection device and a cut detection method ””.
[0061]
FIG. 9 shows a state in which a cut point is detected from an image according to the third embodiment of the present invention. The figure shows a diagram 801 schematically showing the temporal arrangement of video frames, and then shows frames 802, 803, and 804 at the detected cut points. In the example shown in the figure, the times of the three detected cut points are 1:33 (frame 802), 2:05 (frame 803), and 2:40 (frame 804).
[0062]
In step 307, the cut start point detected in step 306 is compared with the predicted start time of each scene of the index generated in step 305.
[0063]
For example, in the case of the example of the index in FIG. 4 generated in step 305, the predicted start time of the scene is “00:00, 01:40, 02:36”.
[0064]
Since the exact start time of the scene is likely to be one of the cut points detected in step 306, the predicted start time of this index is compared with the cut point detection time to correct the predicted start time. As a correction method, the detection time of the cut point closest to the prediction start time of the scene is set as the corrected scene prediction start time.
[0065]
FIG. 10 shows a modification of the predicted start time according to the third embodiment of the present invention. In the figure, the correspondence between the predicted start time of the scene and the time of the cut point of the edited video is shown, and the edited video is represented by a timeline 901, 905, and the detected cut points 902, 903, 904. , Predicted start times 906 and 907 obtained from the scenario. Using the cut point closest to the scene prediction start time as shown in FIG. 10, 1:44 (906) is 1:33 (902), and 2:36 (907) is 2:40 (904). Will be corrected.
[0066]
However, when detecting a cut point, there is a possibility that a detection omission has occurred. Therefore, if the interval between the predicted start time of the scene and the time of the closest cut point is equal to or longer than a predetermined time length, the correction may not be performed.
[0067]
Further, since the closest cut point is not always the start point of the correct scene, the operator may visually determine and select a corresponding cut point from cut points existing in the range D before and after the predicted start time. (D is a predetermined time length). Alternatively, it may be similarly selected from M cut points existing before and after the predicted start time (M is a predetermined number).
[0068]
In step 308, an index is generated using the predicted start time for each scene corrected in step 307. The generated index is as shown in FIG.
[0069]
According to the present embodiment, a cut point is detected from an edited video, and compared with the predicted start time for each scene of the index generated in the above-described first or second embodiment. , The prediction time information in the index can be made more accurate.
[0070]
[Fourth Embodiment]
In the present embodiment, in the index generation, not only the time but also a keyword having a high frequency of appearance for each scene, a keyword included in a title, a position of a person, a composition, BGM, and the like are extracted from a scenario, and a meta-data for the scene is extracted. An example of adding data (attached information) will be described.
[0071]
FIG. 12 is a flowchart of index generation according to the fourth embodiment of the present invention. Steps 401 to 405 are the same as steps 101 to 105 of the first embodiment, and thus description thereof is omitted. Further, in the present embodiment, instead of steps 401 to 405, the procedure of steps 201 to 205 in the second embodiment or steps 301 to 305 of the third embodiment may be used.
[0072]
In step 406, metadata is extracted from the scenario. Examples of the metadata include a title, a performer name, the number of performers, a keyword included in a narration or comment, a used BGM, a telop, a composition, and the like.
[0073]
For information such as title, performer name, number of performers, BGM used, telop, structure, etc., the operator transfers information described in the scenario to the index for each item. Items not described in the scenario are left blank in the index. Index items may be added as necessary, or only a part thereof may be used.
[0074]
With regard to the keyword extraction method, for example, the narration text is cut into word units using existing technology such as “Japanese Patent Application Laid-Open No. 1996-95982”, and all words are used as keywords as they are. The frequency of appearance of words may be counted, and the top 10 words may be selected from those having a high frequency of appearance, or the operator may manually extract keywords from the scenario.
[0075]
Here, it is desirable that the keywords described in the index are not included in other scenes as much as possible. For this purpose, the appearance frequency of each word included in the scenario is counted for each scene, and if the same word is included in the top 10 appearance frequencies of different scenes, the common word is excluded from the keywords. Alternatively, if the same word is included in more than a predetermined number of scenes, the word is excluded from the keywords. In this manner, metadata that expresses the features of each scene is obtained by deleting common keywords or making them less likely to become keywords.
[0076]
For example, in step 406, the operator extracts the keywords “first and last name” and “4.6 billion years” from scene 1 of the scenario in FIG. 3, “evolution” from scene 2, and “humanity” from scene 3 to obtain the title, telop, and BGM. If the information described in the scenario is transcribed, the index generated in step 407 is as shown in FIG.
[0077]
When an index to which such metadata is added is generated, in addition to the effects of the first, second, and third embodiments described above, the operator performs a scene search using a keyword and sets the predicted start time of the scene. Based on this, it is possible to find the corresponding video section.
[0078]
It should be noted that the flowcharts shown in FIGS. 2, 7, 8, and 12 in the above embodiment can be constructed as a program, installed on a computer used as an index generation device, and executed by a control unit such as a CPU. It is possible.
[0079]
Further, the constructed program is stored in a hard disk device connected to a computer used as an index generation device, or in a portable storage medium such as a flexible disk or a CD-ROM. It is also possible to install it.
[0080]
It should be noted that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible within the scope of the claims.
[0081]
【The invention's effect】
As described above, according to the present invention, the start time information of the scene described in the generated index can be expected to be a value close to the start time of each scene of the edited video. Therefore, in order to obtain the correct scene start time from the edited video, the operator only has to search for the scene start time described before and after the scene described in the generated index, without having to look at the entire video, thereby saving the operator's work. can do.
[0082]
In the present invention, it is possible to generate an index even when the scenario does not include information on the time for each scene. In addition, by using the time information of the cut point of the edited video, the index can be generated. It is possible to make the scene start time information close to an accurate value.
[0083]
Further, when the operator wants to search for a scene related to a certain keyword, the target scene can be easily found by referring to the start time information of the scene including the search keyword from the generated index.
[0084]
According to the present invention, no special device or procedure is required at the time of capturing an image to implement the above means.
[0085]
In addition, since an index that describes the start point of a structural unit (scene) whose content is meaningful based on a scenario, rather than a structural unit such as a signal feature of an image or an audio, is generated, images of the start point of the scene are arranged. This makes it possible to grasp the structure of the entire image and the outline of the image.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining the principle of the present invention.
FIG. 2 is a flowchart of index generation according to the first embodiment of the present invention.
FIG. 3 is an example of an input scenario according to the first embodiment of the present invention.
FIG. 4 is an example of an index generated in the first embodiment of the present invention.
FIG. 5 is an example of generating an index in consideration of an error according to the first embodiment of the present invention.
FIG. 6 is an example of a scenario according to the second embodiment of the present invention.
FIG. 7 is a flowchart of index generation according to the second embodiment of the present invention.
FIG. 8 is a flowchart of a method for correcting a predicted scene time according to the second embodiment of the present invention.
FIG. 9 is a diagram illustrating a state where a cut point is detected from an image according to the third embodiment of the present invention.
FIG. 10 is a modified example of the estimated start time according to the third embodiment of the present invention.
FIG. 11 is an example of a generated index according to the third embodiment of the present invention.
FIG. 12 is a flowchart of index generation according to the fourth embodiment of the present invention.
FIG. 13 is an example of a generated index according to the fourth embodiment of the present invention.
[Explanation of symbols]
201 Scene Length
202 title
203 Sketch
204 Narration
205 ticker
206 BGM
301 each scene number
302 Predicted start time
303 Predicted end time
304 Prediction time length
305 Start time of entire video
306 end time
307 hours long
801 Temporal arrangement of video frames
802, 803, 804 Frame at detected cut point

Claims (6)

An index generation method for generating an index for a video that associates a video with a scenario based on an edited video and a scenario document,
A real-time length measuring step of obtaining a real-time length of the entire video from the edited video,
A predicted time calculation step of calculating a predicted time length for each scene and a predicted time length of the entire video from the scenario,
A time length comparing step of calculating a ratio of a real time length of the entire edited video and a predicted time length of the entire video calculated from the scenario,
A predicted time length correction step of correcting the predicted time length for each scene calculated from the scenario using the ratio,
An index generating step of generating an index describing a predicted start time, a predicted end time, and a predicted time length for each scene. In the predicted time calculation process,
Narration and comment for each scene described in the scenario, count all or part of the characters of the dialogue,
Find the ratio of the number of characters for each scene,
2. The index generation method according to claim 1, further comprising calculating a predicted time length for each scene using the ratio. After completion of the index generation process,
Detecting a cut point and the time of the cut point from the edited video,
Comparing the time of the detected cut point to be a correction candidate and the predicted start time of each scene calculated in the predicted time calculation process, any of the cut points is used as the corrected predicted start time,
3. The index generation method according to claim 1, further comprising the step of generating an index describing time information obtained by correcting the prediction start time and the prediction time length for each scene calculated from the scenario. After completion of the index generation process,
Extracting metadata for each scene from the scenario,
4. The index generation method according to claim 1, further comprising a step of generating an index in which the extracted metadata is described for each scene. An index generation program for generating an index for a video that associates a video with a scenario based on a edited video and a scenario document,
A real time length measuring step of obtaining a real time length of the entire video from the edited video,
The predicted time length of each scene and the predicted time length of the entire video are calculated from the scenario, and if there is no description about time in the scenario, all of the narration, comment, and dialog for each scene described in the scenario or A predicted time calculation step of counting the number of partial characters and calculating a ratio of the number of characters for each scene;
A predicted time calculation step of calculating a predicted time length for each scene using the ratio,
A time length comparing step of determining a ratio of a real time length of the entire edited video and a predicted time length of the entire video calculated from the scenario,
A predicted time length correction step of correcting the predicted time length for each scene calculated from the scenario using the ratio,
An index generation step of generating an index describing a prediction start time, a prediction end time, and a prediction time length for each scene;
A cut point detection step of detecting a cut point and the time of the cut point from the edited video,
The time of the detected cut point that is a correction candidate is compared with the predicted start time of each scene calculated in the predicted time calculation step, and any one of the cut points is set as the corrected predicted start time. A start time setting step;
A corrected index generation step of generating an index describing time information obtained by correcting the predicted start time and the predicted time length for each scene calculated from the scenario,
A metadata extraction step of extracting metadata for each scene from the scenario,
An index generation step of generating an index describing the extracted metadata for each scene. A storage medium storing an index generation program for generating an index for a video that associates a video with a scenario based on a edited video and a scenario document,
A real time length measuring step of obtaining a real time length of the entire video from the edited video,
The predicted time length of each scene and the predicted time length of the entire video are calculated from the scenario, and if there is no description about time in the scenario, all of the narration, comment, and dialog for each scene described in the scenario or A predicted time calculation step of counting the number of partial characters and calculating a ratio of the number of characters for each scene;
A predicted time calculation step of calculating a predicted time length for each scene using the ratio,
A time length comparing step of determining a ratio of a real time length of the entire edited video and a predicted time length of the entire video calculated from the scenario,
A predicted time length correction step of correcting the predicted time length for each scene calculated from the scenario using the ratio,
An index generation step of generating an index describing a prediction start time, a prediction end time, and a prediction time length for each scene;
A cut point detection step of detecting a cut point and the time of the cut point from the edited video,
The time of the detected cut point that is a correction candidate is compared with the predicted start time of each scene calculated in the predicted time calculation step, and any one of the cut points is set as the corrected predicted start time. A start time setting step;
A corrected index generation step of generating an index describing time information obtained by correcting the predicted start time and the predicted time length for each scene calculated from the scenario,
A metadata extraction step of extracting metadata for each scene from the scenario,
A storage medium storing an index generation program characterized by storing a program including: a metadata-added index generation step of generating an index in which extracted metadata is described for each scene.

Images