JP2008236729A - Method and apparatus for generating digest - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To optimally classify a plurality of moving-image files obtained by photographing into scenes, and to generate a digest extracting an optimal portion for each genre. <P>SOLUTION: A scene classifying section 42 classifies a plurality of moving-image files classified as the same event by an event classifying section 41 for each scene in accordance with a threshold predetermined correspondingly to the genre of the event on the basis of tags 22. A digest generating section 43 extracts a part of the moving-image files in accordance with a digest generation technique corresponding to the genre of the event for each moving-image file classified to the same event on the basis of the tags 22 and generates an image linking a plurality of portions extracted, respectively, as a digest. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a digest generation apparatus and a digest generation method for generating a digest that is a summary of a moving image file captured by a video camera, a digital camera, and a moving image capturing apparatus such as a mobile phone having a moving image capturing function.

  There is a method in which captured images are divided into a plurality of groups on the basis of the shooting date interval, and representative images in each group are determined (see, for example, Patent Document 1).

JP 2004-295231 A

  However, the above method has a problem in that an optimum representative image cannot be extracted for each genre because the photographed images are divided into a plurality of groups based on the photographing date interval regardless of the genre of the photographed image.

  In particular, in the case of a moving image file shot by a moving image shooting device such as a video camera, there is a problem that the representative portion or characteristic portion included in the moving image file differs depending on the genre.

  Therefore, the present invention provides a digest creation apparatus and digest generation method capable of generating a digest that optimally classifies a plurality of moving image files obtained by shooting and extracts an optimal part for each genre. For the purpose.

  In order to achieve the above object, the digest generating apparatus or digest generating method of the present invention, when generating a digest from a plurality of moving image files obtained by shooting, each of the moving image files attached to the plurality of moving image files. The plurality of video files are classified for each event based on the event identification information of a tag in which genre information indicating an event genre and event identification information for identifying the event from each other are described, and for each event The plurality of classified video files are classified for each scene based on the shooting interval between the video files, and a plurality of classified for each scene by the digest generation method based on the genre information described in the tag. A part of the video file is extracted and generated as a digest.

  Here, when a plurality of video files classified for each event are classified for each scene based on the shooting interval between the video files, a threshold is set in advance for each genre, and the video files are classified for each event. Based on the genre information described in the tags of the plurality of moving image files, the threshold value is set according to the genre of the plurality of moving image files, and based on the set threshold value and the shooting interval between the moving image files. The threshold value may be classified for each scene, the maximum value or the minimum value of the threshold value is set in advance for each genre, the threshold value is determined based on the attribute information of the video file constituting each event, Based on the genre information described in the tag of the plurality of video files classified for each event, the most suitable for the genre of the plurality of video files. Comparing the value or minimum value with the threshold value determined based on the attribute information, and if the threshold value determined based on the attribute information is greater than or equal to the maximum value or less than the minimum value, while selecting the maximum value or minimum value, When the threshold value determined based on the attribute information is smaller than the maximum value or larger than the minimum value, the threshold value determined based on the attribute information is selected, and based on the selected threshold value or maximum value and the shooting interval between the video files. You may make it classify | categorize for every scene. Further, the threshold value is determined based on the attribute information of the moving image file constituting each event, and the threshold value determined based on the attribute information is based on a predetermined coefficient only when the genre of each event is a specific genre. May be adjusted so as not to be limited by the maximum or minimum threshold value for each genre. The attribute information may be the length of the event, the average of the shooting intervals of the moving image files constituting the event, the position information indicating the shooting positions of the moving image files constituting the event, or a combination thereof. There may be.

  According to the digest generation apparatus and method of the present invention, a tag in which genre information indicating an event genre of each moving image file assigned to each moving image file and event identification information for identifying the event from each other is described. Based on event identification information, multiple video files are classified for each event, multiple video files classified for each event are classified for each scene based on the shooting interval between video files, and digest generation based on tag genre information Since a part of a plurality of video files classified by scene is extracted and generated as a digest, the digest that optimally classifies a plurality of video files and extracts the optimal part for each genre is extracted. Can be generated.

Embodiment 1. FIG.
Next, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows a block diagram of an embodiment of a digest generating apparatus according to the present invention. The digest generation apparatus according to the present embodiment is applied to a video camera. As shown in FIG. 1, a content holding unit 11, a photographing unit 31, a tag creation unit 32, an event classification unit 41, a scene classification unit 42, A digest generation unit 43 and an input instruction unit 44 are provided.

  The content holding unit 11 holds a plurality of moving image files 21 photographed by a video camera and a tag 22 attached to each moving image file. That is, the content holding unit 11 stores a plurality of moving image files 21 to which shooting dates / times or recording dates / times taken by the shooting unit 31 are added. Each video file 21 includes a genre information indicating the genre of an event recorded in each video file, that is, the type or content of an event such as an athletic meet or a trip, created by the tag creation unit 32 based on the user's operation or the like. And event identification information for identifying events from each other, that is, for example, non-overlapping numerical information that can be distinguished from each other even in the same genre, such as athletic meet 1, athletic meet 2, etc. The described tag 22 is given. It should be noted that it is also possible to attach a tag to each moving image file outside the apparatus, and in this case, the tag creation unit 32 can be omitted from the apparatus. Similarly, the content holding unit 11, the photographing unit 31, and the input instruction unit 44 may of course be provided outside the apparatus.

  Here, the term “event” refers to the contents taken during a travel or athletic meet in a broad sense, but in the present invention and this embodiment, the tag 22 assigned to each video file by the tag creation unit 32 is a common video. A unit of file. “Genre” indicates the content of an event such as “athletic meet” or “travel”. Therefore, one genre corresponds to one event, the genre of the moving image file is specified by the genre information given to each moving image file, and the events of the same genre are identified by event identification information such as numerical information that does not overlap. . The event identification information may of course be numerical information that does not overlap not only between events of the same genre but also between different types of events.

  FIG. 2 shows examples of the genre of the event. The genre includes, for example, travel / leisure, athletic meet, sports, children / pets, wedding / party, and the tag creation unit 32 assigns each event as event information.

  In the present embodiment, a scene classification threshold and a digest generation method are determined for each of these genres. Here, for example, for the category of “travel / leisure”, the scene classification threshold a and the digest generation method A, Scene classification threshold b and digest generation method B for the genre “sport”, scene classification threshold c and digest generation method C for the genre “children / pets”, and scene classification threshold d and genre “wedding / party” Digest generation method D is set. The scene classification thresholds A and D and the digest generation method will be described later.

  Returning to FIG. The event classification unit 41 classifies the moving image file for each moving image file obtained by photographing the same event based on the genre information described in the tag 22 and the event identification information. For example, a plurality of moving image files photographed at different athletic meet or trip are classified for each event based on genre information such as athletic meet 1, athletic meet 2 and travel 1 described in the tag 22 and event identification information. In addition, when an event can be identified only by event identification information such as “1”, “2”, etc. that follows the genre information, that is, when event identification information is given so as not to overlap even in events of different genres, An event can be identified only by the identification information.

  Subsequently, the scene classification unit 42 classifies the moving image files classified as the same event by the event classification unit 41 for each scene based on thresholds a to d for each genre as shown in FIG. Sort files into the same scenes.

  FIG. 3 is a flowchart showing a flow of processing in the scene classification unit 42 according to the first embodiment.

  First, the scene classification unit 42 calculates a shooting interval of scenes continuously shot from the moving image file 21 classified for each event based on the tag 22 by the event classification unit 41 (S101). This is because if this shooting interval is greater than or equal to a certain threshold, it is classified as a separate scene before and after that interval. In the subsequent steps, this threshold value is determined for each genre. Hereinafter, the moving image file may be simply abbreviated as a file.

  For example, when the genre is “travel / leisure”, it is often necessary to divide the scene into scenes for each place because there are many places to move. In general, shooting is often not performed during movement, and a certain amount of time is often spent for movement. Therefore, in this genre “travel / leisure”, the scene is divided when the recording time interval of the file is equal to or longer than a certain long threshold.

  FIG. 4 shows an example of scene classification of 13 moving image files taken at the same event such as a trip.

  In each of the 13 moving image files, a shooting start time and a shooting end time are described, and the threshold value in the “travel / leisure” event is described with these files. Considering that it takes 30 minutes or more for a general trip, the threshold “a” for scene classification in a “travel / leisure” event is set to “30 minutes”, for example. Of course, this “30 minutes” is an example, and may be one hour, two hours, or even a date unit.

  The scene classification unit 42 calculates the shooting interval from the previous file for each of the files 1 to 13 in FIG. 4, and classifies the scene when it is equal to or greater than “30 minutes” of the threshold value a. By doing so, it is classified into six scenes as shown in the rightmost column of FIG. 4 (S102 “Yes”, S103).

  Next, as the threshold value b when the genre is “athletic / sports”, in this genre, it is desirable to classify as a scene for each competition or game. In general, the interval between competitions and matches is shorter than the travel time. Therefore, scene classification is performed with the threshold value b set to, for example, “10 minutes” (S104 “Yes”, S105).

  Further, as the threshold value c when the genre is “children / pets”, the following two types of shooting contents can be considered as file groups classified into this genre. “A photograph of a subject taken for a long period of time” and “A photograph of an event such as a trip focusing on children and pets”.

  Therefore, the threshold value c is changed according to the length of the event. When the difference between the shooting start time of the first file of the event and the shooting end time of the last file is 3 days or more, the scene classification unit 42 classifies the scene at every turn of the day (S106 “Yes”, S107 “Yes”, S109). That is, if the date is the same, the scene is the same, while if the date is different, the scene is different.

  On the other hand, if the difference between the shooting start time of the first file of the event and the shooting end time of the last file is two days or less, the scene classification threshold c is set to, for example, “10 minutes” (S106 “Yes”). ", S107" No ", S108).

  Further, as the threshold value d when the genre is “wedding / party”, in this genre, it is desirable to classify scenes by changing the venue or the like. In general, it is considered that it takes 30 minutes or more for the venue to move, and the scene classification unit 42 sets the threshold d for scene classification to, for example, “30 minutes” (S110 “Yes”, S111).

  If the genre is not set, the event classification unit 41 sets the threshold value to “20 minutes”, for example (S112).

  As described above, the event classification unit 41 of the present embodiment sets a different threshold for each event genre, that is, an optimum threshold corresponding to each genre, and classifies the moving image file of each event for each scene.

  Then, the event classification unit 41 compares the threshold determined by the above method and the shooting interval between the moving image files calculated in S101, and classifies the images as the same scene when the shooting interval is less than the threshold. If the shooting interval is greater than or equal to the threshold, the file is classified as a different scene before and after that (S113).

  Returning to FIG. 1 again, the digest generation unit 43 generates a digest from a set of moving image files classified into scenes by the event classification unit 41.

  FIG. 5 is a flowchart showing the processing procedure of the digest generation method by the digest generation unit 43.

  First, the digest generation unit 43 determines the length of the digest to be generated (S301). The length of the digest to be generated may be automatically determined such as being proportional to the total shooting time of all files or the number of files, or may be input by the input instruction unit 44. In the first embodiment, for example, a digest of 90 seconds is generated. When the digest length is automatically set, the input instruction unit 44 can be omitted.

  Next, the digest generation unit 43 generates digests of all scenes, and generates a whole digest by connecting them, so that it determines how many seconds of digests to generate in each scene (S302). ). The digest length of each scene may be equally divided by the number of scenes of the digest time determined in S301, or may be distributed in proportion to the total shooting time of each scene, the number of files, and the like. Here, 90 seconds is equally divided by 3 which is the number of scenes, and extracted from each scene as a digest for 30 seconds.

  Next, the digest generation unit 43 generates a digest of each scene (S303). In the present embodiment, as shown in FIG. 2, an optimal digest that matches the genre is generated by changing the digest generation method according to the genre of the moving image file.

  In other words, in the first embodiment, as shown in FIG. 2, the threshold for classifying the moving image file of each event into a scene can be changed for each event genre, and the digest generation method is changed for each event genre. be able to.

  Next, a digest generation method for each genre will be described in detail.

  First, when the genre is “travel / leisure”, that is, when taking a picture using a video camera in a sightseeing trip or the like, it is often the case that a subject is determined in advance and the video camera is pointed at the subject before starting to shoot. For this reason, there is an image to be extracted immediately after the start of shooting.

  Therefore, the digest generation unit 43 extracts the start portion of each moving image file constituting the scene as shown in FIG. 7A by the digest generation method A shown in FIG.

  That is, when the genre is “travel / leisure”, the digest generation unit 43 uses the digest generation method A shown in FIG. 6A to display all files of the same scene, for example, six moving images in the example of FIG. The first 5 seconds are extracted from the files 1 to 6 (S1), and the sections of the first 5 seconds of all the extracted moving image files 1 to 6 are connected in order of shooting time (S2). As a result, a digest of 30 seconds is generated. Here, if the first 5 seconds of all the scenes are extracted and it is less than 30 seconds, for example, if the number of files included in the scene is less than 6, the length of extraction should be 5 seconds or more, or the center of the file A digest of 30 seconds is generated by extracting two or more sections from the above. Conversely, if the first 5 seconds of all scenes are extracted and it exceeds 30 seconds, for example, if there are more than 6 files in the scene, a file with a long shooting time or a file with a long shooting interval from the previous file is used. A 30-second digest is generated by extracting preferentially.

  Next, if the genre is “athletic / sports”, know in advance when the play you want to shoot occurs, such as “the child runs in front of the video camera in a race” or “hit a hit in baseball” Because it is difficult, it is often on standby while the video camera is in the shooting state. Therefore, a digest is generated by extracting the part after the middle of each moving image file.

  That is, when the genre is “athletic / sports”, the digest generation unit 43 first uses the digest generation method B shown in FIG. 6B to show all files in the same scene, for example, the example of FIG. 7B. Thus, 5 seconds indicated by the oblique lines at the approximate centers of the six files are extracted (S11), and then the approximately 5-second sections of all the extracted files are connected in order of photographing time (S12). In this way, the digest generating unit 43 generates a 30-second digest by connecting the 5-second sections of the six files of the same scene shown in FIG. If the center 5 seconds of all the scenes is extracted but it is less than 30 seconds, the length to be extracted is set to 5 seconds or more, or the time period of 1/3 and 2/3 from the beginning of the file Or the like, by extracting two sections, a 30-second digest is generated. Conversely, if the first 5 seconds of all scenes are extracted and it exceeds 30 seconds, a 30-second digest is generated by extracting preferentially from a file with a long shooting time or a file with a long shooting interval from the previous file. To do.

  Next, when the genre is “children / pets”, as in the case of the genre “travel / leisure”, shooting is often started after capturing the subject. Therefore, there is a video that should be extracted immediately after the start of shooting. In this genre, the subject generally moves and exists in the center of the screen. For this reason, when the central portion of the screen moves differently or has a different color compared to the surrounding area, it is highly likely that the subject exists in the center of the image. Therefore, a digest is generated by extracting a portion where the subject is likely to exist in the center of the image near the start portion of each moving image file.

  That is, when the genre is “children / pets”, the digest generation unit 43 first detects a section in which the subject exists in the central portion of the image of each file by the digest generation method C shown in FIG. S21). For example, if the subject is a child, it can be detected by detecting that the face is in the center of the screen, and whether the face is simply the skin color in the center of the screen. It is detected from the positional relationship between the skin color and the color of the hair, eyes, or mouth. Even if the subject is a pet or a face cannot be detected, the motion vector of the image is extracted, and if the motion near the center is different from the surrounding motion, it is detected that the subject is there.

  Subsequently, for each file, the digest generation unit 43 extracts the first 5 seconds in the section closest to the file start portion and in which the subject is present in the center portion of the detected image (S22), The above-mentioned 5-second sections extracted for each file are connected in order of photographing time (S23).

  In FIG. 7C, in each file, the portion where the subject exists at the center of the image is shown in gray. After that, the part where the subject is present closest to the start part and the subject exists for a certain time or more is extracted from the part where the subject exists in the center of the image. Here, 5 seconds each represented by the hatched portion in FIG. 7C are extracted and connected to generate a 30-second digest. In this way, the digest generating unit 43 generates a 30-second digest by connecting the 5-second sections of the six files of the same scene shown in FIG. In this case as well, if a 5-second section is extracted from all scenes and is not less than 30 seconds, the length to be extracted is set to 5 seconds or more, or two or more sections in which one subject exists from one file. By extracting, a digest of 30 seconds is generated. On the other hand, if the first 5 seconds of all scenes are extracted, if it exceeds 30 seconds, the files with long shooting times, files with a long shooting interval with the previous file, and those with long subjects are extracted with priority. To generate a 30-second digest.

  Next, when the genre is “party”, images are continuously captured by the video camera during the event, and one moving image file 21 often becomes long. For this reason, applause and cheers often occur in important scenes, and other camera flashes are often hit. Therefore, the digest is generated by detecting these from the audio information and image information of the moving image file and extracting before and after.

  That is, in this case, the digest generation unit 43 first detects the time when a flash or applause occurred from the captured video file by the digest generation method D shown in FIG. 6D (S31), and then the detected time After extracting a total of 10 seconds, 5 seconds immediately before and 5 seconds immediately after (S32), the extracted 10 seconds section is connected in order of photographing time (S33).

  FIG. 7D shows a digest generation method in the case where the genre is “party”. In this genre “party”, before and after the portion where the sound is more than a certain level or the portion where the flash is applied. Are extracted for a certain period of time. As shown by the shaded area in FIG. 7 (D), there are three voice peaks and flash times, and a total of 10 seconds is extracted for 5 seconds before and after the three places to generate a 30-second digest. To do. Here, when more than 3 flashes and applause are detected, all the flashes and applause are detected by shortening the extraction interval, and the priority is determined by the length of the period during which larger applauses and flashes occur. And the number of detection points may be narrowed down to three.

  By a similar method, the digest generation unit 43 also generates a digest for scenes 2 and 3 according to the genre of each scene.

  FIG. 8 shows an example in which a certain event composed of 15 moving image files 1 to 15 is divided into three scenes, and a digest corresponding to the genre is generated for each scene by a generation method.

  For example, FIG. 8 shows a case where the genre of an event composed of 15 moving image files 1 to 15 is “travel / leisure”, for example. That is, an example is shown in which a part of 5 seconds from the beginning of each moving image file is extracted for each scene 1 to 3 by the digest generation method A shown in FIG. 6A, as shown in FIG. Since scene 2 shown in FIG. 7B has four moving image files, and scene 3 in FIG. 7C has five moving image files, extraction is performed from other than the first 5 seconds of the moving image file of each scene. The digest of scenes 2 and 3 is 30 seconds.

  Finally, the digest generating unit 43 generates a digest of the entire event by concatenating the digests extracted from the moving image file for each event scene as shown in FIG. 9, for example (S304).

  FIG. 9 is an example in which three 30-second digests extracted from a video file are connected for each scene of an event composed of the 15 video files 1 to 15 shown in FIG. 8, and a digest of the event is generated. Is shown.

  Therefore, according to the digest generating method and generating apparatus of the first embodiment, a plurality of moving image files photographed by the video camera are classified for each event based on the tag 22, and a plurality of moving image files classified as the same event are further classified. After classifying each scene according to a predetermined threshold according to the genre of the event, from each of a plurality of video files classified as the same scene, a video is generated by a digest generation method according to the event genre of each moving image file Since a part of the file is extracted and an image obtained by connecting a plurality of extracted parts is generated as a digest, an optimal digest can be automatically generated for each event genre.

  In the first embodiment, as illustrated in FIG. 2, the threshold value and the digest generation method can be individually set and changed for each event genre. However, the present invention is not limited to this. Of course, only the threshold value may be individually set for each genre, the digest generation method may be common, and further, the threshold value may be common for each genre, and only the threshold value may be individually set for each genre.

Embodiment 2. FIG.
In the first embodiment, the method of classifying scenes by changing the threshold value for each event genre has been described. However, in the second embodiment, threshold values, coefficients, and the like determined according to the attribute information of the moving image file constituting the event. A method for classifying scenes will be described. The attribute information mentioned here is obtained from information attached to each file such as shooting time, shooting time, shooting position, and genre information, or a plurality of files such as the total shooting time of all files and the number of files for each event. Information. Since the configuration itself is the same as that of the first embodiment shown in FIG. 1, operations unique to the second embodiment will be described with reference to the configuration of the first embodiment shown in FIG.

  FIG. 10 is a flowchart showing the procedure of scene classification in the second embodiment.

  In the second embodiment, first, as in the threshold value determination method of the first embodiment shown in FIG. 3, the shooting interval between moving image files is calculated for each event (S201), and then the length of the event, that is, the event A threshold for scene classification of a specific event is determined based on the difference between the shooting start time of the first file and the shooting end time of the last file (S202). For example, assuming that 5% of the length of an event is a threshold, if the length of the event is 10 hours, 30%, which is 5%, becomes the threshold for scene classification, and the length of the event is 5 hours. In this case, 15 minutes, which is 5%, is set as a threshold for scene classification.

  However, in the second embodiment, when the event genre is “travel / leisure” or “wedding / party”, the threshold determined by the event length in S202 is used as it is, but the event genre is “ In the case of “athletic / sports” or “children / pets”, 1/3 of the threshold determined in S202 is set as the threshold (S203 “Yes”, S204). This is because when the event genre is “athletic / sports” or “children / pets”, it is expected that there will be many video files with short time intervals. The genres and coefficients listed here are examples, and any genre may be multiplied by any coefficient. Further, not only the coefficient but also addition or power may be performed. Of course, the processing of S203 and S204 for changing the threshold value of the specific event may be omitted.

  Then, the scene classification unit 42 compares the threshold value determined as described above with the maximum threshold value determined in advance for each genre, and uses the shorter one as the threshold value for scene classification (S205 to 215). .

  That is, if the event genre is “travel / leisure” and the threshold determined in S202 is 30 minutes or more, the threshold is 30 minutes (S205 “Yes”, S206), while the event genre is “athletic / sports”. If the threshold determined in S204 is 10 minutes or more, the threshold is 10 minutes (S207 “Yes”, S208), while the event genre is “children / pets”, the event shooting period is less than 3 days, and S204. If the threshold value determined in step 10 is 10 minutes or more, 10 minutes is set as the threshold value (S209 “Yes”, S210 “No”, S212 “Yes”, S213).

  On the other hand, when the threshold determined according to the length of the event in S202 or S204 does not exceed the maximum value of each genre, the threshold is set as the threshold for scene classification, and the moving image file is set according to the threshold determined as described above. Are classified into scenes (S216).

  The method of creating a digest for each genre after classifying events into scenes is the same as in the first embodiment.

  Therefore, according to the digest generating method and generating apparatus of the second embodiment, a plurality of moving image files photographed by the video camera are classified for each event based on the tag 22, and a plurality of moving image files classified as the same event are further classified. A maximum threshold value is set for each genre, and if the threshold value determined according to the length of the event exceeds the maximum threshold value for each genre, the maximum value is set as the threshold value. Since the scenes are classified according to the threshold value determined according to the length, and then the digest is generated in the same manner as in the first embodiment, the maximum value of the threshold value is limited for each genre. It is possible to classify each scene by setting a corresponding threshold value, and to automatically generate an optimal digest for each event genre.

  In the second embodiment, the maximum value of the scene classification threshold for each genre is set, but conversely, the minimum value of the scene classification threshold for each genre is set and the threshold determined according to the length of the event. Is smaller than the minimum value, the minimum value may be used as a threshold for scene classification. In this way, for example, in the case of a genre of “travel / leisure”, there is a possibility that the length of the event is short and the threshold value becomes extremely short. However, in the “travel / leisure” genre, it is desirable to classify scenes at intervals of at least movement. For this reason, in the category of “travel / leisure”, the minimum threshold value for scene classification is set to, for example, 10 minutes, which is the shortest time during which movement is performed, and the threshold value is less than this value. Uses 10 minutes as a threshold value. Of course, the maximum and minimum values mentioned above are examples, and other values may be given.

  In the second embodiment, the scene classification threshold is determined based on the length of the event. However, the present invention is not limited to this, and the average of the shooting intervals between the video files constituting the event is determined. Of course, it may be set as a threshold value.

  Further, in the second embodiment, as shown in FIG. 10, with respect to the threshold value determined based on attribute information such as the length of the event and the average of the shooting interval between files for only a specific genre by the processing of S203 and S204. Although it has been described that the maximum value and the minimum value are set for each genre, or that only a specific genre is adjusted by multiplying the threshold value by a coefficient or the like, the present invention is not limited to this and is based on event attribute information. The steps S203 to S215 in FIG. 10 are omitted with respect to the determined threshold value, so that the limitation of the maximum value and the minimum value for each genre, the threshold value adjustment to a specific genre is not performed, or S205 to S215 in FIG. This step is omitted and the threshold adjustment to the specific genre is performed, but it is of course possible not to provide the maximum value or the minimum value for each genre. This also applies to Embodiments 3 and 4 described later.

Embodiment 3. FIG.
In the third embodiment, the threshold value determination method in S202 of the second embodiment shown in FIG. 10 is changed, and only that portion will be described.

  That is, in the second embodiment, the threshold value for scene classification is determined by the length of the event as attribute information. In the third embodiment, the average of the shooting intervals between all files of the event is also used. Shall. Hereinafter, an example in which the threshold value is obtained using the average of the shooting intervals will be described.

  For example, in a certain event, when five moving image files are shot and the shooting intervals are 5, 10, 30, and 3 minutes, the average value of 12 minutes is set as a threshold based on the average shooting interval. .

  In the fourth embodiment, the threshold value for scene classification is determined based on the larger or smaller threshold obtained from the two attribute information, that is, the length of the event and the average of the shooting intervals between all moving image files constituting the event. . Of course, the average of the two obtained threshold values may be used.

  For example, if there is only one moving image file with a long shooting interval in an event of shooting a moving image file at a short interval, the threshold based on the average shooting interval becomes long. For this reason, even a moving image file of another scene whose shooting interval is shorter than the average may be classified in the same scene. Therefore, in the present embodiment, when there is a moving image file with a shooting interval equal to or greater than a predetermined value in an event, a threshold value determined based on the length of the event is used.

  On the other hand, when many moving image files are shot at a substantially constant shooting interval, the event length increases in proportion to the number of files, so the threshold value due to the event length increases. Therefore, when the number of moving image files in the event is greater than or equal to a predetermined value, a threshold value based on the average value of shooting intervals is used.

  Therefore, according to the digest generation method and generation apparatus of the third embodiment, a plurality of video files photographed by the video camera are classified for each event based on the tag 22, and a plurality of video files classified as the same event are further classified. Scene classification is performed using either the threshold determined based on the length of the event or the threshold based on the average of the shooting intervals between all moving image files constituting the event, and then digested in the same manner as in the first and second embodiments. Therefore, an optimal digest can be automatically generated for each event genre after optimal scene classification according to the length of the event or the shooting interval of the file is performed.

Embodiment 4 FIG.
In the fourth embodiment, as in the third embodiment, the threshold value determination method in S202 of the second embodiment shown in FIG. 10 is changed, and only that portion will be described.

  That is, in the second embodiment, the threshold value for scene classification of a specific genre is determined based on the length of the event. In the fourth embodiment, the threshold value is determined based on the position information of the shooting point of each file of the event. For this reason, in the case of the fourth embodiment, a video camera that captures a moving image file has a GPS function, and adds position information of a shooting point as attribute information to the captured moving image file.

  Therefore, in the fourth embodiment, for example, the shooting position of each moving image file is obtained based on the GPS information of the attribute information given to each moving image file constituting the event, and the total moving distance between the moving image files is obtained. A threshold is used for scene classification by applying a certain coefficient to the total movement distance. For example, when the moving distance is 60000 (m), 30 (minutes) multiplied by 1/2000 is set as the threshold for scene classification.

  Therefore, according to the digest generating method and generating apparatus of the fourth embodiment, the plurality of moving image files photographed by the video camera are classified for each event based on the tag 22, and the positions of the plurality of moving image files classified as the same event are further classified. Since scenes are classified based on threshold values based on information, and then digests are generated in the same manner as in the first to third embodiments, an optimal scene classification is performed based on the position information of the moving image file, and then the event genre An optimal digest can be automatically generated every time.

  In the description of the first to fourth embodiments, the video camera has been described as an example of the moving image shooting device. However, this is an example, and in the present invention, in addition to the video camera, a mobile phone having a digital camera and a moving image shooting function The apparatus and method of the present invention can be applied to any apparatus having a moving image shooting function such as a telephone.

  In the second to fourth embodiments, examples of the attribute information when classifying the scene include the length of the event, the average of the shooting intervals between all the moving image files constituting the event, and the position information of each moving image file. However, in the present invention, the attribute information is not limited to these, and the threshold for scene classification may be obtained using other attribute information, or a plurality of attribute information may be combined as in the third embodiment. good.

  In Embodiments 1 to 4 described above, after scene classification, a digest is generated by extracting a part of a video file from each scene by a digest generation method according to the genre. However, the present invention is not limited to this, and the threshold value is set by the method of Embodiments 1 to 4 even in the apparatus and method that does not generate a digest only by setting the threshold value by the method of Embodiments 1 to 4 and classifying the scene. Then, scene classification may be performed, and digest generation may be applied to an apparatus and method that employs a method different from the digest generation method of the first to fourth embodiments.

It is a block diagram which shows the structural example of Embodiment 1 of the digest production | generation apparatus of this invention. It is a figure which shows each example of the genre of the event in Embodiment 1, and the setting example of a scene classification threshold value and a digest production | generation method for every genre. 4 is a flowchart illustrating a procedure of scene classification processing according to the first embodiment. It is a figure which shows an example which classified the image | photographed moving image file into the scene. 6 is a flowchart illustrating a procedure of digest generation processing according to the first embodiment. (A)-(D) is a flowchart which shows an example of the digest production | generation methods AD according to a genre, respectively. (A)-(D) are the figures which show an example of the cut extracted from each scene by the digest production | generation methods AD according to the genre shown to FIG. 6 (A)-(D), respectively. It is a figure which shows the example of digest extraction for every scene of a certain event by this Embodiment 1. FIG. It is a figure which shows the example which produces | generates the digest of the whole event which connected the digest extracted for every scene by FIG. 10 is a flowchart illustrating a procedure for scene classification in the second embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Content holding part 21 Movie file 22 Tag 31 Shooting part 32 Tag creation part 41 Event classification part 42 Scene classification part 43 Digest generation part 44 Input instruction part

Claims (9)

A digest generation device that generates a digest from a plurality of video files obtained by shooting,
Tag creation means for creating a tag in which genre information indicating the genre of an event of each video file and event identification information for identifying the events from each other are described for each of the plurality of video files;
Event classification means for classifying the plurality of video files for each event based on the event identification information described in the tag;
Scene classification means for classifying the plurality of video files classified for each event into scenes based on shooting intervals between the video files;
Digest generating means for extracting a part from a plurality of video files classified for each scene and generating as a digest by a digest generating method based on the genre information described in the tag;
A digest generation device. In the digest production | generation apparatus of Claim 1,
The scene classification means includes
A threshold is set in advance for each genre,
When classifying the plurality of video files classified for each event, for each scene based on the shooting interval between the video files,
Based on the genre information described in the tags of the plurality of video files classified for each event, the threshold value corresponding to the genre of the plurality of video files is set, and the set threshold value and the video file A digest generation device that classifies scenes based on the shooting interval between them. In the digest production | generation apparatus of Claim 1,
The scene classification means includes
The maximum threshold value is set in advance for each genre,
When classifying the plurality of video files classified for each event, for each scene based on the shooting interval between the video files,
Determine the threshold based on the attribute information of the video file that constitutes each event,
Based on the genre information described in the tags of the plurality of video files classified for each event, the maximum value according to the genre of the plurality of video files, and the attribute information determined Compare with the threshold,
When the threshold value determined by the attribute information is greater than or equal to the maximum value, the maximum value is selected, while when the threshold value determined based on the attribute information is smaller than the maximum value, the threshold value is determined based on the attribute information Select the threshold,
A digest generation apparatus that classifies each scene based on the selected threshold value or the maximum value and a shooting interval between the moving image files. In the digest production | generation apparatus of Claim 1,
The scene classification means includes
The minimum threshold value is set in advance for each genre,
When classifying the plurality of video files classified for each event, for each scene based on the shooting interval between the video files,
Determine the threshold based on the attribute information of the video file that constitutes each event,
Based on the genre information described in the tags of the plurality of video files classified for each event, the maximum value according to the genre of the plurality of video files, and the threshold value for determining the attribute information Compare
When the threshold value determined based on the attribute information is less than or equal to the minimum value, the minimum value is selected, while when the threshold value determined based on the attribute information is greater than the minimum value, based on the attribute information Select the determined threshold value,
A digest generation device that classifies each scene based on the selected threshold value or the minimum value and a shooting interval between the moving image files. In the digest production | generation apparatus of Claim 1,
The scene classification means includes
When classifying the plurality of video files classified for each event, for each scene based on the shooting interval between the video files,
The threshold value is determined based on the attribute information of the video file constituting each event, and the threshold value determined based on the attribute information is adjusted based on a predetermined coefficient only when the genre of each event is a specific genre. A digest generator. In the digest production | generation apparatus as described in any one of Claims 3-5,
The said attribute information is a digest production | generation apparatus which is the length of an event. In the digest production | generation apparatus as described in any one of Claims 3-5,
The said attribute information is a digest production | generation apparatus which is the average of the imaging | photography space | interval of the moving image file which comprises an event. In the digest production | generation apparatus as described in any one of Claims 3-5,
The said attribute information is a digest production | generation apparatus which is the positional information which shows the imaging | photography position of the moving image file which comprises an event. A digest generation method for generating a digest from a plurality of video files obtained by shooting,
Based on the event identification information of a tag in which genre information indicating an event genre of each moving image file assigned to the plurality of moving image files and event identification information for identifying the events from each other are described. Categorizing video files by event,
Classifying the plurality of video files classified for each event into scenes based on shooting intervals between the video files;
Extracting a part from a plurality of moving image files classified for each scene by the digest generation method based on the genre information described in the tag, and generating as a digest;
A digest generation method comprising:

Images