JPH1139343A - Video retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a suitable video retrieval device by using it for speedily detecting a video which a viewer desires from a large quantity of spread videos. SOLUTION: A change position detection part 13 detects the switch of a scene. A representative picture extraction part 14 extracts one to several pieces of representative picture VS(video software) from the respective scenes. A character data extraction part 23 extracts character data from them. Character data is accumulated in a character data accumulation part 23 with information on the extracted position. When a user desires to check something in business or personally, accumulated character data is retrieved with a keyword showing it. The representative picture VD which is hit and original video software VS are displayed on a display 26. Data can be selected from the display of a table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video search device, and more particularly to a video search device suitable for use in quickly finding what a viewer wants from a large amount of distributed video. In this specification, an image corresponding to one frame of a television broadcast is referred to as an "image" or a "screen", and a series of a plurality of images is referred to as an "image". This includes audio signals and caption codes (caption codes)
The one to which the information is added is referred to as “video software (VS in FIG. 3 (hereinafter, this symbol VS is omitted))”. These include motion picture film, videotape, television broadcast, DVD
(Digital video disc or digital versatile disc), the Internet, computer storage media, and the like.

[0002]

2. Description of the Related Art An era has come when a large amount of video software is supplied, such as for movies and sports broadcasting. Besides,
There is a wide variety of video software in the fields of drama, music, education, news coverage, etc., and the supply is expected to increase further in the future. As the saying goes, the picture software is very useful. Whether it's work or personal hobbies, it's very valuable if it can be accurately grasped and used as an occasional decision source or as a library.

Video software follows the flow of time. One needs time to see this. Therefore, when the supply amount becomes enormous in this way, it is no longer possible to record it on a video tape or the like and watch it later slowly as in the past. Thus, for example, a method is conceivable in which a keyword, a search index, or the like is attached to a necessary image (frame) or video, and a search is performed later using this.

[0004]

However, in this method, it takes time and effort to input a keyword, an index, and the like. Specifically, the content of each piece of video software is checked, and keywords or indices are assigned to all or individual images. Such a processing method used to be a rabbit in the past when the supply amount was small. However, it cannot be used in today's multimedia age, or in the future in an ever increasing volume. It is always a matter of what keywords and indexes to add.

[0005] As the calculation speed of the computer has been improved, in the information retrieval centering on characters, the character data body itself is subjected to "full-text search (full text search)" without indexing or registering keywords in advance. And came to pull out the relevant thing. In this case, it is known that it is possible to perform a fairly accurate search by targeting nouns in the text, and when a user enters a word that is used daily, the text containing that word is searched at high speed. .

Similarly, it is conceivable to perform "all image search (full image search)" in image search. However, for an image, searching for a match is a very time-consuming process and is often meaningless. For this reason, some other method is needed.

[0007] An object of the present invention is to provide a video search apparatus suitable for use in quickly finding what a viewer wants from such a large amount of distributed video.

[0008]

According to the first aspect of the present invention, a character data extracting means for extracting character data from video software, and a character data storage for storing the character data together with information on the extraction position. Means, selecting means for selecting a desired one from the stored character data, and extracting means for extracting an image related to the selected character data from the video software.

In the invention according to claim 2, the character data extracting means extracts the character data from a representative image extracted from the video software. In the invention according to claim 3, the extracting means according to claim 1 or 2 extracts a representative image relating to the selected character data.

The present invention focuses on characters included in video software. Here, the characters in the video software have the following characteristics. (1) In movies and the like, characters are used to convey important information such as titles and performers. (2) Subtitles for foreign movies are designed to convey conversations and stories with minimal characters. (3) In TV (television) news, etc., characters are used to sort out the outline of the case. Here, the characters are used to collect information compactly. (4) Numerical information is often displayed on the screen for sports results, horse racing results, and the like.

As described above, the video software itself conveys subtle emotions and sensations, while the characters in the software are used for transmitting information having an explicit meaning. The present invention focuses on the character data, and extracts and accumulates the character data according to the configuration described in the claims, so as to form a so-called image and video database.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail based on illustrated embodiments. First, an embodiment (FIG. 1)
An outline of the video search device 100) will be described. (1) The received original video software such as a TV broadcast (television broadcast) is supplied to the original video storage unit 12 via the input interface 11 and stored therein. It should be "video software storage unit", but "software" is omitted because it is redundant). It is also supplied to the change position detection unit 13 to detect a position where the state of the video software changes, such as switching of a scene (scene, cut). Video software supplied in another form, for example, video tape, DVD, Internet, CD-ROM, etc., is also taken in from the input interface 11.

(2) When the change position detecting section 13 detects a state change, for example, a scene change,
In response, the representative image extracting unit 14 extracts the representative image VD (FIG. 2 (hereinafter, symbol VD is omitted)). Here, the representative image is a single still image or a short moving image of about 2 to 5 seconds extracted from the vicinity of the state change position, and a set thereof is accumulated as a summary VE (FIG. 2 (hereinafter abbreviated as VE)). Is done. The summary is 1/10 compared to the original video software
The data amount is about 1/100. Therefore, the subsequent processing is simplified. It is also used as a material for determining whether or not to view the original video software.

The representative image may be extracted from the original video software at a fixed cycle. Characters in video software are displayed for a relatively long time. This is because the time until the viewer finishes reading the displayed text is considered. Conversely, the same characters are displayed during this time. Therefore, at this time interval, for example, 5 seconds or 10 seconds,
If a still image or a moving image is extracted, characters included in the video software can be extracted with almost no leakage. When using this method, it is not necessary to extract a representative image, but an image is extracted at regular time intervals from an original video and characters included therein are extracted. The images extracted here may also be collected and used as a summary.

(3) Extract characters from the representative image. Using character recognition software, character data is extracted from the extracted representative image. The extracted character data is stored as a keyword. In some cases, this keyword may be limited to nouns only. When accumulating the keyword, the summary, and the original video software, heading information such as a channel number, a broadcast date, hours, minutes, seconds, and a frame number, which can maintain a mutual link relationship, is attached.

The accumulated keywords, summaries, and original video software are used, for example, as follows. (1) If the user wants to know whether related video software is stored for a theme that the viewer is interested in, for work or personal reasons, a word that expresses the event simply, for example, PATHOLIS Enter words such as free keywords as keywords whenever necessary. (2) This is collated with the stored character data. Exact match, partial match and other search rules are set by the viewer. If there is any, the next step is to extract the corresponding summary and check the contents of the image. (3) Looking at the summary, if it is considered to be related to the desired theme, take out the original video software and check the contents.

As described above, the summary is one-tenth to one-hundredth of the size of the original video, and it does not take much time to see it. Therefore, it is easier to perform a search by arranging the summary in two stages and determining whether the summary is the desired one or not.

It is also possible to register a keyword in advance and, if there is a keyword corresponding to the keyword, separately record table-of-contents information and the like relating to the video software. For example, the viewer checks the accumulation state once a day. Applicable items are displayed on the screen in a list, and the viewer is selected, and the summary and the original video software are displayed. Instead of entering a keyword, the stored character data is displayed in the order of the Japanese syllabary, for example, and the viewer selects a desired one from among them, and displays the summary and the original video software. Is also good.

Hereinafter, the video search apparatus 100 according to the embodiment will be described.
Will be described with reference to the drawings. That is, in FIG.
The video software supplied from the broadcast receiving unit 16, DVD player 17, video tape deck 18, Internet connection unit 19, CD-ROM drive 20, etc.
The video is retrieved by the video search device 100 via the input interface 11.

The TV broadcast receiving section 16 has a number of receiving channels. For example, terrestrial TV broadcasting (VHF, UH
F) Simultaneously receive a plurality of analog and digital TV broadcasts desired by the viewer for a plurality of channels such as satellite broadcasting and cable TV. From the Internet connection unit 19, desired video software on the Internet is taken in periodically or in response to a viewer's operation.

The video software fetched therefrom is stored in the original video storage unit 12. The accumulation should be done digitally. The analog type is digitized by the input interface 11 and stored in the original video storage unit 12. MUSE, MPEG, JP to reduce data volume
It is preferable to compress the data by EG or another method. Digital television broadcasting and video software on the Internet can be as it is. Video software, for example, 10
It is divided into minutes and stored in the original video storage unit 12 as one file. The table of contents information (index) required for later reading, such as the reception date and time of each file, the reception time of the file, and the channel number, is recorded in the file allocation table (FAT).

The storage in the original video storage unit 12 is performed for a specified channel continuously for a certain period of time, for example, 72 hours. Voice information is also recorded. In addition,
The viewer sets the recording date, time, and recording target channel using the user operation unit 22. There are times when it is known that there is no need to record from the beginning, and there are programs where the viewer is not interested. Therefore, if the time zone can be set not only by the date and time but also in units of hours and minutes, it is easy to check the recorded contents later. The time zone setting procedure may be the same as that of an existing video tape recorder or the like.

The recording (accumulation) is preferably performed cyclically. For example, the storage capacity of the original video storage unit 12 is
It is possible to record for 3 consecutive days for 5 channels. When the viewer has time, the viewer checks the stored character data, abstract, and original video software, and deletes unnecessary ones. As a result, the storage areas of the original video storage unit 12 and the summary storage unit 21 become free. The storage capacity may be exhausted by recording broadcast contents continuously for a certain period. In this case, the next new video signal is overwritten on the oldest one of the recorded video.

In this case, it is not necessary to specify the date, time, and the like to be recorded, and only the desired channel (the channel desired to be monitored) need be specified by the user operation unit 22. However, as described above, there is a time zone that does not need to be recorded from the beginning. Therefore, even in the case of cyclic recording, setting a time zone in which recording is not required enables a smaller storage capacity to be consumed.

Returning to FIG. 1, the change position detecting unit 13 detects the television broadcast taken at that time or the DVD stored video software being reproduced at that time.
When a state change occurs in the designated element of interest, the fact is notified to the representative image extracting unit 14.

According to the present invention, character data included in video software is extracted, stored, and used as a key for video search. As described above, the character data may be directly extracted from the original video software (broken line L in FIG. 1), but here, as described above, the processing is simplified to facilitate the check of the recorded contents. As shown in FIG. 2, a representative image VD is extracted from the original video software, and character data is extracted from the representative image VD (hereinafter, the symbol VD of the representative image is appropriately omitted).

The detection of the state change position is for giving a motivation for extracting the representative image. In response, the representative image extracting unit 14 extracts the representative image. In addition, each representative image includes not only one still image but also a moving image of several seconds related to the change position.

The video software (broadcast program) includes various detectable elements (features). For example, voice (sound)
There are human voices, music, cheers, pulsed sounds, and many other detectable elements. Also, the video has several detectable elements such as being stationary, sudden changes in screen brightness, sudden changes in the color of the entire screen, and the like. Here, the representative image is extracted to extract character data existing in the video software. Therefore, for the element of interest, it is preferable to select an element that changes in relation to the appearance of a new character on the screen.

For example, in the case of a news program, after the reading of the news manuscript by the announcer or caster ends and the voice is interrupted, new characters are projected on the screen in synchronization with the start of reading of the next news manuscript. Therefore,
This change in the factor of intermittent speech can also be used to motivate the representative image extraction. The same tendency is seen in educational programs.

The relationship between the change in the element and the appearance of the character changes according to the genre (field) of the video software. Therefore,
If the characteristics of the video software to be stored at that time are clear from a program guide or the like, when setting the recording target channel, the viewer can set which element to focus on.

Some elements do not depend on characteristics or fields. When a recording target is not specified, that is, when recording is continuously performed on several channels and the like and character data in the channel is extracted as a search key, it is good to pay attention to such a universal element. An example is shown below. (1) Extraction of head image The image at the start of the original video software is simply extracted as a title (representative image). If the broadcast start time of the program can be grasped from a program table or the like, the leading image is extracted. For example, if recording is performed continuously for a long period of time in units of 10 minutes, the leading image of each segment is extracted as a title.

(2) Periodic clipping An image is clipped periodically from the video software at regular intervals. If an abstract consisting of M representative images is generated for the original video software composed of N images, images for every N / M images are extracted from the original video software as representative images. This is the simplest method. In image compression methods such as MPEG, JPEG, and MUSE, complete images are transmitted every few images.
The difference from the previous image is transmitted. Therefore, with respect to such image data, a complete image sent to some of the image data may be extracted as a representative image.

This method of periodic clipping is useful when character data is directly extracted from video software without generating a summary. That is, the characters in the video software are displayed just for the audience to read, but it takes a certain amount of time for a person to read the characters. For this reason, the characters on the screen are often displayed for a relatively long time, for example, 5 seconds or 10 seconds in consideration of this. Therefore, if it is assumed to be 5 seconds, in the case of television, one image is displayed every 150 frames for 10 seconds, and if one image is taken out every 300 frames and inspected, the video software Most characters appearing in can be extracted.

Therefore, according to this method, a character is directly extracted from original video software without generating an abstract. However, as mentioned above, the summary is also a source of judgment as to whether or not to view the original video software. Therefore, the character data is directly extracted from the original video as referred to herein, while, on the other hand, the elements that will generate a summary that expresses the contents of the video software in a more accurate manner are separately selected, and the state change is motivated. Then, a summary of the video software may be generated.

In this case, the following are examples of the elements of interest. Image genre Featured elements News Cut with pattern (flip) Drama Cut with subtitle Cut with audio Documentary Cut with audio English conversation Cut with subtitle Sport Applause, cheering cut and its surroundings (audio climax) Animation with subtitle Cut Long, stationary cut Cut with sound TV shopping Cut with subtitles (You can see information such as price) Song program Cut at the beginning of music (discriminate from voice) Educational program Cut with pattern Variety show Cheerful cut Orchestra Music (Periodical extraction) Weather forecast Stationary cut (pattern (flip) = board on which characters and images are written. Speakers and others use TV.) Broadcast channel characteristics Depends on Therefore, summarization and checking may be performed several times, and a preferable element of interest, a predetermined position, and the like may be determined.

(3) Clipping for Each Scene When the scene changes, new characters are often displayed in the image. Therefore, it is also possible to detect this, extract the appropriate number of representative images for each scene, and check for the presence or absence of character data. It is arbitrary from which position in the scene the representative image is taken. For example, if the scene changes in CS1 and CS2 in FIG. 3, those and the center PC of the scene are one candidate.

The switching of the scene is determined, for example, as follows. (1) Pixel aggregation processing Needless to say, an image has a two-dimensional spread. For example, the NTSC system is a group of pixels (pixels, dots) such as approximately 250 dots × 525 lines, and each pixel has brightness, saturation, and hue. 30 images are transmitted per second.

Needless to say, the number of pixels is large. Therefore, in order to speed up and simplify the processing, pixel aggregation is performed here. For example, if the sum and average value of 4 × 4 pixels are determined, the video data is 1/1
When the same processing is performed on 8 × 8 pixels, the video data becomes 1/64.

This aggregation has the effect of ignoring gradual changes in the image and facilitating the determination of scene switching. In other words, the scene switching determination is basically to compare the pixels of the two preceding and succeeding images with each other and to determine whether there is a large difference therebetween. Specifically, for example, the data of both pixels are compared in several areas of the image, and if many of them are in common, there is, for example, a scene of one scene such as a person moving without changing the background. Think of it as the image inside.

However, for example, when there is a zoom up / down or a slow change in the direction of the camera, comparing the pixels of the preceding and succeeding images without performing the aggregation processing here, the appearance is slight, but almost at the pixel level. Although the data is different in any part and is actually in the middle of the scene, many of them are erroneously determined to be the scene switching positions.

Looking at the same thing at the coarse pixel level after the aggregation processing, the images up to a certain number have the same data and no change because the pixels are coarse. Therefore, these are correctly determined to be images belonging to the same scene. Therefore, first, as a pre-process, the values of these coarse pixels, a (t, x, y), are extracted here. Here, t: time x, y: coordinates of the image after the aggregation processing a: color value of the point (x, y). a
Is to adopt R, G, B as they are, or C ₁ · R + C ₂ ·
G + C ₃ · B (R, G, B are values of three primary color information, and C ₁ , C ₂ , C ₃ are weighting coefficients).

(2) Aggregation processing in the time direction Next, regarding the video data on which the above aggregation processing has been performed,
For each image, the difference between its respective pixel and each pixel of the previous image is determined. Then, it is determined whether or not the difference is large enough to be determined to be the scene switching position. To simplify the processing, a fixed threshold value is set for the magnitude of the difference at this time. If the difference between a certain number of coarse pixels constituting the image exceeds a threshold value, it is determined that the image has changed there, that is, that a scene change has occurred.

Even if they belong to the same scene, the magnitude of the difference between pixels differs depending on the content. In a scene where one frame changes quickly, the difference between the preceding and following images is large, and when the scene is slow, the difference is small. Therefore, the simple processing as described above is one method, but here, the following processing is further added in order to more accurately extract the representative image. "Process of temporal difference comparison of video" A temporal difference is obtained for the data a (t, x, y) of each point constituting one image. That is, d (t, x, y) = a (t, x, y) −a (t− △ t,
x, y). Where Δt is an appropriate time width. this is,
The difference (change amount) between pixels of two images that are temporally adjacent to each other by Δt is shown.

The difference d (t, x, y) is calculated for a predetermined period.
For example, it is determined for each image during one minute. In terms of the formula,

(Equation 1) Becomes

FIG. 3 shows an example of the difference Da (t) for an aggregated video image of, for example, one minute thus obtained.
The position where the value is large, that is, the position CS1,
CS2 indicates that there is some major change in the image, where it is likely that the scene has changed. Therefore, the area between CS1 and CS2 is estimated as one scene, and an appropriate number of representative images are extracted from the scene. These are the “temporal difference comparison processing of video” referred to in the above heading.

For each frame constituting the video software, for example, a video signal corresponding to one horizontal scanning line near the center of the image is focused on, and a difference accompanying a transition of the screen is determined with respect to the video signal to detect a scene change. You may do it. Specifically, for example, the video signal for one horizontal scanning line is divided into N sections, and a sum or an average value is obtained for each section. Then, for each section, a difference from the average value of the section on the previous screen is obtained. The difference may be summed for each frame, and a position where the value is large, that is, a position corresponding to CS1 or CS2 in FIG. 3 may be obtained, and this may be set as a scene switching position similar to the above.

Returning to FIG. 1, each representative image extracted in this way is supplied to the summary storage unit 21 and the character data extraction unit 23. The summary storage unit 21 stores the supplied set of representative images as a summary corresponding to the video software. In addition, the date and time when the original image was recorded and the time when the recording of the original image was started are recorded in the file so that the relationship between the accumulated representative images and the original image can be understood. Each representative image is provided with information on the position where the representative image is taken out.
This can be represented, for example, by using the recording start time of the original video as the base point, the number of seconds elapsed from that point, and the frame number.

The character data extraction unit 23 detects character data directly from the supplied representative image or directly from the original video software. There are several forms of character data included in video software. Therefore, the detection here is also performed according to this mode. Since one piece of video software may include character data in a plurality of forms, detection may be performed by a plurality of methods.

(1) A case where a character is incorporated in an image as an image, that is, as an image or dot data. In this case, characters are detected by applying the kanji OCR (optical character reading) technique. In general, characters are displayed below or to the left and right of the image. Therefore, if there is no problem if the accuracy is slightly reduced, the presence or absence of a character may be inspected in these portions. Then, the processing can be simplified and speeded up.

(2) When a character code is added. Like text broadcast subtitles, character data (caption data) may be supplied as character code as image gap information in a frame signal at the time of vertical feedback. This is simple. These character codes may be sequentially stored in the character data storage unit 28 together with the video position information. For further development, the video software to which the character code is added may be extracted by focusing only on the character code and the positional information of the video without extracting the representative image.

Referring back to FIG. 1, the user operation unit 22 includes a keyboard, a mouse, and the like. In response to these operations by the viewer, the search control unit 24 causes the character data storage unit 2
8 to extract the character data stored in the memory 8, extract abstracts corresponding to the selected character data, extract original video software, display on the display 26, and the like. Reference numeral 27 denotes a keyword holding unit which holds a keyword input by a viewer in advance. The search control unit 24 performs registration processing of the keyword, the keyword registration processing, and the character data storage unit 28.
Also, a collation process with the stored data is performed.

The original video, the summary, and the detected character data are
As described above, indexes such as the broadcast date, hour, minute, second, and frame number are assigned to each of them, and are stored in the original video storage unit 12, the summary storage unit 21, or the character data storage unit 28.
These are the video software databases themselves. Therefore, there are various uses. The description of the general operation described above is the same as that of the first embodiment, but the following is an example.

(A) Whenever the viewer needs it, search by inputting a keyword. This search method is the most common database search method. Specifically, (1) When a theme that the viewer wants to find out comes out because of work or personal matters, some words that would express the theme are assumed, Operation unit 22
Type from.

(2) In response, the search control unit 24 collates the keyword with the character data stored in the character data storage unit 28, and checks whether or not the keyword exists. The collation is performed in a perfect match, a partial match or any other form desired by the viewer. (3) If there is a corresponding summary or video software, the search controller 24 lists the table of contents on the display 26. If stored in a medium such as a DVD, a volume name or a title name representing the DVD or the like,
A list of track numbers and the like is displayed on the display 26.

(4) The viewer selects a desired one from the list. Using the table of contents information, such as the broadcast time and the DVD title, provided in the list as a clue makes it easy to approach a desired summary. Search control unit 24
Extracts an abstract corresponding to the character data selected by the viewer from the abstract storage unit 21 and displays it on the display 26.

Various methods of displaying the summary are also conceivable. For example, each time a viewer presses a key or the like, a representative image may be displayed one by one in a manner similar to turning a page. The representative images may be displayed one after another at a certain speed in an automatic flip type. Needless to say, video software has a story. Therefore, the viewer can unexpectedly and accurately understand the contents of the summary even if the images are displayed one after another at a high speed. Functions such as slow playback and review are also provided,
You may want to make this available as needed.

(5) After looking at an appropriate number of summaries or looking at the end, the video software or a part thereof
When it is determined that the representative image is related to or is likely to be related to the desired theme, the reproduction of the original video software from the position where the representative image is extracted is instructed. The summary is one-tenth to one-hundredth the size of the original video software and does not take much time to view. Therefore,
With the two stages, the time required to reach the desired original video can be reduced.

In response to the viewer's instruction, the search control unit 24 sequentially reads out the original video software from the original image storage unit 12 onward after the summary extraction position and displays it on the display 26. This processing is performed in the same manner as the reproduction processing on a video tape deck or other ordinary video equipment.
Again, it is convenient to provide fast forward, review, pause and other keys.

(B) A keyword is registered in advance,
Applicable list is displayed. In this case, keywords indicating items that the viewer wants to watch are stored in the keyword holding unit 27 first. For example, when the viewer performs a confirmation operation once a day, in response to the confirmation operation, the summary including the character data matching the registered keyword or the table of contents information of the original video software is displayed on the display 26 as a list. Is done.

That is, the types of information that we need on business or personally or want to keep an eye on are limited to some extent. Entering the same keyword repeatedly every day can be a little annoying. Thus, in this way, the keywords used in this kind fixedly are registered in the keyword holding unit 27 in advance. Then, when the viewer wants, a list of applicable items is displayed on the display 26. For example, if you want to keep track of the video software related to patents, you can use “Patents”, “Industrial Property Rights”,
Is registered in the keyword holding unit 27, and the corresponding video software is captured. politician,
By registering the names of celebrities and movie stars, you can quickly find the video you want to know.

(C) List of stored character data. For example, the stored character data is displayed on the display 26 in alphabetical order. The viewer may select a desired one from among them and display the corresponding summary or original video software.

After the summary and the video software are confirmed by these various methods, if the representative image, the summary, the video software, etc. to be stored come out, the user operates a key of "save" (not shown). In response, the search control unit 24 stores them in the desired software storage unit 29. This allows you to refer to it later, whenever you need it.

The method of reproducing the sound is arbitrary. For example, when playing back the original video software, it may be played back at the same speed. However, when the summary is reproduced, it cannot be reproduced as it is. Therefore, for example, audio data for several seconds near each representative image extraction position is extracted, and is reproduced at the same or slightly earlier than the original video. Alternatively, after digitization, the data may be thinned out and stored, and then reproduced.

A modification will be described. In summary playback, four or six representative images may be displayed on one screen of the display 26. As described above, the representative image has a story. Therefore, even if a plurality of these are displayed on the display 26 at once, the viewer can easily perceive the contents. Given the story,
It can be said that this way is easier to understand.

Therefore, if, for example, X representative images are displayed on the screen at a time, the summary can be simply reproduced in 1 / X time for each image. Since the required time can be reduced, the display time of the display screen in which the X representative images are combined into one image may be increased. This further facilitates understanding of the video software.

The present invention may be realized as a dedicated device or as an application program of a personal computer. A dedicated chip may be manufactured and incorporated in a personal computer, or incorporated in a DVD player, video tape deck, game player, or the like.

For one piece of video software, respective representative images motivated by changes in different elements may be extracted and stored as separate summaries. When checking the recorded contents,
Each summary can be replayed and the subsequent checks performed by those who are easy to understand.

The original video storage section 12 and the summary storage section 21 may use a home video tape or other magnetic tape. If the recording time is not enough, arrange a plurality of decks (drives). The storage of the original video software requires a certain storage capacity. Therefore, it seems that the above-mentioned cyclical method is preferable for the recording. However, the summary and the character code may be stored for a considerably long time because the required storage capacity is small.

Although the original video software is erased by overwriting, it is somewhat difficult to confirm. However, it is possible to grasp to some extent what the contents of the video software corresponding to the keyword were by simply leaving the character data and the summary. . Using these as clues, it is also possible to make inquiries to the broadcast source and data supplier. Therefore, even with this embodiment, the utility as a database is great.

It is better not to accumulate unnecessary summaries and character codes because they also increase search noise. For this reason, it is also good for a process in which the viewer checks the data stored for several days, stores only the data determined to be stored, and deletes the rest on the spot (search noise = applicable to the keyword) Data that does not serve the purpose of search).

[0071]

As described above, the present invention focuses on the property that characters present in video software simply represent the contents of the video software, and retrieves and accumulates them, making them usable as a database. did. Therefore, the viewer can easily extract a desired one from the large amount of supplied video software by using a desired search format, for example, a keyword of a word used everyday, whenever necessary. You can use them for work, work, and personal use.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an example of an embodiment of the present invention.

FIG. 2 is a conceptual diagram showing an example of video software, a representative image, and character data.

FIG. 3 is a waveform diagram showing a relationship between a difference between video data at time t, video data at time t + Δt, and scene switching.

[Explanation of symbols]

REFERENCE SIGNS LIST 100 video search device 11 input interface 12 original video storage unit 13 change position detection unit 14 representative image extraction unit 16 TV broadcast reception unit 17 DVD player 18 video tape deck 19 Internet connection unit 20 CD -RO
M drive 21 ... Summary storage unit 22 ... User operation unit 23 ... Character data extraction unit 24 ... Search control unit 26 ... Display 27 ... Keyword storage unit 28 ... Character data storage unit 29 ... Desired software storage unit CS ... Scene switching Da (t ) ... time t
And the difference between the video at time t + △ t
PC: Scene center L: Direct extraction of character data VD: Representative image VE: Summary VS: Video software

Claims (3)

[Claims] 1. Character data extracting means for extracting character data from video software, character data storing means for storing the character data together with the extraction position information, and selecting a desired one from the stored character data. A video search device, comprising: a selection unit that performs the selection and an extraction unit that extracts an image related to the selected character data from the video software. 2. The video search device according to claim 1, wherein the text data extraction unit extracts the text data from a representative image extracted from the video software. 3. The video search device according to claim 1, wherein the extraction unit extracts a representative image related to the selected character data.