JP2002049625A - Image retrieval device and image retrieval method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to retrieve an image by a spontaneous sentence, and to directly retrieve a corresponding spot of the image. SOLUTION: This invention is constituted so that, in registering images, an editing cut point is detected and a key word is inputted regarding this corresponding portion, and, in retrieving, an independent word morpheme obtained by analyzing a natural language is collated with the key word, and the corresponding image or the corresponding cut is retrieved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image retrieval apparatus and an image retrieval method for retrieving and extracting a required image.

[0002]

2. Description of the Related Art A conventional apparatus of this kind is described in, for example, "Knowledge Engineering and Artificial Intelligence", 56-9, pp. 139-143. 65-72. Hereinafter, a conventional image search system will be described with reference to FIG. FIG. 4 is a diagram showing a configuration diagram of a conventional image search system.

In FIG. 4, various images are recorded in an image storage section 40 in advance. The inference engine 41 infers an input sentence input from the interactive interface 42 by the user. The interactive interface 42 is an input device such as a keyboard. The image description unit 44 stores the image stored in the image storage unit 40 into an image storage frame 44a, a component instance frame 44b, and a component class frame 44c during the image detection operation. The image recording frame 44a is a part for storing image information.

[0004] The component instance frame 44b is:
This part stores a tree-structured heading used for detecting an image. The component class frame is linguistic information such as a noun and a verb used for a heading used for detecting an image.

[0005] The image interpretation knowledge section 43 includes an inference engine 41.
Gives the knowledge for image detection. Specifically, the image interpretation knowledge unit 43 stores the image stored in the image storage unit 40 and the component instance frame 4 serving as a heading storage unit.
4b is a section for giving more detailed information to 4b. For example, even if it is a "house", there are a "large house", a "small house", a "red house", and the like. The image interpretation knowledge unit 43 gives these information to the inference engine 41, and gives information that these all indicate the information of "house". Therefore, even if the information “house” is input to the inference engine, the image storage unit 40 outputs all the “houses” stored in the image interpretation knowledge unit 43 as candidates to the image storage frame 44a as candidates.

Next, an image search operation of the conventional image search system configured as described above will be described.

[0007] When the user wants to obtain an "image showing a person", the user operates the interactive interface 42 and inputs "person" as an appropriate image search request. Then, the inference engine 41 executes the component class frame 44
With reference to c, an image with a heading "person" is detected from the image recording frame 44a.

When the user wants to obtain “an image of a house with a square window”, the user must use the interactive interface 4.
2. Operate 2 and select "house",
Input the hierarchical information such as “window” and “square”.
Then, the inference engine 41 finds the heading of the tree structure described in the component instance frame 44b,
By tracing “house”, “window”, “square”, and the like, the house having a square window from the image recording frame 44a ”
Search for images that match.

As described above, by making the headings used for image detection into a tree structure, a long image search request such as "a house having a square window" can be used for "house", "window", "square".
An image search can be made using a short general image search request such as that described above. In addition, by making the image search request a short and general word such as “house”, “window”, and “square”, the search hit rate increases and the image search efficiency increases.

[0010]

However, the above-described configuration is intended for still images, and in the case of various scenes such as moving images, the number of image search requests is enormous. . Therefore, even when a heading for searching for an image is configured in a tree structure, the number of types of trees is enormous, and it is difficult to describe the heading in a tree structure.

[0011] Further, in the conventional system, words used for headings are limited. For this reason, the operator must use the image search information corresponding to the headline of the system. For this reason, there is a disadvantage that the operator has to search for usable image search information before searching.

It is also conceivable to try to search for an image by analyzing a given natural language, but in this case, it is necessary to make an enormous number of words included in the natural language a headline. Therefore, even when the headings are configured in a tree structure, the number of types of trees is enormous, and it is practically difficult to realize.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide an image retrieval apparatus capable of retrieving a video corresponding to the content of a large-capacity moving video. I do.

[0014]

SUMMARY OF THE INVENTION The present invention divides a moving image into cut images, inputs cut attributes for each cut image, and compares input image detection information with input cut attribute information. Then, a cut image including the matching cut attribute information is extracted.

As a result, the number of cut attributes is reduced. Therefore, a search system using a cut attribute for a large-capacity video becomes possible.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image retrieval apparatus according to a first aspect of the present invention comprises: video storage means for storing an input video; edit cut point detection means for detecting an edit cut point of the input video; Cut attribute information input means for inputting cut attribute information corresponding to the cut image between the edit cut points, search information input means for inputting search information for searching for a video in the video storage unit, A comparison unit that compares the search information with the cut attribute information, and from the comparison result obtained by the comparison unit, the cut image having the cut attribute information that matches the input search information is output from the image storage unit. Extracting means for extracting and outputting,
Is adopted.

According to this configuration, the number of cut attributes can be reduced because the cut attribute information can be stored in units of cut images. In addition, by dividing the input image at each edit cut point where the change in the content is large, the moving image can be appropriately divided according to the content. Further, by inputting a cut attribute for each cut image in which the content is expected to change greatly, a desired search can be performed with an accuracy close to having cut attribute information for all image information. Can be.

According to a second aspect of the present invention, in the image search device according to the first aspect, the cut attribute information corresponding to the cut image is input to the image corresponding to the position of the edit cut point. Take.

According to this configuration, by inputting the cut attribute to the still image at the edit cut point where the content of the image greatly changes, the cut attribute appropriately representing the feature of the image can be entered. As a result, an image search can be reliably performed.

According to a third aspect of the present invention, in the image retrieval apparatus according to the first aspect or the second aspect, the syntax of an input character string inputted by the retrieval information input means is analyzed, and A syntactic analysis means for extracting search information is provided, and the comparing means adopts a configuration for comparing the search information obtained by the syntactic analysis means with the cut attribute information storage means.

According to this configuration, since search information and cut attribute information can be extracted by inputting a natural language, a video corresponding to an arbitrary input natural language can be searched. This allows the operator to input various types of information in the language he or she usually uses, so that the operator can easily input search information and cut attribute information.

According to a fourth aspect of the present invention, in the image retrieval apparatus according to any one of the first to third aspects, a plurality of cut images having the cut attribute information corresponding to all of the inputted search information are provided. In some cases, the plurality of cut images are displayed on a display unit, and one cut image selected from the displayed plurality of cut images is output.

With this configuration, even when there are a plurality of cut images including cut attribute information corresponding to the inputted search information, the operator can view appropriate cut images as a plurality of candidates while viewing the cut images. You can search for cut images.

According to a fifth aspect of the present invention, in the image retrieval apparatus according to any one of the first to third aspects, the cut image having at least one or more of the cut attribute information corresponding to the inputted search information is provided. When there is a plurality, the cut images having the highest match with the cut attribute information input for each of the plurality of cut images are ranked from the cut image, and the cut images corresponding to the rank are displayed together and displayed. And outputting one cut image selected from the cut images.

With this configuration, even when there is no cut image including cut attribute information corresponding to all the input search information, ranking is performed in accordance with the number of matched cut attribute information, and the ranking is determined. Is displayed, the operator can search for an appropriate cut image while looking at the displayed plurality of candidates.

According to a sixth aspect of the present invention, in the image retrieval device according to the first to fifth aspects, the information input means is provided with a recognition means for converting voice or handwritten characters into a character string code. The character string code recognized by the means is used as the cut attribute information and the search information.

With this configuration, it is possible to input cut attribute information and search information by handwritten text or voice. Thus, even an operator who is not good at inputting characters by operating a keyboard or the like can easily input cut attribute information and search information.

According to a seventh aspect of the present invention, a step of detecting an edit cut point of an input image, inputting cut attribute information corresponding to a cut image between the detected edit cut points, and searching the input image Inputting search information for performing the search, comparing the input search information with the cut attribute information, and storing the cut image having the cut attribute information that matches the input search information with the image storage Extracting from the means and outputting.

According to an eighth aspect of the present invention, there is provided a computer comprising:
The edit cut point of the input video is detected, and the cut attribute information corresponding to the input cut video between the edit cut points is input, and the input search information and the cut attribute information are compared. A computer-readable storage medium storing a program for extracting the cut image having the cut attribute information matching the search information from the input image and outputting the cut image.

Hereinafter, an image retrieval apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a block diagram of the video search device in the above embodiment.

In FIG. 1, an image storage unit 1 is means for storing an externally input image, and stores a large amount of images as a database.

The edit cut point detector 2 is a means for detecting an edit cut point of a video when the video to be stored is input. The edit cut point detection method performed by the edit cut point detection unit 2 generally employs a method in which a portion where a difference between a feature amount (such as a DCT coefficient) of an input video frame and a feature amount of a subsequent frame exceeds a certain threshold is set as an edit cut point. It is possible to use an improved technique such as that disclosed in Japanese Patent Application No. 09-101429.

The edit cut point storage section 3 is means for storing position information (such as a time code or a frame number) corresponding to the edit cut point detected by the edit cut point detection section 2. The video management unit 4 assigns an address indicating the position of the video storage unit 1 to the input video, and the edit cut point detection unit 2
Is a means for associating the edit cut point detected by the above with the edit cut point address of the relevant video stored in the video storage unit 1.

The display section 5 sequentially displays cut images starting from the edit cut point detected by the edit cut point detection section 2. The display unit 5 also sequentially displays still images at the positions of the edit cut points detected by the edit cut point detection unit 2.
Note that the cut video is a video from an edit cut point to the next edit cut point. The display unit 5 also displays cut attribute information of an edit cut point, which will be described later.

The cut attribute information input section 6 is means for inputting cut attribute information for a cut image. Further, the cut attribute information storage unit 7 stores the cut attribute information input from the cut attribute information input unit 6 in comparison with the storage address of the edited cut point storage unit 3.

The search information input section 8 is a means for inputting search information for the input material video as a sentence. The cut attribute information input unit 6 and the search information input unit 8 are means capable of inputting characters such as a keyboard.

The syntax analysis section 90 performs syntax analysis on the input text input from the search information input section 8. The syntax analyzer 90 includes a morphological analyzer 9 and a linguistic morphological dictionary 10. The morphological analysis unit 9 includes a language morphological dictionary 10
The morphological analysis is performed by taking the matching with, and the content of the sentence is determined. The linguistic morphological dictionary 10 includes an independent word dictionary which is a dictionary of independent words extracted from a Japanese dictionary, and an auxiliary word dictionary in which auxiliary words are stored.

The comparing section 11 sequentially compares the contents of the sentence analyzed by the syntax analyzing section 90 with the contents of the cut attribute information storage section 7.

The judging unit 12 judges the order of the cut images having the largest sum of the corresponding cut attribute information in all the cut images of the images to which the result of the comparison by the comparing unit 11 is applicable. Further, the determination unit 12 determines the order of the edit cut point of the cut video having a large number of cut attribute information that matches from all the edit cut points of all the cut videos to which the result of the comparison by the comparison unit 11 is applicable. Then, the determination unit 12 provides the video management unit 4 with the start address of the cut video and the edit cut point address.

Based on the judgment result from the judging section 12, the picture management section 4 retrieves all or only the cut video portion from the start address of the video stored in the video storage section 1 from the start address of the cut video or the corresponding edit cut point address. Output as video output.

Based on the result determined by the determination unit 12, the display method input unit 13 displays only the cut image having the largest number of matches of the cut attribute information in all the cut images included in the relevant image as the first candidate on the display unit 5. indicate. Whether the cut video extracted as a result of the detection is output as a search video output (material video output) or the first, second,... Candidate cut video is sequentially output is determined by a display method input by the operator. Part 1
3 can be determined. Whether to display only the cut image having the largest number of cut attribute matches or to sequentially display the first candidate, the second candidate,... Of the cut image can be determined by the input from the display method input unit 13 by the operator. Whether the search result is displayed as a cut image or as a still image at an edit cut point may be displayed by the operator's display method input unit 1.
3 can be determined.

The information input from the display method input unit 13 is input to the determination unit 12, and the determination unit 12 selects a cut image output method or a cut image display method.
The display method input unit 13 is a general input unit such as a computer keyboard.

When inputting voice, handwritten characters, and the like, the recognizing unit 14 converts them into a character string code. As a result, even if the user inputs the cut attribute information and the search information by voice, handwritten characters, or the like, the recognition unit 14 can convert them into a character string code. Therefore, the cut attribute information and the search information can be input by voice, handwritten characters, and the like. A well-known voice recognition device or character recognition device can be used for inputting voice or handwritten characters. Japanese Patent Laid-Open Publication No. Sho 59-160199 discloses an example of a speech recognition apparatus, such as Fumio Maehara and others.

Hereinafter, the operation of the video search apparatus configured as described above will be described with reference to FIGS. FIG.
FIG. 8 is a diagram for explaining correspondence between an input video, an edit cut point, and cut attribute information according to the embodiment. FIG. 3 is a diagram illustrating an example of a display on the display unit according to the embodiment.

In the figure, reference numeral 21 denotes a video storage unit 1
Is the input video stored in the. The input video 21 is composed of a plurality of cut videos, where an edit cut point is detected by the edit cut point detection unit 2. The one indicated by 22 is
This is a video (still image or moving image) at the edit cut point position corresponding to the input video 22. The edit cut point video 22 is stored in the edit cut point storage unit 3. Reference numeral 23 indicates cut attribute information of each cut input corresponding to the edited cut point video 22. The cut attribute information 23 is stored in the cut attribute information storage unit 7.

An operation for registering an image having a content "Yamada and Suzuki are climbing Mt. Fuji" will be described.

The input video 21 is input to the video storage unit 1 shown in FIG. At this time, the storage address where the input video 21 is stored in the video storage unit 1 by the video management unit 4 is stored. At the same time as the input video 21 is input, the edit cut point detector 3 detects an edit cut point in editing. Generally, there are a plurality of edit cut points of the input video 21. For example, an input video 21 of about one hour has about 50 edit cut points on average.

Then, the display unit 5 first lists the first screen of the cut image as one (still image at the position of the edit cut point) or a plurality of still images at the position of the edit cut point at the same time. Display in the form of FIG. 3 shows an example of an image displayed on the display unit 5 at this time.

According to the input from the cut attribute information input unit 6 by the operator, a continuous cut display image 32 (an example of simultaneous display of four images in the figure) from the successive edit cut point to the next edit cut point is displayed on the display unit 5. Is displayed on the display screen 31. On this occasion,
According to the instruction input from the cut attribute information input unit 6, the cut image for which the cut attribute information is currently to be input is input.
Display a moving image or a still image by stopping it.

The operator inputs the relevant cut attribute information from the cut attribute information input unit 6 in accordance with the content of the cut image displayed on the display screen 31. Then, in response to the corresponding cut image displayed on the display unit 5, the input cut attribute information is displayed and input to the cut attribute information storage unit 7. Cut image 22 and cut attribute information 2
2 is shown in FIG. One embodiment of the display of the cut image 22 and the display of the cut attribute information 23 is shown by 31 and 33 in FIG.

In this case, for example, if the first cut image is titled “Future of Shizuoka Prefecture”, the operator can use, for example, “Title”, “Shizuoka”, “Future” as cut attribute information.
And so on. And in the third cut, "Title"
“Shizuoka” and “feature” are registered in the cut attribute information storage unit 7 as cut attribute information.

If the image of "Mt. Fuji" is shown in the second cut image, the operator judges that the cut attribute information is related to, for example, "Mt. Fuji", "Best in Japan", "High mountain", and the like. Enter the cut attribute. In the second cut image, “Mt. Fuji”, “Best in Japan”, and “High Mountain” are registered in the cut attribute information storage unit 7 as cut attribute information.

If the characters "Yamada" and "Suzuki" appear in the third cut image, the operator also inputs this as cut attribute information. And in the third cut, "Yamada"
“Suzuki” is registered in the cut attribute information storage unit 7 as cut attribute information.

If there is a cut image in which two persons are climbing in the fourth cut image, the operator inputs cut attribute information “climbing” from the cut attribute information input unit 6. For the fourth cut, “climbing” is registered in the cut attribute information storage unit 7 as cut attribute information.

The above operation is performed for all cut images of the input image 21. Actually, other various cut attribute information is input according to the judgment of the operator.

Next, the operation for retrieving a video will be described. As an example, consider a case where a sentence “Yamada and Suzuki are climbing Mt. Fuji” is input from the search information input unit 9.

When "Yamada and Suzuki are climbing Mt. Fuji" is input to the search information input unit 8, the morphological analysis unit 9
Refers to the independent word dictionary and the attached word dictionary of the language morphological dictionary 10. "Mt. Fuji" and "Yamada" in the independent dictionary
Independent words such as "Suzuki" and "climbing" are included in the attached dictionary,
Includes the attached words "te", "ni", "no", "wo", "ga", and "is". The morphological analysis unit 9 compares input sentences from the search information input unit 8 in order from the independent word dictionary, and regards a result of the matching as an analysis result. The morphological analysis unit 9
If the input sentence from the search information input unit 8 matches a plurality of independent words in the independent word dictionary, the longest independent word is used as the matching result.

Next, the morphological analysis unit 9 performs matching by using an attached word dictionary for a part of the sentence input from the search information input unit 8 from which the independent word is omitted, and determines an attached word.
Next, the morphological analysis unit 9 checks whether or not it is possible to combine the attached word and the independent word with respect to the attached word cut attribute. As a result, if a pair of a connectable independent word and an adjunct word can be determined, the syntactic analysis ends, and if it cannot be determined, the same operation is performed on the matched second candidate adjunct word. Then, the morphological analysis unit 9 outputs “Mt. Fuji”, “Yamada”, “Suzuki”, and “climbing” as independent words.

As described above, the morphological analysis unit 9 can analyze an independent word from an input natural language.

The comparison unit 11 compares the independent word output from the morphological analysis unit 9 with the character stored in the cut attribute information storage unit 7. Then, the comparison unit 11 outputs the address of the video corresponding to the edit cut point in which “Mt. Fuji” is stored as the cut attribute information and the position of the edit cut point to the determination unit 12 as a cut video candidate. Less than,
The comparing unit 11 repeats the same operation as “Yamada”, “Suzuki”
Select the edit cut point corresponding to and the storage address of the cut image. Then, the comparison unit 11 performs the above operation on all cut images stored in the image storage unit 1.

Naturally, since a large number of cut images are stored in the image storage unit 1, the number of cut images containing the same cut attribute information as the input search information is not limited to one, but may be plural. In addition, there is not always a cut video including all of the cut attribute information.

Accordingly, the determination unit 12 determines, based on the result of the comparison unit 11, the sum of the number of pieces of cut attribute information matching in all the cut images of the corresponding video, and the matching of the registered cut images with the corresponding cut attribute information. The candidate ranking is determined in the order of the most-cut video, such as the first candidate and the second candidate.

Then, the determination unit 12 displays the determined candidate order on the display unit 5 in accordance with the cut image. Alternatively, the determination unit 12 displays only the cut images with the highest matching of the cut attribute information on the display unit 5 in order. in this case,
Whether to display only the first candidate of the cut image on the display unit 5, display the cut images of a plurality of candidates, display only the image of the first candidate cut unit, or display the image of the plurality of cut units. Is displayed by the operator in the display method input unit 1.
3 can be selected by inputting that effect. This selection method can be performed by a known means used for a computer or the like.

Then, according to the display content, the operator selects the required video from the beginning or from the cut video containing the corresponding cut attribute information and arbitrarily selects the required video by inputting from the display information input unit 13. Take out as.

As described above, even when there is no cut image including cut attribute information corresponding to all the input search information, ranking is performed in accordance with the number of matched cut attribute information, and ranking is performed. By displaying the result, the operator can search for an appropriate cut image while viewing a plurality of candidates displayed on the display unit 5.

As a result of ranking in this way, even when there are a plurality of cut images as the first candidates, a plurality of cut images as the first candidates are displayed on the display unit 5 and the operation is performed. The user can search for an appropriate cut image while viewing the plurality of first candidate cut images displayed on the display unit 5.

Further, as a result of ranking, a cut image as a first candidate may be output as a search image without the operator making a selection.

Even when there are a plurality of cut images including cut attribute information corresponding to the input search information, a plurality of cut images including cut attribute information corresponding to the display unit 5 are displayed, and Can search for an appropriate cut image while viewing a plurality of candidates displayed on the display unit 5.

In the above embodiment, the cut attribute information input unit 6, the search information input unit 8, and the display method input unit 13
Although shown in FIG. 1 as separate input devices, these can be realized by a computer keyboard or the like, and can share the same input means.

Further, by using the recognizing unit 14, it is possible to input cut attribute information and search information by handwritten text or voice. Thus, even an operator who is not good at inputting characters by operating a keyboard or the like can easily input cut attribute information and search information.

In the above-described embodiment, the operations of the edit cut point detection unit 2, the video management unit 4, the syntax analysis unit 90, the comparison unit 11, and the determination unit 12 are implemented as programs as CD-Rs.
Stored in a storage medium such as an OM, and an edit cut point detection unit 2;
Video management unit 4, syntax analysis unit 90, comparison unit 11, determination unit 1
Assuming that the operation of step 2 is realized by software using a computer, it is assumed that a storage unit of the computer is used for the video storage unit 1, the edit cut point storage unit 3, and the cut attribute information storage unit 7. I assume.

However, the present invention is not limited to this, and the video detecting apparatus according to the above-described embodiment may be a dedicated video processing device having the functions of the edit cut point detecting unit 2, the video managing unit 4, the syntax analyzing unit 90, the comparing unit 11, and the determining unit 12. It is also possible to realize by the hardware circuit of FIG.

Further, in the above-described embodiment, the description has been made of the form in which the material video is input from the outside. However, the material video may be stored in advance in a storage medium such as a CD-ROM together with the program. In the above embodiment,
In the above description, the cut attribute information is input by the operator later, but the cut attribute information may be input in advance by the creator of the material video.

As described above, according to the above embodiment,
Since a cut attribute is provided for each cut image, the number of cut attributes can be suppressed even for a large-capacity moving image stored image. In addition, by dividing the moving image at each edit cut point where the change in the content is large, the moving image can be appropriately divided according to the content. Further, by inputting a cut attribute for each cut image in which the content is expected to change greatly, a desired search can be performed with an accuracy close to having cut attribute information for all image information. Can be.

Further, in the above embodiment, it is also possible to input a cut attribute for a still image at an edit cut point. As described above, by inputting the cut attribute to the still image at the edit cut point where the change of the content of the image is large, it is possible to insert the cut attribute appropriately representing the feature of the image. As a result, an image search can be reliably performed.

Further, according to the above embodiment, independent words can be extracted from a natural language, so that a video corresponding to any input natural language can be searched. That is, in the case of an image search, for example, a video can be searched from a cut attribute described in a series of sentences such as “Yamada is climbing Mt. Fuji”. This allows the operator to input various types of information in the language he or she usually uses, so that the operator can easily input cut attributes and search information.

Further, it is possible to directly search for a video portion having specific cut attribute information in the relevant video.
That is, in the case of image search, for example, when "Mt. Fuji" is input, several cuts including "Mt. Fuji" can be directly searched and rubbed.

Further, according to the above embodiment, by using the pattern recognition unit 14, it is possible to input cut attributes and search information by handwritten character recognition and voice input.

[0079]

As described above, according to the present invention, it is possible to efficiently input a cut attribute (keyword) accurately representing the content of a moving image. Further, according to the present invention, a video can be efficiently searched for from a sentence in a natural language. Further, according to the present invention, it is possible to directly extract a cut image closest to the words of the input sentence from the images. As a result, a large-capacity video database can be searched quickly and efficiently.

[Brief description of the drawings]

FIG. 1 is a block diagram of a video search device according to an embodiment of the present invention.

FIG. 2 is a view for explaining the correspondence between an input video, an edit cut point, and cut attribute information according to the embodiment;

FIG. 3 is a view for explaining an example of display of an edit cut point, a video cut and its cut attribute information according to the embodiment;

FIG. 4 is a block diagram of a conventional image search system.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Video storage part 2 Edit cut point detection part 3 Edit cut point storage part 4 Video management part 5 Display part 6 Cut attribute information input part 7 Cut attribute information storage part 8 Search information input part 11 Comparison part 12 Judgment part 13 Display method input Part 14 Recognition part 90 Syntax analysis part

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H04N 5/76 G10L 3/00 551G 5/91 H04N 5/91 NF Term (Reference) 5B075 ND12 NK02 NK25 PP02 PP03 PP07 PP13 PP24 PQ02 PQ46 PQ72 5C052 AB02 AB04 AC08 CC06 CC20 EE03 5C053 FA14 HA29 HA40 JA01 JA16 JA30 KA26 LA06 LA11 5D015 KK01 KK02

Claims (8)

[Claims] 1. An image storage means for storing an input image, an edit cut point detecting means for detecting an edit cut point of the input image, and cut attribute information corresponding to a cut image between the detected edit cut points. Cut attribute information input means for inputting, search information input means for inputting search information for searching for a video in the video storage unit, and comparison means for comparing the input search information with the cut attribute information; ,
Judging means for extracting, from the video accumulating means, the cut image having the cut attribute information that matches the input search information from the comparison result by the comparing means, and outputting the cut image. Image search device. 2. The image search apparatus according to claim 1, wherein the cut attribute information corresponding to the cut video is input for a video corresponding to the position of the edit cut point. 3. A syntactic analysis means for analyzing the syntax of an input character string input by the search information input means and extracting the search information, wherein the comparison means is provided by the syntax analysis means. 3. The image search device according to claim 1, wherein the search information obtained is compared with the cut attribute information storage unit. 4. When there are a plurality of cut images having the cut attribute information corresponding to all the inputted search information, the plurality of cut images are displayed on a display means, and the displayed plurality of cut images are displayed. The image search device according to claim 1, wherein one of the cut images selected from the images is output. 5. When there is a plurality of cut images having at least one or more pieces of the cut attribute information corresponding to the input search information, the cut image includes a plurality of cut attribute information corresponding to each of the plurality of cut images. The cut image having the highest number of matches is ranked, the cut images corresponding to the rank are displayed together, and one cut image selected from the displayed cut images is output. The image search device according to claim 1. 6. The information input means includes a recognition means for converting voice or handwritten characters into a character string code, and the character string code recognized by the recognition means is used as the cut attribute information and the search information. The image search device according to any one of claims 1 to 5, wherein 7. A step of detecting an edit cut point of an input video, inputting cut attribute information corresponding to a cut video between the detected edit cut points, and inputting search information for searching the input video. And comparing the input search information with the cut attribute information, extracting the cut image having the cut attribute information that matches the input search information from the image storage unit, and outputting the cut image. An image search method comprising: 8. A computer which detects an edit cut point of an input video, and calculates a relationship between input cut attribute information corresponding to a cut video between the edit cut points, input search information, and the cut attribute information. A computer-readable storage medium storing a program for causing a comparison, extracting the cut video having the cut attribute information that matches the input search information from the input image, and outputting the cut video.