JP2006053802A - Image type determining method, image type determining device, and image type determining program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image type determining method, image type determining device, and image type determining program for supporting classification and selection of an image type on the basis of a font of a telop character (including modification information). <P>SOLUTION: The image type determining method is provided with a frame extracting step (S1) for extracting a frame with the telop character displayed from an image of an input frame construction, a character recognizing step (S6) for recognizing the telop character of the frame with the telop character displayed, a font recognizing step (S10) for recognizing the font of the recognized character, and an image determining step (S16) for determining the image type of the input image by referring to a font-image type associating database for associating the font of the character with the image type. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a video type determination method, a video type determination device, and a video type determination program.

  With the advance of computer technology, telop characters are inserted in various fonts in images. Character modification information such as the telop font and its color and size often reflects the atmosphere of the program.

  Conventionally, a telop character recognition technique for extracting a telop character from an image and discriminating the character is known. This technique is used for understanding videos or automatically organizing videos. Specifically, video indexing in which the time when a telop character appears is regarded as a video break, and characters determined by performing character string recognition are added to video as metadata.

For example, Patent Document 1 discloses a technique for detecting a frame in which a telop character is displayed from a video with high accuracy.
FIG. 1 shown in Patent Document 1 shows an example of a telop character display frame detection unit of a telop character display frame detection device. The telop character display frame detection unit includes a partial block division unit 11, an unprocessed partial block determination unit 12, an edge detection unit 13, a scan control unit 14, an edge gradient direction determination unit 15, an inter-edge luminance change calculation unit 16, and an edge pair account. It comprises a unit 17, an edge pair total number calculating unit 18, an edge pair number determining unit 19, a flag information setting unit 20, and a plug information determining unit 21. The partial block dividing unit 11 divides the input frame into partial blocks. FIG. 2 shows an example in which an input frame is divided into 8 partial blocks. The unprocessed partial block determination unit 12 determines whether flag information setting processing has been completed for the target partial block among the divided partial blocks. The edge detection unit 13 uses a predetermined method, and includes an edge with direction information of an edge pixel that is a part of a pixel in which the luminance value changes locally and discontinuously among a plurality of pixels constituting the image. Perform detection processing. The scan control unit 14 designates a scanning line (horizontal or vertical) for calculating an edge pair. As shown in FIG. 2, the edge gradient direction determination unit 15 determines the gradient direction (up and down) of the detected edge pixel in the scanning line direction. The inter-edge luminance change calculation unit 16 calculates a change in luminance value between adjacent edge pixels as shown in FIG. The edge pair account unit 17 counts the edges as edge pairs when the gradient directions of adjacent edges are reverse (in the order of up and down or in the order of down and up), and the luminance change between the edges is small. In the example of FIG. 2, the number of edge pairs is thirteen. The edge pair total number calculation unit 18 calculates the total of edge pairs calculated for each scanning direction. The edge pair number determination unit 19 compares the total of edge pairs with a preset value. When the detected number of edge pairs is equal to or greater than a predetermined value, the flag information setting unit 20 determines that the character is a telop character and sets flag information in the partial block. The flag information determination unit 21 outputs the frame as a telop character display frame.

  Patent Document 2 discloses a technique for extracting a character region from an image including a telop character or the like.

  In FIG. 4 shown in Patent Document 2, reference numeral 2-10 denotes a horizontal line unit binarization unit, which performs luminance binarization within each horizontal line in the image. Reference numeral 2-11 denotes a vertical line unit binarization unit, which performs luminance binarization within each vertical line in the image. 2-12 is a binarization result integration unit, which compares two binary images obtained by the horizontal line unit binarization unit 2-10 and the vertical line unit binarization unit 2-11, Both connected components that are present at the same position and have the same area value are determined as character regions, and the other connected components are determined as background noise portions and removed from the image. A processing control unit 2-13 controls the execution order of the units 2-10 to 12-12. FIG. 5 is a block diagram of the horizontal line unit binarization unit 2-10 in FIG. In FIG. 5, reference numeral 2-14 denotes a horizontal line maximum luminance area detection unit, which examines the luminance distribution in a certain horizontal line in the image and connects the luminance value higher by a preset value locally in the horizontal line. A pixel area is detected. Reference numeral 2-15 denotes a high-luminance character area extraction unit which, on a certain horizontal line in the image, has a plurality of pixels near the left and right ends of the connected pixel area obtained by the horizontal line luminance maximum area detection unit 2-14. A pixel having the maximum absolute value of the luminance gradient in the horizontal direction from the inside is obtained on each of the left and right sides, and a range having the pixel as both ends is extracted as a high-luminance character region. Reference numeral 2-16 denotes a horizontal line minimum luminance area detection unit, which examines the luminance distribution in a certain horizontal line in the image and detects a connected pixel area whose luminance value is locally lower in the horizontal line by a preset value. To do. Reference numeral 2-17 denotes a low-luminance character area extraction unit which, on a certain horizontal line in the image, includes a plurality of pixels near the left and right ends of the connected pixel area obtained by the horizontal line luminance minimum area detection unit 2-16. A pixel having the maximum absolute value of the luminance gradient in the horizontal direction from the inside is obtained on each of the left and right sides, and a range having these pixels at both ends is extracted as a low-luminance character region. 2-18 is a binarization result integration unit for each horizontal line, integrating the character regions on each horizontal line obtained by the high luminance character region extraction unit 2-15 and the low luminance character region extraction unit 2-17, Create a character area image for the entire image. Similarly, the vertical line unit binarization unit 2-11 integrates the character regions on the vertical lines obtained by the high luminance character region extraction unit and the low luminance character region extraction unit, and converts the character region image as the entire image create.

  Thereby, a character area is extracted from an image including a telop character or the like from the binarization result integration unit for each horizontal line and the binarization result integration unit on each vertical line.

  Further, a technique is known in which the read characters read as image data are replaced with approximate fonts built in a personal computer, a printing apparatus, or the like.

  For example, in Patent Document 3, a plurality of image data created using a plurality of built-in fonts and image data of read characters are compared with each other to detect differences, and those having the least differences are approximated fonts. The technique to select as is disclosed.

In FIG. 6 shown in Patent Document 3, a blocking unit 3-21 blocks image data of a document read by the scanner device 3-3 for each character string. The in-block character separating unit 3-22 separates a plurality of read characters in each character string blocked by the blocking unit 3-21 for each read character. The read character recognition means 3-23 recognizes each read character separated by the in-block character separation means 3-22 using the existing OCR (Optical Character Reader) technology. Specifically, the read character recognition unit 3-23 converts, for example, image data of each read character into a vector amount, and collates the converted vector amount with a built-in database configured for each character category. , A character having many similarities is recognized as the read character. Each read character recognized by the read character recognizing means 3-23 is associated with code data such as a JIS code and temporarily stored in, for example, the RAM 3-12. The read character size determination unit 3-24 determines the size of each read character recognized by the read character recognition unit 3-23. The image data creation means 3-25 uses several types of built-in fonts stored in the HDD 3-14 to determine the read character size of the image data having the same character as the read character recognized by the read character recognition means 3-23. Each character is created with a size substantially the same as the size of the read character determined by means 3-24. In other words, the read character that has been character-recognized is temporarily stored in the RAM 3-12 in association with code data such as a JIS code, as described above. On the other hand, data of several types of built-in fonts are usually stored in the HDD 3-14 as vector data. Based on the code data corresponding to the read character, the image data creation means 25 extracts vector data of the same character as the read character for each built-in font, and these vector data have the same size as the read character. The image data is converted into image data such as bitmap data and expanded in the RAM 3-12. The approximate font selection means 3-26 detects the difference between the image data creation means 3-25 and the image data created by using several kinds of built-in fonts by comparing the pattern with the image data of the read characters. Then, the built-in font of the image data with the least difference from the image data of the read character is selected as an approximate font for forming an image corresponding to the read character.
JP-A-11-178007 JP 2000-194851 A JP2003-50971

  However, with the conventional telop recognition technology, the type of program cannot be determined even if characters are recognized. For example, even if the telop “Unidentified organism found” is included, it cannot be distinguished from the contents of the text whether the program is news or a program as a joke in variety.

  The telop character font and character modification information often reflect the atmosphere (or content) of the program, but conventional telop character recognition recognizes the font and character modification information. not going. For this reason, even if characters are recognized, it is impossible to judge from the telop font to the program characteristics.

  Even if the video is expressed as “video”, the type and genre are diverse, such as news, sports, drama, documentary, variety, and movie. Therefore, if various types of video content are classified and managed appropriately, video content can be easily searched.

  By the way, although the type and genre of the video are important classification items, it is necessary to understand the content of the video in order to classify them. There is a problem that becomes larger.

  The problem to be solved by the present invention is to provide a video type determination method, a video type determination device, and a video type determination program that support classification or selection of video types based on the font of telop characters (including character modification information). It is to be.

  In order to solve the above problems, the present invention employs means for solving the problems having the following characteristics.

  The invention described in claim 1 is recognized as a frame extracting means for extracting a frame in which characters are displayed from an input image having a frame structure, and a character recognizing means for recognizing characters in the frame in which the characters are displayed. A font recognizing means for recognizing a font of a character, and a video determining means for determining a video type of the inputted video with reference to a font / video type correspondence database that associates the font of the character with a video type. A video type determination device characterized in that it is provided.

  Note that “font” here means not only a character font but also one including “character font and character modification information”. Character modification information is information that modifies characters, such as the color and size of characters. A broad concept of “font” is one of character modification information, and it is complicated in a strict sense. This is because the font cannot distinguish between “character font” and “character modification information”. Therefore, the “font” here means not only a so-called “character font”, but also includes “character font and character modification information”. However, the case of only “character modification information” excluding the so-called “character font” is not included.

  According to a second aspect of the present invention, in the video type determination apparatus according to the first aspect, a characteristic font for determining that the font recognized most by the font recognition means is a characteristic font of the video is determined. Characterized in that the video determination means determines the video type based on the characteristic font determined by the characteristic font determination means and the font / video type correspondence database. To do.

  According to a third aspect of the present invention, in the video type determining apparatus according to the first aspect, the font recognized by the font recognizing unit is integrated for each font, and the font used to calculate the font usage rate of the input video is calculated. A ratio calculation unit is provided, wherein the video determination unit determines the video type based on the font ratio calculated by the font ratio calculation unit and the font / video type correspondence database.

  According to a fourth aspect of the present invention, in the video type determination apparatus according to the third aspect, the font / video type correspondence database includes data on a font use ratio for each video genre in advance.

  The invention described in claim 5 is recognized as a frame extracting step for extracting a frame in which characters are displayed from an input image having a frame structure, and a character recognizing step for recognizing characters in the frame in which characters are displayed. A font recognition step for recognizing the font of the character, and a video determination step for determining the video type of the input video with reference to a font / video type correspondence database that associates the font of the character with the video type. A video type determination method characterized by comprising the above.

  The invention described in claim 6 is a frame extracting means for extracting a frame in which characters are displayed from an input image having a frame structure, a character recognizing means for recognizing characters in a frame in which characters are displayed, and a recognized character. The computer functions as image recognition means for determining the image type of the input image with reference to a font / image type correspondence database that associates the font of the character with the image type. This is a video type determination program.

  According to a seventh aspect of the present invention, in the video type determination program according to the sixth aspect, as the characteristic font determination means for determining that the font most recognized by the font recognition means is the characteristic font of the video. The computer is made to function, and the video determination unit determines the video type based on the characteristic font determined by the characteristic font determination unit and the font / video type correspondence database.

  The invention described in claim 8 is the video type determination program according to claim 6, wherein the fonts recognized by the font recognition means are integrated for each font, and the font usage ratio of the input video is calculated. The computer is made to function as a ratio calculation means, and the video determination means determines the video type based on the font ratio calculated by the font ratio calculation means and the font / video type correspondence database.

  It is possible to provide a video type determination method, a video type determination device, and a video type determination program that support classification or selection of a video type based on a telop font (including modification information).

  FIG. 7 shows an example of the video type determination apparatus of the present invention. 7 includes a video input device 42 that inputs a video material 41 having a plurality of frames, a frame extraction device 43 that extracts a frame in which a telop character is displayed from the input video, and a telop character is A character recognition device 44 for recognizing the telop character of the displayed frame, a character recognition information database 45 used for character recognition, a font determination device 46 for determining the font of the recognized character, a font image, etc. A font information database 47 in which font information is stored, a video type determination device 48 that determines the video type of the input video, a font / video type correspondence database 49 that associates fonts with video types, a meta The data generation device 50 and the metadata management device 51 are included.

  The video material 41 may be a video material stored in various video media such as HDCAM, Digital Beta Cam, DVCAM, Beta Cam, DV, VHS, and DVD, or from broadcast waves such as terrestrial and digital broadcasts. The video may be used directly. Further, it may be one image corresponding to one frame.

  The video input device 42 is a device that acquires video content from various media. It may have functions such as VTR device control and video digitization.

  The frame extracting device 43 automatically extracts a frame in which telop characters are displayed from the input video. The character recognition device 44 refers to the character recognition information database 45 and automatically recognizes the telop character of the frame in which the telop character extracted by the frame extraction device 43 is displayed.

  The font recognizing device 46 refers to the font information database 47 and recognizes the character font and / or character modification information recognized by the character recognizing device 44. The video type determination device 48 automatically determines the video type (genre etc.) from the recognition results of the character recognition device 44 and the font recognition device 46 and the font / video type correspondence database 47 in which the font video type is associated. To do.

  The metadata generation device 50 generates and assigns the video type (genre and the like) determined by the video type determination device 48 as one of the target video metadata.

  Note that one video type may be given for each input video, or may be given at regular intervals. In the case where the fonts are given at regular intervals, the font usage rate is calculated at regular intervals.

  The metadata in the case of video is, for example, genre, video title, file name, producer, name of performer, copyright, time, number of scenes, and the like. These metadata are described by, for example, MPEG (Moving Picture Experts Group) -7.

  The metadata management device 50 is a disk for managing metadata. The correspondence between video content and metadata is managed by issuing a content ID.

  By managing the metadata in this way, it is possible to perform a search using the metadata. For example, it is possible for a user to search for a desired video with the metadata management apparatus and acquire specific video data.

  Next, a video type determination method will be described using the flowchart of FIG. First, video data to be recognized is input from the video input means (S1). In step S1, the frame extracting means (frame extracting device) detects the frame of the video on which the telop is displayed. Since the same telop character is displayed between certain consecutive frames, the same telop character is displayed and the section is detected. Note that the detection of the video frame in step S1 can be performed by the telop character display frame detection device described in Patent Document 1 described above.

  In step S2, the frame area extracting means extracts the telop area displayed in the frame. In step S3, the character separation means (part of the character recognition device) separates the extracted telop area for each character. Here, it is assumed that the image data of the separated characters is a read character (n). The telop area extraction in step S2 can be performed by the character area extraction technique described in Patent Document 2 described above.

  In step S4, the character information determination means (a part of the character recognition device) determines the size and color of the nth character of the read character. In step S5, the character size determination means determines whether the size of the nth read character is equal to or larger than a certain size (m × k). If it is less than a certain size (mxk), it is considered that the size of the character is too small to determine the font, the font of the character is not determined, and the process proceeds to step S11.

  If the size of the nth read character is equal to or larger than a certain size (m × k), in step S6, the character recognition means (a part of the character recognition device) refers to the character recognition information database built in the system, Recognize characters. Note that the character recognition in step S6 can be performed by the read character recognition means described in Patent Document 3 described above.

  In step S7, the font data recognition means (font data recognition device) acquires the font data of the nth recognized character from the font information database built in the system. For example, when the n-th read character is recognized as “A” in step S6, all “A” font data such as Mincho, Gothic, Round Gothic,... Are acquired from the font information database. In step S8, the font recognition unit generates image data corresponding to the size and color of the nth read character based on the font size and color acquired in step S4. Next, in step S9, the font data recognizing means performs pattern comparison between the created image data and the read image data of the read character, and detects a difference. In step S10, the font of the image data with the least difference is selected as the font of the character. Note that the font selection in step S10 can be performed by the approximate font selection means described in Patent Document 3 described above.

  In step S11, it is determined whether or not font recognition has been performed for all of the read characters that have been read. At this time, if the font has not been recognized for all of the read characters that have been read, the value of n is decremented in step S12 and the process returns to step S4. Process.

  On the other hand, if it is determined in step S11 that the font has been recognized for all of the read characters that have been read, in step S13, the font most recognized in one telop is the font that matches the telop. Selected as.

For example, in FIG. 9, one telop has a plurality of fonts. The telop in FIG.
・ Gothic: 3
・ Pop body: 1
Therefore, the telop font is determined to be Gothic.

  In step S14, it is determined whether or not telop detection has been completed for all frames of the video. If not, in step S15, the process proceeds to a frame where the next different telop is displayed, and returns to step S1. The subsequent processing is repeated.

  On the other hand, if it is determined in step S14 that telop detection has been completed for all frames of the video, the process proceeds to step S16.

  In step 16, the video type determination means (video type determination device) determines the video type of the input video with reference to a font / video type correspondence database in which character fonts and video types are associated with each other.

  Here, two examples of the video type determination method in step 16 will be described.

  In the first method, it is determined that the most recognized font in one entire telop selected in step S13 is a characteristic font in the entire video, and the font and video type corresponding to FIG. This is a method for determining a video type of an input video by referring to a database.

  Here, for example, as shown in FIG. 10, a case is considered where the video is composed of three scenes with telops. In FIG. 10, the font of telop 1 is Gothic, the font of telop 2 is pop, and the font of telop 3 is Gothic.

in this case,
・ Gothic: 2
・ Pop body: 1
Therefore, since the font most recognized as the font of each telop in the entire video is Gothic, it is determined that the characteristic font is Gothic.

  If a characteristic font is found, the video type of the input video can be determined with reference to the font / video type correspondence database of FIG.

  As shown in FIG. 11, the font / video type correspondence database associates fonts with video types. For example, in FIG. 11, the round Gothic font, POP font, logo character, design font, etc. are a variety, and Mincho, Gothic, textbook, font, etc. are a news font and a subtitle font. Is associated with the movie system.

  Therefore, in step S13, if the most recognized font in the telop is a round Gothic font, the video in which the telop has flowed is determined to be a variety system, and if it is a text type, it is determined to be a news system. In the case of a subtitle typeface, it is determined that the movie type.

  Since the video shown in FIG. 10 is Gothic, it is determined that the input video is a news system.

  The second method is a method of determining the font for each telop over the entire video and determining the most reliable video type from the font ratio based on the ratio of the telop font used in the entire video.

  In the second method, the font / video type correspondence database of FIG. 12 is used as the font / video type correspondence database.

  In the font / video type correspondence database of FIG. 12, the probability ratio (%) is stored for the video type. In FIG. 12, when the video is a variety, the round Gothic type is 100, the POP type is 100, the round type is 100, the logo type is 90, the design type is 80, the Mincho type is 0, the Gothic type is 20, and the textbook. The body is 0, the subtitle font is 0 ..., and in the case of news, the round Gothic font is 0, the POP font is 0, the round font is 0, the logo is 10, the design font is 20, and the Mincho font is 100, Gothic 80, textbook 100, subtitle font 0 ... are stored.

  Using this database, the most reliable video type is determined. For example, as shown in FIG. 13, it is assumed that the ratio of the telop font used in the entire video is calculated and the ratio is as follows.

・ Gothic: 50%
・ Design font: 10%
・ Mincho: 40%
From the font / video type correspondence database in FIG. 12, the value of the video type possibility is obtained as follows.

The value α of the possibility that the video is a variety system is
α = 50% × 20% (Gothic)
+ 10% x 80% (design body)
+ 10% x 80% (Mincho)
= 18%
The probability β that the video is news-related is
β = 50% × 80% (Gothic)
+ 10% x 20% (design body)
+ 10% x 100% (Mincho)
= 82%
Thereby, it is found that the video type in this case is a news system.

  FIG. 14 shows an application example of the present invention. This application example mainly includes a content input device, a metadata input device, a search device, a playback device, and a user terminal.

  The content input device includes a video input unit 62 that inputs a video material 61 having a plurality of frames, a format conversion unit 63 that converts the content of the input video, and a content management unit 64 that manages the content.

  Further, the metadata input device includes a telop recognition unit 65, a font recognition unit 66, a genre selection unit 68, a metadata generation unit 69, and a metadata management unit 70.

  The video material 41 may be a video material stored in various video media such as HDCAM, Digital Beta Cam, DVCAM, Beta Cam, DV, VHS, and DVD, or from broadcast waves such as terrestrial and digital broadcasts. The video may be used directly.

  The video input unit 62 is a device that acquires video content from various media. It may have functions such as VTR device control and video digitization.

  The format conversion unit 63 performs format conversion of the acquired video. There are various formats of video data such as MPEG, MPEG2, MPEG4, AVI, WMV, etc. at present. The format conversion unit 63 performs processing such as converting the captured video data in various formats into a format that is unified throughout the system.

  The content management unit 64 manages the acquired video. For example, media storage or disk.

  The telop recognition unit 65 automatically recognizes the displayed telop characters in the video. The font recognition unit 66 recognizes the font and / or modification information of the displayed telop characters in the video. The genre selection unit 68 automatically selects a video genre (video type) based on the characters recognized by the telop recognition unit 65 and the phone and / or character modification information recognized by the font recognition unit 66.

  The metadata generation unit 69 generates and assigns the video type (genre etc.) selected by the video genre selection unit as one of the target video metadata.

  The metadata management unit 70 is a disk for managing metadata. The correspondence between video content and metadata is managed by issuing a content ID.

  The content of the content management unit 72 and the metadata (content ID, file name, start time, end time, title, producer name, etc.) of the metadata management unit 70 are linked inside the database.

  The user terminal 73 may be a general personal computer, a PDA (Personal Digital Assistant), or a portable terminal as long as it can be connected to a network and can reproduce video content.

  The search unit 72 performs a search from the metadata management unit 10 when an inquiry is received from the user terminal 73 regarding the genre or keyword of the video. The reproduction unit 72 reproduces the video so that the user terminal 73 can view the video.

  In FIG. 14, the user of the system accesses the metadata management device 70 using the user device 73. When the user selects a video genre that the user wants to watch, the metadata management device 70 searches stored metadata and selects from the content management device 64 video data that matches the selected video genre.

  The content reproduction device 72 distributes the video data selected earlier to the user terminal 73.

  The user terminal 73 acquires and browses video data distributed by the playback device 12.

  Although a detailed description is omitted, FIG. 15 shows the relationship between content and metadata. As shown in FIG. 15, by managing the content and metadata, it is possible to perform keyword search, quick reference of contents, preview viewing, and search of physical information amount from the user device 73.

  7 and 14, each block may be configured by hardware (independently or in combination). In addition, each block may be a block that represents a function played by a program related to a video type installed in a computer.

  Further, the video type determination method shown in FIG. 8 may be processed as a program, and this processing program may be installed in a computer so that the video type is determined.

  In the best mode for carrying out the invention described above, the video type determination is determined based on the font. However, the video type determination is performed in consideration of the modification information of characters other than the font. You may go. In this case, the character recognition information other than the font is recognized by font recognition, and the font / video type correspondence database stores information that associates the font of the character and the character modification information with the video type.

It is FIG. (1) for demonstrating a prior art example. It is FIG. (2) for demonstrating a prior art example. It is FIG. (3) for demonstrating a prior art example. It is FIG. (4) for demonstrating a prior art example. It is FIG. (5) for demonstrating a prior art example. It is FIG. (1) for demonstrating a prior art example. It is a figure for demonstrating the example of the video classification determination apparatus of this invention. It is an example of the flowchart for demonstrating the image | video classification determination method. This is a case where a single telop has a plurality of fonts. This is a case where the video is composed of three scenes with telops. It is a figure for demonstrating the example (the 1) of a font and video classification correspondence database. It is a figure for demonstrating the example (the 2) of a font and image | video classification corresponding | compatible database. It is a figure for demonstrating the telop font used in the whole image | video. It is a figure which shows the example of application of this invention. It is a figure which shows the relationship between a content and metadata.

Explanation of symbols

41, 61 Video material 42 Video input device 43 Frame extraction device 44 Character recognition device 45 Character recognition information database 46 Font judgment device 47 Font information database 48 Video type judgment device 49 Font / video type correspondence database 50 Metadata generation device 51 Meta Data management device 62 Video input unit 63 Format conversion unit 64 Content management unit 65 Telop recognition unit 66 Font recognition unit 67 Font dictionary 68 Genre selection unit (video type determination unit)
69 Metadata generation unit 70 Metadata management device 71 Search unit 72 Playback unit 73 User terminal

Claims (8)

A frame extracting means for extracting a frame on which characters are displayed from the input frame configuration video;
Character recognition means for recognizing the character of the frame in which the character is displayed;
A font recognition means for recognizing the font of the recognized character;
A video type determination apparatus comprising: a video determination unit that determines a video type of the input video by referring to a font / video type correspondence database in which character fonts and video types are associated with each other. Characteristic font determination means for determining that the font recognized most by the font recognition means is a characteristic font of the video,
2. The video type determination according to claim 1, wherein the video determination unit determines the video type based on the characteristic font determined by the characteristic font determination unit and the font / video type correspondence database. apparatus. A font ratio calculating means for accumulating the fonts recognized by the font recognition means for each font and calculating a font usage ratio of the input video;
2. The video type determination apparatus according to claim 1, wherein the video determination unit determines a video type based on a font ratio calculated by the font ratio calculation unit and the font / video type correspondence database.   4. The video type determination apparatus according to claim 3, wherein the font / video type correspondence database includes data on a font use ratio for each video genre in advance. A frame extracting step of extracting a frame in which characters are displayed from the input frame configuration video;
A character recognition step for recognizing the character of the frame in which the character is displayed;
A font recognition step for recognizing the font of the recognized character;
A video type determination method comprising: a video determination step of determining a video type of the input video by referring to a font / video type correspondence database in which character fonts and video types are associated with each other. Frame extracting means for extracting a frame on which characters are displayed from the input frame configuration video,
Character recognition means for recognizing the character of the frame in which the character is displayed,
Font recognition means for recognizing the font of recognized characters,
A video type determination program for causing a computer to function as video determination means for determining the video type of the input video with reference to a font / video type correspondence database in which character fonts are associated with video types. A video type determination program for causing a computer to function as characteristic font determination means for determining that the font recognized most by the font recognition means is a characteristic font of the video,
7. The video type determination according to claim 6, wherein the video determination unit determines the video type based on the characteristic font determined by the characteristic font determination unit and the font / video type correspondence database. program. A video type determination program for causing a computer to function as a font ratio calculation unit that calculates a font usage ratio of an input video by accumulating fonts recognized by the font recognition unit for each font,
7. The video type determination program according to claim 6, wherein the video determination unit determines the video type based on the font ratio calculated by the font ratio calculation unit and the font / video type correspondence database.

Images