KR100707205B1 - Method and apparatus for detect play section in sports video - Google Patents

Abstract

The present invention relates to a method and apparatus for detecting a play section of a sports video by extracting frames at which play starts and frames at which play ends from a video of a sports game such as baseball, tennis, and volleyball. The method includes extracting play start keyframes among keyframes of a sports video; Detecting a field color from the extracted play start keyframes; And extracting play-end keyframes among keyframes of the sports video using the detected field color.

According to the present invention, when detecting a play section in a video of a sport and using the video summary, etc., the field color is detected using play start keyframes extracted from the video using repeatability, and the detected field color is used. By extracting the play end keyframe, it is possible to provide a method and apparatus for detecting a play section which is adaptive to all sports videos and has high accuracy.

Description

Method and apparatus for detecting play section of sports video {Method and apparatus for detect play section in sports video}

1 is a block diagram showing the overall configuration of an apparatus for detecting a play section of a sports video according to the present invention.

2A, 2B, and 2C are diagrams showing examples of play sections in a baseball, tennis, and volleyball game video, respectively.

3 is a block diagram illustrating an embodiment of a start frame extractor of FIG. 1.

FIG. 4 is a block diagram illustrating an embodiment of the candidate extracting unit of FIG. 3.

FIG. 5 is a diagram for describing repeatability between play start keyframes in a baseball game video.

FIG. 6 is a diagram illustrating an embodiment of a method in which a start keyframe group selector of FIG. 4 selects a start keyframe group from among a plurality of keyframe groups.

FIG. 7 is a block diagram illustrating an embodiment of the field color detection unit of FIG. 1.

8A, 8B, and 8C are diagrams for explaining an embodiment of a method of detecting field colors in volleyball, tennis, and baseball videos, respectively.

FIG. 9 is a block diagram illustrating an embodiment of the end frame extractor of FIG. 1.

FIG. 10 is a diagram illustrating an embodiment of a method of dividing a frame into a plurality of slices by the slice division of FIG. 9.

11 is a view for explaining an embodiment of a method of extracting a play end keyframe.

12 is a flowchart illustrating a method of detecting a play section of a sports video according to the present invention.

FIG. 13 is a flowchart illustrating an embodiment of a play start keyframe extraction step of FIG. 12.

FIG. 14 is a flowchart illustrating an embodiment of extracting play start candidate keyframes of FIG. 13.

FIG. 15 is a flowchart illustrating an embodiment of detecting the field color of FIG. 12.

FIG. 16 is a flowchart illustrating an embodiment of extracting a play end keyframe of FIG. 12.

The present invention relates to a method and apparatus for analyzing video, and more particularly, to a method and apparatus for detecting a play section in a sports video such as baseball, tennis, and volleyball.

Video playback apparatuses such as PVR (Personal Video Recorder) generally have a function of decoding and outputting encrypted video data, which is mainly used for playing a video stored in a storage device for viewing by a user. In accordance with the development of networks, digital storage devices, image compression and decompression technology, devices for storing and reproducing digital images in storage devices have been widely used.

For videos recorded on sports events, such as baseball, which require more than two hours per game, such as the pitcher's pitching scene, batter's swing scene, and the beast's muzzle scene, In other words, it is very necessary to select only the play section to play and edit quickly and easily.

The present invention has been made in an effort to provide a method and apparatus for detecting a play section, which is adaptive to all sports videos and has high accuracy, in detecting a play section of a sports video and using the video summary.

According to an aspect of the present invention, there is provided a method of detecting a play section of a sports video, the method comprising: extracting play start keyframes among keyframes of the sports video; Detecting a field color from the extracted play start keyframes; And extracting play-end keyframes among keyframes of the sports video using the detected field color.

Preferably, extracting the play start keyframes extracts play start keyframes among keyframes of the sports video by using repeatability of edge information or color information of the playstart keyframes.

Extracting the play start keyframes may include extracting play start candidate keyframes among keyframes of the sports video; Generating a play start frame model using the extracted play start candidate keyframes; And extracting play start keyframes among keyframes of the sports video using the generated play start frame model.

Preferably, extracting the play start candidate keyframes comprises: classifying keyframes of the sports video into a plurality of groups using edge information of the keyframes; Calculating a similarity between adjacent keyframes in the group for each of the classified plurality of groups, and selecting a group having the largest calculated similarity among the plurality of groups as a play start keyframe group; And extracting key frames having a similarity or more among key frames belonging to the selected play start key frame group as a play start candidate key frame.

The generating of the play start frame model may be generated using an average value of values of edge information or color information of the extracted play start candidate keyframes.

Preferably, the detecting of the field color comprises: detecting, for each of the extracted play start keyframes, a color of the maximum ratio among colors included in a predetermined lower region of the play start keyframe; Calculating the field color using the detected maximum ratio colors for each of the play start keyframes.

The calculating of the field color may include calculating, for each of the detected maximum ratio colors, a ratio of color information of the maximum ratio color and the maximum ratio color in the play start keyframe; Comparing the calculated color information and the ratio with the color information and the ratio of a predetermined base field color, extracting colors having a similarity with the predetermined base field color among the maximum ratio colors equal to or greater than a second predetermined value as candidate field colors. step; And detecting the average of the extracted candidate field colors as a field color.

Preferably, extracting the play end keyframe comprises: dividing each keyframe of the sports video into a plurality of slices; By comparing the color of the pixels included in the divided plurality of slices with the detected field color, a pixel having a similarity between the color of the pixel and the detected field color equal to or greater than a third predetermined value is determined as a field region pixel. Doing; Determining a slice including a predetermined number of pixels determined as the field region pixels among the divided slices as a field region slice; And extracting a play end keyframe from among keyframes of the sports video by using the determined number of field region slices or proximity between field region slices.

According to an aspect of the present invention, there is provided an apparatus for detecting a play interval of a sports video, including: a start frame extracting unit configured to extract play start keyframes among keyframes of the sports video; A field color detector for detecting a field color from the extracted play start keyframes; And an end frame extracting unit configured to extract play end key frames among key frames of the sports video by using the detected field color.

Preferably, the start frame extracting unit comprises: a candidate extracting unit extracting play start candidate key frames among key frames of the sports video; A start frame model generator configured to generate a play start frame model using the extracted play start candidate keyframes; And a model comparison unit configured to compare the generated play start frame model with key frames of the sports video to extract play start key frames among the key frames of the sports video.

The field color detector includes: a maximum ratio color detector configured to detect, for each of the extracted play start keyframes, a color having a maximum ratio among colors included in a predetermined lower region of the play start keyframe; And a field color calculator that calculates the field color using the maximum ratio colors detected for each of the play start keyframes.

Preferably, the end frame extraction unit comprises: a slice dividing unit for dividing each of the key frames of the sports video into a plurality of slices; By comparing the color of the pixels included in the divided plurality of slices with the detected field color, a pixel having a similarity between the color of the pixel and the detected field color equal to or greater than a third predetermined value is determined as a field region pixel. A first field pixel determination unit; A field slice determination unit which determines a slice including a predetermined number or more of pixels determined as the field region pixels among the divided slices as a field region slice; And a slice analyzer configured to extract a play end keyframe from among keyframes of the sports video by using the determined number of field region slices or proximity between field region slices.

The method for detecting a play section of the sports video may be embodied as a computer-readable recording medium that records a program to be executed in a computer.

Hereinafter, a method and apparatus for detecting a play section of a sports video according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing the overall configuration of the apparatus for detecting a play section of a sports video according to the present invention. The illustrated detection apparatus includes a shot splitter 100, a key frame extractor 110, a start frame extractor 120, It comprises a field color detection unit 130 and the end frame extraction unit 140. The operation of the play section detection apparatus shown in FIG. 1 will be described with reference to a flowchart illustrating a method for detecting a play section of a sports video according to the present invention shown in FIG. 12.

The shot dividing unit 100 detects the scene transition of the input sports video and obtains time information of the same scene section divided by the scene transition boundary time point, thereby dividing the sports video into shot (same scene section) units (step 1200). In operation 1210, the key frame extractor 110 extracts a key frame for each of the divided shots. As a method of extracting keyframes for each shot, the keyframe extractor 110 selects and extracts a fixed position of the shot, for example, the first image, the last image, or the intermediate image of the shot, or the image or image with less movement. A method of selecting a clear image or an image in which a face is clearly present may be used.

The sports video may be divided into a play section in which a game is played and a break section including content outside the game. The play section may be defined differently for each sport game. Referring to FIG. 2A, in the case of a baseball game, play is performed from the frame 200 containing the pitcher throwing the ball to the frame 210 containing the close-up scene after the batter swings or the beast catches the ball. It is preferable to define the interval.

Referring to FIGS. 2B and 2C, in the case of a tennis or volleyball game, the play section includes frames 220 and 240 containing scenes in which the player puts the serve, and frames 230 and 250 containing close-up scenes after one attack is completed. It is desirable to define.

The start frame extractor 120 extracts play start keyframes at which a play section starts among the extracted keyframes (step 1220). The play start keyframes have the same repeating shape or color. Therefore, the start frame extractor 120 extracts the start keyframes by using the repeatability of edge information or color information between the play start keyframes.

The field color detector 130 detects a field color using the extracted play start keyframes (step 1230). Since the field color is distributed at the lower end of the play start keyframes as the color of the stadium floor, the field color detection unit 130 preferably detects the field color using the lower area of the extracted play start keyframes. Do.

The end frame extractor 140 extracts play end key frames in which the play section ends among the key frames by using the detected field color (step 1240). The play end keyframe is preferably defined as a close-up scene. Since the close-up scene has a high proportion of the non-field area having a color different from the field color and the non-field area is concentrated, the end frame extractor 140 uses the ratio or the density of the non-field area. Preferably, the play end keyframes are extracted.

FIG. 3 is a block diagram illustrating an embodiment of the start frame extractor 120 of FIG. 1. The shown start frame extractor 120 includes a candidate extractor 300, a start frame model generator 310, and The model comparison unit 320 is made. The operation of the start frame extractor 120 shown in FIG. 3 will be described in conjunction with the flowchart shown in FIG. 13.

The candidate extractor 300 extracts keyframes having repeatability among the keyframes as play start candidate keyframes using edge information and color information of keyframes for each shot (step 1300). As shown in FIG. 5, the play start keyframes 500, 510, and 520 in the baseball game video have repeatability since the shape included in the keyframe, that is, the edge information and the color information are very similar. The candidate extracting unit 300 calculates the similarity between the edge information and the color information between the keyframes, and adds one keyframe group having the calculated similarity more than a predetermined threshold value as the play start candidate keyframes. It is preferable to ship.

The start frame model generation unit 310 generates a play start frame model using the extracted play start candidate key frames (step 1310). The start frame model generation unit 310 preferably calculates an average of edge information and color information of the play start candidate keyframes and determines the average value of the calculated edge information and color information as the play start frame model. .

The model comparison unit 320 calculates a similarity between the generated play start frame model and key frames for each shot, and extracts the key frame as a play start key frame when the calculated similarity is equal to or greater than a predetermined threshold similarity value (see FIG. Step 1320). When the start frame model generation unit 310 generates the play start frame model with edge information and color information, the model comparison unit 320 uses the edge information and color information of the start frame model as edge information and color of keyframes. In comparison with the information, it is preferable to extract the play start keyframe when the two calculated similarities are equal to or more than a predetermined threshold similarity value. In another embodiment, when any one of the calculated two similarities is equal to or greater than a preset threshold similarity value, the keyframe may be extracted as a play start keyframe.

4 is a block diagram illustrating an exemplary embodiment of the candidate extractor 300 of FIG. 3. The candidate extractor 300 illustrated in FIG. 3 includes a keyframe classifier 400, a start keyframe group selector 410, and the like. It comprises a start candidate frame extraction unit 420. The operation of the candidate extractor 300 illustrated in FIG. 4 will be described in conjunction with the flowchart illustrated in FIG. 14.

The keyframe classification unit 400 divides the keyframes into a plurality of groups such that keyframes having similar edge information are included in one group by using edge information of keyframes for each shot (step 1400). The start key frame group selector 410 selects a key frame group having the largest similarity among the key frames included in each group among the divided key frame groups as the play start key frame group (step 1410).

FIG. 6 illustrates an embodiment of a method in which the start keyframe group selector 410 selects a start keyframe group among a plurality of keyframe groups, and assumes that keyframes are divided into four groups. For each of the divided groups, the start key frame group selection unit 410 configures three key frames adjacent to each other in time as a window, so that edge information between key frames included in each of the windows or The similarity of the color information is calculated, and the number of the similar windows is calculated by determining a similar window as a window having the calculated similarities greater than a predetermined threshold similarity value.

Referring to FIG. 6, the start key frame group selection unit 410 may measure the similarity of edge information or color information between three key frames 620, 630, and 640 included in the first window 600 of the first group. If the calculated three similarities are larger than the threshold similarity value, the first window 600 is determined as a similar window. The start key frame group selection unit 410 determines whether or not the similar window is the same for the second window 610 of the first group, and then consists of three key frames adjacent to each other included in the first group. For all windows, the method determines whether the window is the similar window and calculates the number of similar windows included in the first group. In this manner, the start key frame group selection unit 410 calculates the number of similar windows included in each of the second group, the third group, and the fourth group, and then calculates the largest number of the similar windows. Select the keyframe group as the play start keyframe group.

The start candidate frame extracting unit 420 selects key frames included in the similar window among the key frames included in the play start key frame group (step 1420).

FIG. 7 is a block diagram illustrating an embodiment of the field color detector 130 of FIG. 1. The field color detector 130 includes a maximum ratio color detector 700, a candidate field color extractor 710, and an average. The calculation unit 720 is included. The operation of the field color detector 130 shown in FIG. 7 will be described in conjunction with the flowchart shown in FIG. 15.

As shown in FIGS. 8A, 8B, and 8C, since the field color is included at the bottom of the play start scene, the maximum ratio color detection unit 700 is located at the bottom 1/2 area for each of the extracted play start keyframes. The color of the maximum ratio among the included colors is detected (step 1500). 8A, 8B, and 8C, the maximum ratio color detection unit 700 preferably detects a maximum ratio of colors included in the lower half areas 800, 810, and 820 of the play start keyframe. .

The candidate field color extractor 710 stores color information of a preset global field color and a ratio occupying the lower half of the frame. The candidate field color extractor 710 compares the color information and the ratio of the maximum ratio colors detected for each of the play start keyframes with the color information and the ratio of the stored global field color. If the similarity of the values is equal to or greater than a predetermined threshold similarity value, the maximum ratio color is extracted as the candidate field color (step 1510).

The average calculator 720 calculates an average of color information of the extracted candidate field colors, and determines the calculated color information value as the field color (step 1520).

Since the field of the baseball stadium consists of two parts, the grass and the ground, the maximum ratio color detection unit 700 detects the two largest ratio colors with the highest ratio for each of the play start keyframes, and extracts the candidate field colors. The unit 710 extracts two candidate field color sets similar to the two basic field colors by comparing the detected two maximum ratio colors with two basic field colors, and the average calculating unit 720 performs the two It is desirable to determine the two field colors by calculating the average for each candidate field color set.

FIG. 9 is a block diagram illustrating an embodiment of the end frame extractor 140 of FIG. 1. The end frame extractor 140 illustrated in FIG. 1 includes a slice divider 900, a field pixel determiner 910, and a field. A slice determination unit 920 and a slice analyzer 930 are included. The operation of the end frame extractor 140 shown in FIG. 9 will be described in conjunction with the flowchart shown in FIG. 16.

The end frame extractor 140 performs one of the play end keyframes by performing the following operation on keyframes located between two play start keyframes adjacent in time among the extracted play start keyframes. Extract.

The slice dividing unit 900 divides each of the key frames into a plurality of slices (step 1600). FIG. 10 illustrates an embodiment in which the slice dividing unit 900 divides a keyframe into a plurality of slices.

The field pixel determination unit 910 compares the color information of the pixels included in the slice with the color information of the detected field color with respect to each of the divided slices, and the color similarity is equal to or greater than a preset threshold similarity value. The pixels are determined as field pixels (step 1610). If the pixel information of the pixel is not similar to the field color detected by the field color detector 130, the field pixel determination unit 910 has color information of the basic field color and color information of more than a predetermined threshold similarity value. It is desirable to determine the pixel as a field pixel.

The field slice determiner 920 determines a slice in which the number of field pixels included in each of the plurality of slices is equal to or greater than a preset threshold (step 1620).

The slice analysis unit 930 configures adjacent slices as one window, and then, for each of the windows, checks whether a slice that does not correspond to a field slice, that is, a ratio including a non-field slice is greater than or equal to a preset threshold ratio. If a window that satisfies the condition exists in the key frame, the key frame is extracted as the play end key frame (step 1630).

FIG. 11 illustrates an embodiment of a method for extracting the play end keyframe, wherein four adjacent slices included in the keyframe are configured as one window. The slice analyzer 930 checks whether all four slices 1120, 1130, 1140, and 1150 included in the first window 1100 are non-field slices, and extracts the keyframe as the play end keyframe in this case. When the slice analyzing unit 930 performs the above-described checking step on all windows such as the second window 1110, and there are windows in which all four slices among the windows included in the keyframe are all non-field slices. Preferably, the key frame is extracted as a play end key frame.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which are also implemented in the form of carrier waves (for example, transmission over the Internet). Include.

The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. And functional programs, codes and code segments for implementing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

Although preferred embodiments of the present invention have been described in detail with reference to baseball, tennis, and volleyball game videos, other sports videos such as soccer, hockey, handball, etc. can be applied, and the general knowledge in the art to which the present invention pertains. Those skilled in the art will appreciate that various modifications or changes of the present invention can be made without departing from the spirit and scope of the invention as defined in the appended claims. Accordingly, modifications to future embodiments of the present invention will not depart from the technology of the present invention.

As described above, according to the method and apparatus for detecting a play section of a sports video according to the present invention, the field color is detected using play start keyframes extracted from the video using repeatability, and the play is performed using the detected field color. By extracting the end keyframe, it is possible to provide a method and apparatus for detecting a play section which is adaptive to all sports videos and has high accuracy.

Claims (26)

In the method for detecting a play section of the sports video, Extracting play start keyframes among keyframes of the sports video; Detecting a field color from the extracted play start keyframes; And And extracting play end keyframes among keyframes of the sporting video using the detected field color. The method of claim 1, Dividing the sports video by a shot unit; And And extracting a keyframe for each of the divided shots. The method of claim 1, wherein extracting the play start keyframes comprises: And extracting the play start keyframes among the keyframes of the sport video using the repeatability of the edge information or the color information of the play start keyframes. The method of claim 1, wherein extracting the play start keyframes comprises: Extracting play start candidate key frames among key frames of the sports video; Generating a play start frame model using the extracted play start candidate keyframes; And And extracting play start keyframes from among keyframes of the sporting video using the generated play start frame model. The method of claim 4, wherein extracting the play start candidate keyframes comprises: Classifying key frames of the sports video into a plurality of groups using edge information of the key frames; Calculating similarity between keyframes belonging to the group for each of the classified plurality of groups, and selecting a group having the largest calculated similarity among the plurality of groups as a play start keyframe group; And And extracting, as a play start candidate keyframe, keyframes having a similarity level equal to or greater than a first predetermined value among keyframes belonging to the selected play start keyframe group. 6. The method of claim 5, wherein the similarity between the keyframes is And a similarity between edge information or color information of the keyframes. The method of claim 4, wherein generating the play start frame model And generating an average value of values of edge information or color information of the extracted play start candidate keyframes. The method of claim 1, wherein detecting the field color Detecting, for each of the extracted play start keyframes, a color having a maximum ratio among colors included in a predetermined lower region of the play start keyframe; And calculating the field color using the maximum ratio colors detected for each of the play start keyframes. 9. The method of claim 8, wherein calculating the field color Calculating, for each of the detected maximum ratio colors, a ratio of color information of the maximum ratio color and the maximum ratio color in the play start keyframe; Comparing the calculated color information and the ratio with the color information and the ratio of a predetermined base field color, extracting colors having a similarity with the predetermined base field color among the maximum ratio colors equal to or greater than a second predetermined value as candidate field colors. step; And And detecting the average of the extracted candidate field colors as a field color. The method of claim 1, wherein the field color is If the sports video is a baseball video, the sports video play section detection method characterized in that it is detected in two colors. The method of claim 1, wherein extracting the play end keyframe Dividing each of the keyframes of the sports video into a plurality of slices; By comparing the color of the pixels included in the divided plurality of slices with the detected field color, a pixel having a similarity between the color of the pixel and the detected field color equal to or greater than a third predetermined value is determined as a field region pixel. Doing; Determining a slice including a predetermined number of pixels determined as the field region pixels among the divided slices as a field region slice; And And extracting a play end keyframe from among keyframes of the sports video by using the determined number of field region slices or proximity between field region slices. The method of claim 11, Comparing the color of the pixels included in the divided plurality of slices with a predetermined base field color to determine a pixel having a similarity between the color of the pixel and the predetermined base field color as a fourth predetermined value or more as a field region pixel; Sports video play section detection method further comprises. The method of claim 11, wherein the dividing into a plurality of slices comprises The method of claim 1, wherein each of the key frames positioned between two play start key frames adjacent in time among the extracted play start key frames is divided into a plurality of slices. A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 1 to 13. In the device for extracting the play section of the sports video, A start frame extracting unit which extracts play start key frames among key frames of the sports video; A field color detector for detecting a field color from the extracted play start keyframes; And And an end frame extracting unit configured to extract play end key frames among the key frames of the sport video using the detected field color. The method of claim 15, A shot splitter for dividing the sports video in shot units; And And a key frame extracting unit extracting a key frame for each of the divided shots. The method of claim 15, wherein the start frame extraction unit And a play start keyframe is extracted from keyframes of the sport video using the repeatability of the edge information or the color information of the play start keyframes. The method of claim 15, wherein the start frame extraction unit A candidate extracting unit for extracting play start candidate key frames among key frames of the sports video; A start frame model generator configured to generate a play start frame model using the extracted play start candidate keyframes; And And a model comparison unit configured to compare the generated play start frame model with key frames of the sport video and extract play start key frames among the key frames of the sport video. 19. The method of claim 18, wherein the candidate extracting unit A key frame classification unit classifying key frames of the sports video into a plurality of groups using edge information of the key frames; A start key frame group selection unit for calculating similarity between keyframes belonging to the group with respect to each of the classified groups, and selecting a group having the largest calculated similarity among the plurality of groups as a play start keyframe group; ; And And a start candidate frame extracting unit for extracting key frames having a similarity or more as a play start candidate key frame among key frames belonging to the selected play start key frame group. 20. The method of claim 19, wherein the similarity between the keyframes is And a similarity between edge information or color information of the keyframes. The method of claim 18, wherein the start frame model generation unit And the play start frame model is generated using an average value of values of edge information or color information of the extracted play start candidate keyframes. The method of claim 15, wherein the field color detection unit A maximum ratio color detection unit for detecting each of the extracted play start keyframes, a color having a maximum ratio among colors included in a predetermined lower region of the play start keyframe; And a field color calculator configured to calculate the field color by using the maximum ratio colors detected for each of the play start keyframes. The method of claim 22, wherein the field color calculation unit For each of the detected maximum ratio colors, the ratio occupied by the color information and the play start keyframe is compared with the color information and the ratio of the predetermined basic field color, and the similarity with the predetermined basic field color among the maximum ratio colors is compared. A candidate field color extracting unit extracting colors having a second predetermined value or more as candidate field colors; And And an average calculator configured to detect an average of the extracted candidate field colors as a field color. The method of claim 15, wherein the end frame extraction unit A slice dividing unit dividing each of the key frames of the sports video into a plurality of slices; By comparing the color of the pixels included in the divided plurality of slices with the detected field color, a pixel having a similarity between the color of the pixel and the detected field color equal to or greater than a third predetermined value is determined as a field region pixel. A first field pixel determination unit; A field slice determination unit which determines a slice including a predetermined number or more of pixels determined as the field region pixels among the divided slices as a field region slice; And And a slice analyzer configured to extract a play end keyframe from among keyframes of the sports video by using the determined number of field region slices or proximity between field region slices. The method of claim 24, Comparing the color of the pixels included in the divided plurality of slices with a predetermined base field color to determine a pixel having a similarity between the color of the pixel and the predetermined base field color as a field area pixel equal to or greater than a fourth predetermined value; Sports video play section detection apparatus further comprises a two-field pixel determination unit. The method of claim 24, wherein the slice splitter And a plurality of slices of each of the key frames positioned between two play start key frames adjacent in time among the extracted play start key frames.

Images