JP2003061038A - Video contents edit aid device and video contents video aid method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To freely edit video contents in a short time by the intention of an end user. SOLUTION: A video contents analyzer 82 analyzes contents of an original video image, and a contents-providing device 83 provides a particular scene extracted from the contents. When an edit instruction is sent, while confirming the provided particular scene, a video configuration device 89 provides an edited object video image resulting from applying edit processing to the original image along with the edit instruction this time. Upon the receipt of revision of edit processing, a new edited object video image subjected to revised edit processing is provided to an end user. When receiving an edit processing end instruction, the edited object video image provided, immediately prior to this is converted into a desired format and the resulting image is distributed as the edited video image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video content editing support device and a video content editing support method which allow an end user to edit video content.

[0002]

With the spread of high-performance digital video cameras in recent years, general users can easily take and own digital video contents. Along with this, the demand for video editing is increasing, and general-purpose personal computers equipped with a video content editing environment (authoring tool)
It is provided as a function on hardware and software. However, such an editing function at present has no choice but to look at all the video contents in chronological order, and because it takes more editing time than recording time, it is not always easy to use for general users.

A moving image taken by a digital video camera (eg, travel, athletic meet, graduation album, personal history,
In order to reduce the data volume, performances such as concerts are converted to MPEG (Moving Picture Experts Group) compressed video content on a personal computer, but the content of video content shot by general users is often blurred and redundant. Since it is often the case, it is desirable to cut unnecessary parts and edit them into short scenes. However, when editing, it is necessary to follow the evaluation of the video content itself with the eyes, and since the editing process requires enormous effort and time, in reality, editing at the general user level has hardly been executed.

Therefore, in view of the above problems, it is an object of the present invention to provide a video content editing support device and a video content editing support method which allow the end user to freely edit video content in a short time.

[0005]

According to a first aspect of the present invention, there is provided a video content editing support device, a receiving means for receiving an original video to be edited, a video content analyzing means for analyzing the content of the original video, Content presentation means for extracting and presenting a specific scene from the analysis content of the video content analysis means, editing command receiving means for receiving an editing command based on the specific scene, and editing processing of the original video along the editing command. Edited candidate video presenting means for presenting the edited edited candidate video, and edit processing for accepting a change in the edit processing, and causing the edited edited candidate video presenting means to present a new edited candidate video by the changed edit processing. Upon receiving the change accepting means and the completion command of the editing process, the edited candidate video presented immediately before that is converted into a desired format and distributed as the edited video. Configured to include a video distribution means.

Further, the video content editing support method according to claim 4 of the present invention receives an original video to be edited, presents a specific scene extracted by analysis contents of the original video,
When an edit command based on this specific scene is accepted, an edited candidate video in which the original video is edited according to the edit command is presented, and when a change in the edit process is accepted, a new one is created by the modified edit process. When the edited candidate video is presented and the editing processing completion command is received, the edited candidate video presented immediately before that is converted into a desired format,
The feature is that it is delivered as an edited video.

In this case, the content of the acquired original image is analyzed, and the specific scene extracted from it is presented to the end user. While confirming the specific scene presented, the end user sends out as an editing command what kind of editing should be performed on the original image, this time the original image is edited according to this editing command The candidate video is presented to the end user. When the editing process change from the end user is accepted, the new edited candidate image by this modified editing process is presented to the end user, and when the editing process completion command is accepted, the edited candidate image presented immediately before that is received. Is converted into a desired format and distributed as an edited video.

As described above, when the original video is edited on the end user side, not the original video itself but a specific scene obtained by analyzing the content of the original video or an edited candidate video subjected to the editing process. All you have to do is send the appropriate commands for editing while watching. Therefore, it is possible to save time and effort to evaluate the content of the video content (original video) itself with eyes and edit the video content freely in a short time at the will of the end user.

In the above-mentioned structure, in the video content editing support apparatus according to claim 2, the video content analyzing means divides the original video into a plurality of scenes, and the visibility of the video for each scene is evaluated at a plurality of evaluation levels. In addition to the above, the content presenting means presents a scene of an evaluation level that matches a requested condition as the specific scene.

Further, in the above method, the video content editing support method according to claim 5 divides the original video into a plurality of scenes when presenting a specific scene extracted by the analysis content of the original image, and for each scene. Is determined from a plurality of evaluation levels, and a scene of an evaluation level that matches the requested condition is presented as the specific scene.

In this case, the received original video image is divided into a plurality of scenes, and the visibility of the video image for each scene is determined from a plurality of evaluation levels. Then, the scene of the evaluation level that matches the condition requested by the end user is presented to the end user as a specific scene. Therefore, the end user can confirm the scene of the required evaluation level before editing without evaluating each scene in the original image.

In the above structure, in the video content editing support apparatus according to claim 3, the content presenting means presents a representative still image in the specific scene.

Further, in the above method, the video content editing support method according to claim 6 is characterized in that, at the time of presenting the specific scene, a representative still image in the specific scene is presented.

In this case, the end user only needs to confirm the representative still image of the required specific scene, and it is possible to reduce the amount of information presented as the specific scene.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of a video content editing support device according to the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the overall configuration of the apparatus. This device is composed of a video content editing support system 71 equivalent to the device main body installed in the center by a video content editing support company, and an information network 72 as a communication means for transmitting information to end users in the Internet environment. Composed. Reference numeral 73 is a terminal such as a personal computer or a mobile phone, which is connected to an end of the information network 72. As is well known, the terminal 73 includes an operation unit having a plurality of operation keys and a display unit such as a liquid crystal display or a CRT. Then, by properly connecting the terminal 73 to the information network 72, various information can be acquired from the video content editing support system 71 through the home page on the Internet managed by the video content editing support system 71, and vice versa. Can be sent out.

Next, the support system main body 1 will be described in detail. Reference numeral 81 is a receiving device corresponding to a receiving means for receiving a video content which is an original video serving as a material in response to a request from a user. The receiving device 81 may be, for example, a video input interface for acquiring material video data that is an original video via the information network 72, or may be an original from a storage medium (eg, digital video tape or CD-ROM) entrusted to the user. It may be a reading device that reads an image. In any case, the receiving device 81 here is configured as a device that takes in the original video created in various formats and converts it into an output format that can be used by the video content editing support system 71.

Reference numeral 82 is a video content analysis device corresponding to a video content analysis means for analyzing the content of the original video taken in from the receiving device 81, and 83 is a specific content obtained based on the content analysis result of the original video. The content presentation device corresponds to content presentation means for presenting a scene to the terminal 73 through the information network 72. The video content analysis device 82 individually separates the video contents of the original video compressed in the MPEG2 system, for example, by changing scenes including scene changes and camera work, and stores them in the video storage device 84. The ID number corresponding to the video ID together with the evaluation level, which indicates the evaluation of the ease of viewing the video.
The storage device 85 has a function of storing. Although the video storage device 84 and the video ID storage device 85 are configured as the common storage device 86 here, they may be configured as separate storage devices, for example. And these video storage devices 84
And the video content stored in the video ID storage device 85,
It is centrally managed by a content management device 87 provided in the video content editing support system 71. The details of how the video content analysis device 82 analyzes the contents of the original video will be described later in detail.
From the specific scenes obtained by analysis by 82, a representative still image is displayed by the content presentation device 83 on the information network 72.
These still images can be viewed side by side on the terminal 73 side.

Reference numeral 88 denotes an edit command (which is sent to the video content edit support system 71 via the information network 72 by an input operation of the terminal 73 after the user confirms the video content from the content presentation means 83 through the terminal 73. It is an edit command reception device that receives a command). The editing command here means deleting unnecessary scenes from the original video and adding data (other video, sound,
It is a command to insert subtitles, effects, etc.)
When the edit command receiving device 88 receives the edit command, the content management device 87 retrieves the necessary scene ID number and the original image corresponding thereto from the storage device 86, and the retrieved ID number and original image Based on the editing command from the editing command receiving means 88, the video composing device 89
Generates an edited candidate video, and this edited candidate video is transmitted to the terminal 73 owned by the user via the information network 72.
It is supposed to be sent to.

The edit command accepting means 88 has not only the function as an edit command accepting means for accepting an edit command for the representative still image presented by the content presenting means 83, but also the edited candidate image edited by the image composition device 89. With respect to the above, a function as an edit process change receiving unit for receiving a change command such as addition or correction of the edit process is also provided. Each time the edit command receiving means 88 receives such a change command such as addition or correction, the video composing device 89 regenerates a new edited candidate video based on the content thereof, and regenerates the new edited video.
Send to terminal 73 via 72. Then, the final edited candidate image is confirmed by the output means on the terminal 73 side, and the fact that there is no additional or correction instruction is confirmed as an instruction to confirm the completion of the editing process, and the video is transmitted from the terminal 73 side via the information network 72. When sent to the content editing support system 71, this confirmation command is sent from the edit command receiving means 88 to the content management device 87. In response to this, the video composing device 89 also serving as the edited video distribution means converts the edited candidate video, which has been confirmed for the editing process, into the format desired by the user, and outputs the edited video from the information network 72 to the terminal.
Delivered to 73.

Next, the internal structure of the video content analysis means 82 and its operation procedure will be described. FIG. 2 is a schematic block diagram of the video content analysis means 82, in which 1 is MPE.
It is a camera work extraction means for extracting a camera work scene such as pan or zoom based on the motion information (motion vector) included in the G2 compressed video content (original video). Further, 2 is a video section continuously shot during power-on / off of one camera based on color information (color histogram change in time series) included in the compressed video content of the MPEG2 system, That is, the key frame extraction means is provided with section division means for dividing the inside of a shot into a plurality of sections, and extracts meaningful frames as key frames from the section division means. This section is divided more as the camera movement in the video content is larger, but the camera work extraction means 1 divides the scene by the key frame extraction means 2 in a scene in which the motion vector information has no clear feature. If the number of sections per unit time is equal to or greater than a predetermined value, the determination correction means 3 is provided to correct the determination that the scene is a camera work scene. Then, the video content with this determination result added as a parameter of camerawork is
The camera work extraction means 1 and then the video content analysis device 82 output the data.

Reference numeral 4 is based on the degree of camerawork detection and blurring extracted by the camerawork extraction means 1 and the intensity of image change that can be obtained from the time length of one section detected by the keyframe extraction means 2. The video content automatic structuring means determines the visibility of each scene in the video content from a plurality of evaluation levels and automatically structures the video content according to each evaluation level. The details of the video content automatic structuring means 4 will be described later, but according to the evaluation level for each scene obtained here,
It is possible to extract a representative still image from scenes at the same evaluation level.

FIG. 3 is a flow chart showing the detailed procedure relating to the above configuration and the configuration for realizing it.
In the figure, the prediction residual coding coefficient (DCT coefficient) D from the bit stream for each frame included in the compressed video content of the MPEG2 system in step S1.
The C component is decoded and extracted by the decoder 11 (step S
2). At this time, the motion vector information within the frame is extracted (step S3), and based on this motion vector information, in the next step S4 the blurring of each section and camerawork are detected by the section feature quantity extraction unit 12.

FIG. 4 is a flow chart showing a detailed procedure of the section feature quantity extraction unit 12 for performing processing in frame units and a configuration for realizing it. In the figure,
In step S21, a motion vector is selected according to the camerawork. Specifically, the scene in which the camera is moving is detected in each frame, and P is detected according to the speed of the camera.
In a frame with a short reference frame interval, which controls the use of a motion vector of a picture or a B picture, a motion vector of a B picture is suitable when the camera moves quickly, and in a frame with a long reference frame interval, a P picture The motion vector is suitable when the camera moves slowly. Therefore, when the camera moves fast, the B-picture motion vector is selected, and when the camera moves slow, P
Select a motion vector for the picture. Next, the motion vector usage rate calculator 21 in the same step S21 calculates the motion vector usage rate RVi from the selected motion vector. In a scene in which the camera is moving, the scalar value of the motion vector in the frame often becomes nonzero.
The motion vector usage rate calculator 21 calculates the proportion of macroblocks in which the scalar value of the motion vector is not 0. If the motion vector usage rate RVi is smaller than a predetermined value, the process proceeds to step S22 and it is determined that the scene is fixed. If the motion vector usage rate RVi is greater than or equal to the predetermined value, step S23
Then, it is judged as a moving camera work scene including camera shake, and the two are distinguished. Since the scene fixed in step S22 does not include the blur, the information is the section feature quantity extraction unit.
It is output as it is from 12.

The moving scene in step S23 includes so-called blurring, such as a serious blurring scene that cannot be seen when the power is forgotten to be turned off or when the photographer changes, or a camera shake scene. Therefore, in the next step S24, the intra-frame motion vector smoothness calculator 22 extracts the former severely blurred scene from the moving scene in step S23. In-frame motion vector smoothness calculator 22
Is to calculate the continuous value of the motion vector of each macroblock in the frame, divide the frame in the vertical direction, and calculate the difference value in the vector direction. In a severely blurred scene, the directions of the motion vectors in the frame have variations, so if the motion vector smoothness in the frame exceeds the threshold value, it is determined that the frame is a severely blurred scene (step S25), and the frame If the smoothness of the motion vector is within the threshold, it is determined that the frame is one of pan, zoom, and camera shake scene (step S26). In this way, severely blurred scenes that do not require motion analysis are separated from other scenes.

In the next step S27, the in-frame motion vector histogram deviation degree calculator 23 separates the pan scene and the zoom scene for the scene in which the camera is moving in step S26. However, in this pan scene, in addition to the case where the camera changes in the left and right direction, tilt that changes in the up and down direction, a track that changes the camera position in the left and right direction, a boom that changes the camera position in the up and down direction, All combinations are included. Also, the zoom scene is
It also includes a dolly that moves the camera back and forth. The intra-frame motion vector histogram bias degree calculator 23 clusters the directions of the motion vectors in the frame into, for example, eight ways of classes 1 to 8 and then obtains a histogram (frequency distribution) of each class. The ratio of the maximum frequency of this histogram to the whole is calculated as the intra-frame motion vector histogram bias degree. If this bias degree exceeds a threshold value, it is determined to be a pan scene including camera shake (step S28), and the bias degree is If it is less than or equal to the threshold value, it is determined to be a zoom scene including a camera shake scene (step S29).

In the pan scene and the zoom scene, the motion vector in the frame shows the same property during the section, whereas in the camera shake scene, the motion vector in the frame changes in a short cycle. Utilizing this, in step S31, the intra-frame average motion vector difference value calculator 24 separates the pan scene from the camera shake scene, and in step S32, the intra-frame motion vector histogram difference value calculator 25 calculates the zoom scene and camera shake. Perform scene separation.

In the pan scene, the average motion vector in the frame is constant, whereas in the camera shake scene, the average motion vector in the frame changes in a short cycle. The in-frame average motion vector difference value calculator 24 calculates the difference value of the average motion vector in the frame. If this difference value is less than or equal to the threshold value, it is determined to be a pan scene (step S33), and the difference value is the threshold value. If it exceeds, it is determined to be a camera shake scene (step S34).

Intra-frame average motion vector difference value calculator
25 clusters the directions of the motion vectors in the frame into, for example, eight ways of classes 1 to 8 as shown in FIG.
After obtaining the histogram (frequency distribution) of each class, the change in the histogram for each frame is calculated as the intra-frame motion vector histogram difference value. If this difference value is less than or equal to the threshold value, it is determined to be a zoom scene (step S35). If the difference value exceeds the threshold value, it is determined to be a camera shake scene (step S36). Then, the fixed scene, pan scene, zoom scene, camera shake scene, and severe blur scene are separated for each frame unit by these series of procedures.

In another step S41, the weighting function for each frame for the P-picture filter and the B-picture filter is used as a weighting function so that camerawork and blurring can be detected in GOP units. It is calculated by the calculator 27. In calculating the weighting function, the number of motion vectors in the frame is used for the result of step S21 of detecting the camera motion, and the smoothness of the motion vector is used for the other results. .
More specifically, the weighting function calculator 27 calculates the average of the motion vector usage rate RVi for detecting the motion of the motion vector and the motion vector amount QVI within the GOP, and divides each value by SMV (smoothness). Weighting is done by that. Further, for the motion vector histogram bias degree and the motion vector histogram variance value, average values using weight values are calculated. The weight value here means a ratio of frame values in which a number of motion vectors whose scalar value is not 0 is equal to or larger than a threshold value in a portion in the GOP where the motion of the camera in the frame is known. Further, with respect to the motion vector histogram change, the average motion vector change, and the motion vector histogram variance value change, the maximum value is selected by multiplying the weight value for each frame. The weight value of each frame (step S42) thus obtained is used for the feature detection for each GOP in step S7 described later.

Returning to FIG. 3 again, when the section feature amount extraction unit 12 detects blurring of each section or camera work in step S4, the procedure of step S6 is performed through the procedure of buffering each GOP in step S5. Here, we will describe an example when the number of frames that make up a GOP is 15, but it is judged whether the remainder when the frame number is divided by 15 has reached 14. If the remainder has not reached 14 again, Step S2
Return to the procedure, but if the remainder reaches 14, step S7
At, the characteristic for each GOP is detected by each parameter calculator 13 for GOP detection. As a result, one frame of I picture composed only of intra macroblocks and interframe motion prediction coding method were adopted.
It is possible to configure 15 frames consisting of B picture or P picture of a frame as one unit, and configure these as one GOP.

Each parameter calculator in step S7
Reference numeral 27 is for detecting camerawork and blur for each GOP by using the weight value of each frame derived in step S41. Specifically, the ratio of the number of motion vectors in each GOP and the smoothness of the motion vectors is used as a weight value, and the value for each frame is multiplied by 1 GOP.
Then, the sum is calculated, and the feature amount of camerawork or blurring in the GOP is output as a parameter.

On the other hand, the DC coefficient (step S11) obtained during the decoding in step S2 is detected by the shot detection calculator 14 in the next step 12 by detecting the cut point. , Step 13
At, the intra-shot section detector 15 detects a content change within one shot. Note that the cut point is a transition point of a scene when the power of the camera is turned on / off, and the shot is a video section continuously photographed between the cut points. The shot detection by the shot detection calculator 14 may use any conventionally known method. The in-shot section detector 15 tracks structural changes in the shot by using changes in the color histogram, and this corresponds to the key frame extraction means 2 shown in FIG. Then, the shot number and the cut point (step S14) obtained by the shot detection calculator 14,
The section points (step S15) corresponding to the key frames in the shot obtained by the in-shot section detector 15 are matched in time series with the feature amounts of camerawork and blurring from step S7. This work will be described later.

Now, a method of extracting a key frame by the key frame extracting means 2 will be described. Since the feature of the video content shot by a general user is that the shot is long, it is important to know the temporal change in the shot as in step S13. Therefore,
It is preferable to divide a shot into a plurality of sections, perform clustering on the section, and then preferentially extract a frame closest to the center of gravity of the cluster as a key frame from a cluster having a large number of elements. As a result, it is possible to reduce the processing cost required for random access to moving images and indexing of moving images, even for video contents shot by general users.

In order to correctly extract the key frame, it is necessary to extract from the shot a frame which is considered to reflect the intention of the photographer when photographing the object which is considered significant. In general, what is considered important by the photographer has a longer shooting time than ones which are not important, and tries to shoot by placing the focus of the important object in the center of the frame. In addition, it seems that the photographer considers the shooting to be important, even when shooting with zoom. Therefore, in this embodiment, a key frame is extracted when the shooting time is particularly long and the focus is in the center of the frame.

When determining the similarity of images within a shot, it can be said that the video content is data in which frames are sequentially reproduced along the time axis, so that between several consecutive frames within the same shot. The image composition and the shape of the object being photographed are considered to be substantially the same. Therefore, as the feature quantity used to obtain the similarity,
It is noted that only the color information is taken into consideration here, not the texture, the shape, or the position of the object. Although there are several color models that represent color information, the luminance value Y and the two chrominance values C are taken into consideration in consideration of using the compressed video content of the MPEG2 system as described above.
It is preferable to use a YCbCr color model that represents bCr. In particular, here, the luminance histogram H, the Euclidean distance D in the YCbCr space, the luminance of the frame in the YCbCr space, and the average color difference (Yavr, Cbavr, Cravr) are used as the feature amount indicating the similarity of the images.

The luminance histogram H loses information about the spatial distribution of colors in the image (image composition information and the shape of objects in the image).
The fact that images are similar does not mean that the images are similar. However, in a moving image, a common portion is photographed in several consecutive frames, so when the luminance histograms H are similar between the frames, the composition information of the image and the shape of the object in the image are to some extent. It can be assumed that they are similar. Based on that assumption, the reliability of the similarity using the luminance histogram H in the video content is high. It is also possible to use an inter-frame difference as a feature amount for comparing consecutive frames, but in this case, the human visual similarity cannot be well reflected in the movement over the entire screen generated by the camera operation. On the other hand, the luminance histogram H has an advantage that the human visual similarity can be well reflected. The similarity m between the frame i and the frame j is expressed by the following formula 1 when the number of pixels in the portion where the luminance histograms Hi (s) and Hj (s) (s is the luminance value) overlap each other. .

[0038]

[Equation 1]

Euclidean distance D in YCbCr space
Is used for simple clustering described later.
Here, the value of the YCbCr space of the frame i is set to (Y, C
b, Cr) and the value of the YCbCr space of the frame j is (Y ', Cb', Cr '), the Euclidean distance D in the YCbCr space of the frame i and the frame j can be expressed by the following equation 2.

[0040]

[Equation 2]

The average luminance value and average color difference (Yavr, Cbavr, Cravr) at the frame center in the YCbCr space are
It is calculated using the feature amount at the center of the frame based on the assumption that the focus is placed on the center of the important object to shoot. Further, by paying attention to the center of the frame, it is possible to distinguish them by the difference in focus even if the frames are similar to each other as a whole. The center of the frame here refers to a portion corresponding to the center 3 × 3 by dividing the frame into 5 × 5. As luminance and color difference information, the luminance average of the Y component in the YCbCr space and the Cb component are used. And the respective color difference averages of the Cr component are used. The luminance of the Y component of the pixel at the coordinates (x, y) in the frame image is Y (x, y), and the color difference of the Cb component is Cb.
If (x, y) and the color difference of the Cr component is Cr (x, y), the luminance average Yavr of the Y component in the YCbCr space,
Cb component color difference average Cbavr and Cr component color difference average C
ravr is expressed by the following equation 3.

[0042]

[Equation 3]

However, w is the width of the image, and h is the number of pixels of the height of the image.

By the way, when the shot is divided into a plurality of sections and the time axis is not taken into consideration when the clustering is performed on the sections, the feature values at the center of the frame are similar, but are actually completely different. There is a problem that things belong to the same cluster. For example, assume that after a person's face is photographed, pan is performed to photograph another person's face. At this time, the central brightness and color difference of the frames in which the two persons are moving are determined to be similar because the human skin color is similar, and the frames in which different persons are photographed have the same cluster. More likely to belong to. These phenomena
This is particularly noticeable in images when the camera is operated or images in which the number of shot frames is large.

In order to prevent such a phenomenon, it is necessary to divide a section into a section in which a person is photographed first and a section in which a person is photographed next, and to cluster the section. preferable. That is, if the shot is divided into sections in advance and the key frame extraction means 2 is configured to perform clustering in the divided sections, YCbCr
Average luminance value and average color difference (Yav
Even if r, Cbavr, Cravr) are similar, it is possible to reduce the problem that completely different frames belong to the same cluster. Also, when browsing a moving image, the flow of time becomes important when grasping the content of the moving image.

The procedure for dividing the shot section by the key frame extraction means 2 will be described with reference to FIG. 6. First, the first frame in the frame group 51 in the shot is set as the start frame 51A of the section 1. This start frame
The degree of similarity m between 51A and the subsequent frames is calculated in order along the time axis based on the equation (5). When the similarity m becomes smaller than a certain threshold Th, the frame at that time becomes the next start frame 51B. Hereinafter, the similarity m between the frontmost start frame 51B and the subsequent frames is similarly calculated sequentially along the time axis, and the similarity m is the threshold Th.
Frame when it is smaller than the next start frame
Repeat procedure 51C. Thus each start frame
51A, 51B, 51C as the head, three sections 52A, 52
B and 52C are divided. That is, the similarity m is used in advance to divide the section in the shot.

Since originally, the change in the entire frame is focused on, it is ideal that only the histogram of the entire frame is used as the feature amount indicating the similarity when setting the start frame of the next section. However, in video contents in which the background occupies a large proportion, the entire luminance histogram H is greatly affected by the background. It becomes impossible to divide what should be divided into sections. On the other hand, in the luminance histogram H at the center of the frame, the proportion of objects in the image in the background is larger than in the case of the entire luminance histogram. Therefore, in calculating the similarity m, it is preferable to consider the luminance histogram H at the center of the frame together in order to reduce the influence of the background. In addition, when the luminance value of the frame changes for a moment due to the influence of a flash or the like, it is possible that the originally same section is erroneously detected as a section break. In order to avoid such erroneous detection, when the similarity m to the previous start frame becomes smaller than the threshold Th continuously for a plurality of frames (for example, 2 frames), the section is divided, and the start frame at that time is divided. May be the frame that was previously compared in the plurality of frames.

Next, a procedure for extracting a key frame from each of the divided sections 52A to 52C will be described with reference to the schematic diagram of FIG. First, in each of the sections 52A to 52C, simple clustering is performed using the average luminance value and the average color difference in the YCbCr space of the frame center calculated by the above-mentioned equation 3. As a result of the clustering, for example, in FIG. 7, three clusters in the section 52A, that is, a cluster 54A having 10 elements is used.
, A cluster 54B having 5 elements, and a cluster 54 having 2 elements
C is formed, and there are two clusters in the section 52B, that is, a cluster 54D having 7 elements and a cluster 54 having 10 elements.
E is formed, and in the section 52C, two clusters, that is, a cluster 54F having seven elements and a cluster 54 having seven elements are formed.
G is formed.

Next, from the plurality of clusters 54A to 54G formed in each of the sections 52A to 52C, the respective clusters 54A to 54G are
The frame closest to the center of gravity of G is the key frame candidate 55A.
~ 55G are extracted one by one. Then, a desired key frame is extracted according to the detail desired by the user. At this time, key frame candidates extracted from a cluster having a large number of elements (for example, in FIG. 7, key frame candidate 55
As A and 55E have the largest number of cluster elements), the degree of importance, that is, the presentation priority, is set higher. The reason is based on the assumption that the photographer thinks that the longer the shooting time is, the more important the shooting is. That is, a cluster having a large number of elements (frame number) becomes more significant to the photographer than a cluster having a small number of elements.

When the simple clustering is performed and the time axis in the section is taken into consideration, the possibility of extracting visually similar extra frames increases, whereas the time axis in the section is If not taken into consideration, the possibility of extracting extra frames can be reduced. This is because the photographer may take a picture once and take the picture again. Here, a procedure for simple clustering the n feature vectors shown in Equation 4 will be described.

[0051]

[Equation 4]

First, an arbitrary vector X _i is taken, and this is set as the center Y ₁ (Y ₁ = X _i ) of the first cluster C ₁ . Next, take another vector _{X j,} the Euclidean distance D _{1, j} of the first cluster _{C 1} of the center _{Y 1} and vector _{X j}
Ask for. At this time, if D _{1, j} > T, the vector X _{j is set} to the center Y ₂ (Y ₂ = X _i ) of the second cluster C _2.
And if D _{1, j} ≤T, then the vector X _j is the first
The cluster center Y ₁ is updated as (X _j εC ₁ ) included in the cluster C ₁ .

[0053] Thereafter, take the vector _{X k,} center _{Y 2} center _{Y 1} and the second cluster _{C 2} of the first cluster _{C 1}
And Euclidean distances D _{1, k} and D _{2, k} from At this time, if D _{1, k} > T and D _{2, k} > T, the vector X _{k is set} to the center Y of the third cluster C ₃ .
₃ (Y ₃ = X _i ) and D _{1, k} ≦ T or D _{2, k} ≦ T
Then, the vector X _k belongs to the cluster having the shorter distance from the center.

Thus, the above procedure is repeated for all the vectors, and the clustering is completed. The center of the cluster is updated every time a new element is added. Here, the center of the cluster before the new element increases is (Ycen, Cbcen, Crcen), the number of elements is num, and the elements added to the new cluster are (Ynew, Cbnew,
Crnew), the center of the new cluster (Y'cen,
Cb'cen, Cr'cen) can be expressed by the following equation 5.

[0055]

[Equation 5]

As a result of the above clustering, if the number of clusters is too large or too small, the value of T may be changed and clustering may be performed again.

Next, the experimental results of the key frame extraction processing will be described with reference to FIGS. 8 and 9 and Table 1. The video contents used in the experiment were videos of several people, and the conditions were 720 × 240 pixels, 30 frames / sec.
The G-2 format was saved and converted to ppm format, and the shooting time is 31 seconds. When the similarity Th of the brightness histogram H of the entire frame is calculated, the threshold Th is
When Thall = 121000 and focusing on the luminance histogram H at the center of the frame, Thcenter = 32000.
Moreover, T = 30 was used as the threshold value of the Euclidean distance D. The section number of the section into which the video content is divided, the key frame candidate number representative of the cluster refined from each section, and the number of elements of the cluster are limited to those in which the section range is 5 frames or more, and shown in Table 1. Show.

[0058]

[Table 1]

Further, FIG. 8 shows the luminance average (solid line) and the color difference average (broken line) at the center of each frame, and FIG. 9 shows the similarity m of each frame to the start frame. The whole (solid line) and the center of the frame (broken line) are shown respectively. As a result, when the key frames are extracted from the key frame candidates so as to be over-detected, the frame required by the photographer can be presented. Further, when the key frame is extracted, if the information such as camera work and camera shake is also used, it is possible to more effectively present the key frame. Here, although only the luminance information is used when dividing the shot, it is possible to divide the shot into more appropriate sections in consideration of color difference components and the like.

Subsequently, the change of the color information obtained by the key frame extraction means 2 is used to calculate the motion vector reference frame interval while watching the change of the video content, and a new motion vector is generated to generate the camera work. The procedure used for detection will be described below. As described above, when the section is detected using the color information (change of the luminance histogram H on the time series), the change point of the moving image can be robustly obtained.
That is, it is possible to detect a scene in which the color composition of the entire frame is the same, but on the other hand, in a camera work scene such as pan or zoom, excessive detection of sections occurs.

On the other hand, with the motion information based on the motion vector, it is possible to detect a time-series change of the entire frame, and it is possible to detect a severely blurred scene whose contents are difficult to understand as an image as an image processing unnecessary frame by the above-mentioned procedure. When the motion vector information has no characteristic, camerawork may not be detected or may be detected as wrong camerawork. In the present embodiment, the point that camerawork is detected by using the color information and the motion information in a hybrid by the determination correction means 3 is noted as a novel feature. A specific method for detecting this camerawork will be described with reference to FIG. FIG. 10 schematically shows an example of the detection result, and “CW” in the upper row is a time-sequential arrangement of the scenes extracted by the camerawork extracting means 1.
"P" is a pan scene, "F" is a fixed scene, "S" is a handshake scene, and "N" is a scene in which the motion vector information cannot be discriminated because it does not have clear features. There is. Also, although not shown here,
The zoom work is also detected as one of the camera works by the camera work extraction means 1 like the pan scene. That is, two types of camera work, a pan scene and a zoom scene, are detected from the motion information. On the other hand, the lower "KF"
Is a time-series arrangement of the detection of the sections by the luminance histogram H extracted by the key frame extraction means 2.

The determination correction means 3 corrects the camerawork detection result obtained from the camerawork extraction means 1 and the section detection result obtained from the keyframe extraction means 2 in the following procedure. In a fixed scene (camera fixed) “F” in which motion vector information does not exist, since there is no change in color information, one section extracted by the key frame extraction unit 2 is detected. Therefore, in such a case, it is determined as it is as a fixed scene. The rapid frame change has a characteristic that the motion information is clear, so that the precision of camerawork detection by the camerawork extraction means 1 is high. Therefore, in the panning scene “P” shown in FIG. 10, the judgment correction means 3 puts together the sections that the key frame extraction means 2 has excessively detected, and the camera work extraction means 1 minimizes the frame interval for motion compensation to make the camera Work and camera shake are separated (minimum interframe motion compensation). As a result, it is possible to prevent excessive detection of a section based on the luminance histogram H for a fast frame change such as a pan scene or a zoom scene, and it is possible to more accurately determine camerawork and camera shake.

On the other hand, when the frame changes slowly, there is little change from the neighboring frames before and after that, and therefore the motion vector information has few characteristics as camerawork, and the matching rate of camerawork detection is low. Therefore, the determination correction unit 3 includes a camera including a camera shake when the key frame extraction unit 2 divides a section of a predetermined value or more between scenes "N" in which the camera work cannot be determined as shown in FIG. The camera work extracting means 1 determines that the work scene is a correction, and in response to this, the camera work extraction means 1 widens the interval of the motion compensation frames in this scene (variable inter-frame motion compensation) so that the camera work scene is a pan or zoom camera work scene. Or a camera shake scene. As a result, even if the scene cannot be identified by the motion vector information alone, it is possible to determine whether or not the scene is a camerawork scene by considering the section detection based on the color information.

The frame change amount per section by the key frame extraction means 2 is equal regardless of the change speed. Therefore, when an object (object) is near or when the camera work scene is detected quickly, the frame change per unit time (image change) from the length of one section detected by the key frame extraction means 2. It is also possible to detect the intensity of). The intensity of this video change is a criterion for determining the ease of viewing the video for each scene in the video content automatic structuring means 4, together with the degree of detection and blur of the camerawork scene extracted by the camerawork extraction means 1. Becomes

Table 2 shows an example in which the evaluation criteria for determining the visibility of the image are set to four levels, level 0 to level 3.

[0066]

[Table 2]

In Table 2, the camera-shake scene and the severe camera-shake scene are distinguished by the camera work extracting means 1 as shown in FIG. In other words, “camera shake” in the evaluation criteria of level 0 and level 1 is a camera shake scene that is difficult to see although the contents of the video are understandable, and “camera shake” in the evaluation criteria of level 3 is the understanding of the contents of the video. It is a terrible blur scene where it is difficult to blur. Furthermore, the “quick video change exceeding a certain threshold” at level 2 can be detected by the time length of the section divided by the key frame extraction means 2.

FIG. 11 is a schematic diagram showing an example in which the video contents are automatically structured based on the easiness of viewing the video based on the evaluation criteria of level 0 to level 3. In the figure, “scene change” is a so-called cut point, which is a scene change accompanying the power on / off of the camera. Here, three shots A, with the cut point in between,
There are B and C. As shown in the top row of FIG. 11, the video content is divided by scene change and scene detection, and the level 0 to level 3 are divided for each divided scene.
Is evaluated. In this way, the video content can be automatically structured for each shot A, B, C at the evaluation level based on the visibility of the video.

The video content automatic structuring means 4 extracts only scenes of the same evaluation level or up to a certain evaluation level according to a request from the outside, and extracts representative still images (for example, the first and last scenes, the middle, Or only the first still image) is displayed side by side. Specifically, if the level 0 scene is extracted, only the representative still image of the camera work scene that does not include camera fixing or camera shake can be quickly displayed from the structured video content. Conversely, it is also easy to display the representative still images of the scenes of level 1 to level 3, select some unnecessary scenes there, and cut them from the original video content. In that case, it is not necessary to temporally track all of the video content, and the editing work on the video content can be shortened.

Next, regarding the configuration of FIG. 1 described above, a series of procedures for providing a video content editing service will be described with reference to FIG.
It will be described based on the flowchart of FIG. First, step S41
In the above, when the video content editing support system 71 obtains the material video to be edited as an original video from the receiving device 81, the original video is converted into an MPEG2 format that can be handled in the video content editing support system 71 (step S42). ). Note that the original video is MPEG2 in advance.
If it is the compressed video content of the system, step S42
The step can be omitted. Next, in step S43, the contents of the original video compressed by the MPEG2 system are analyzed by the video content analysis device 82. Specifically, based on the motion vector contained in the compressed original video, individual scenes are separated by changing scenes including scene change and camera work, and for each scene, the visibility of the video is evaluated, for example. Judgment is made to one of four levels from 0 to evaluation level 3. Each divided scene is a video storage device
In addition to being sequentially accumulated in 84, the ID number corresponding to these scenes and the determination result (evaluation level) of the visibility of the image are
It is stored in the video ID storage device 85. As a result, the parameter of the action intentionally performed by the photographer is added to the moving image information of each scene as the evaluation level, but the addition of the evaluation level data does not increase the information amount so much. Rather, using this evaluation level data to reduce the amount of information presented to the end user when the original image is edited is advantageous for the end user because it does not have to handle a huge amount of data.

In this way, when each scene of the original image is automatically structured into an evaluation level according to the ease of viewing the image, the video content editing support system 71 selects what kind of image in editing the original image. Whether the information is desired to be displayed is displayed on the terminal 73 through the information network 72. At that time, the video content editing support system 71 provides the time of each evaluation level scene to the terminal 73 as time information so that the end user can judge in advance to what evaluation level the scene should be viewed. To do so. Upon receiving this, when the necessary condition is input from the terminal 73, the scene of the evaluation level that matches the condition is extracted from the video storage means 84,
A representative still image of this particular scene is sent from the content presenting means 83 to the terminal 73 (steps S44, S45).

For example, when a condition that only the scene of the evaluation level 0 is desired to be viewed is input from the terminal 73, the content presenting means 83 sequentially extracts the scenes of the evaluation level 0, and a typical still image in each scene is input to the terminal 73. Present. Further, for example, when the scene time of the evaluation level 0 is too short, the condition for wanting to see only the scenes of the evaluation level 0 and the evaluation level 1 is input from the terminal 73. In this case, the content presenting means 83 sequentially extracts scenes of evaluation level 0 or evaluation level 1 and presents a representative still image in each scene to the terminal 73. The order in which the still images are arranged and displayed can also be appropriately input from the terminal 73. In this way, the end user cuts off unnecessary scenes in advance and browses only representative still images of necessary scenes,
It is possible to reduce the amount of information presented.

As another procedure, regardless of the condition input from the terminal 73, only the scene of the evaluation level 0, which is a stable camera work section, may be automatically presented to the terminal 73 in advance. At that time, the content presentation device 83 may present the specific scene as a moving image as it is, but in order to reduce the amount of information transmitted to the terminal 73 as much as possible, here, an image representative of the specific scene is used as in the embodiment. It is preferable to present (still image) and the time information. Furthermore, when similar specific scenes are extracted, it is preferable to add an index to distinguish each scene.

When a representative still image of a specific scene is displayed on the terminal 73, the end user also refers to the scene time for each evaluation level, the index, etc. to edit what the original image is. Enter what you want to do as an edit command. That is, in step S46, when the edit command accepting device 88 constituting the video content editing support system 71 accepts the edit command from the terminal 73, the video composing device 89 is required from all the original images based on this edit command. Only scenes from the video storage device 84 are extracted, these scenes are connected appropriately, and the editing candidate video with additional data (other video, sound, subtitles, effects, etc.) inserted here is automatically created if necessary. To generate.
Then, the edited candidate video is sent to the terminal 73 owned by the user via the information network 72 (step S47).

At the subsequent step S48, the terminal 73 side confirms the edited candidate video image sent, and if there is a change such as addition or correction to the content of the editing process performed by the video image composition device 89, that effect is notified. Is sent to the video content editing support system 71 as a change command. When this is received by the edit command receiving means 88, the video composing device 89 regenerates a new edited candidate video based on the change command.
Then, the changed edited candidate video is sent to the terminal 73 via the information network 72 (step S4).
9). By repeating the procedure of steps S48 to S49, the end user can obtain a desired edited video by simply inputting a simple command to the terminal 73.

In this way, when the desired edited image is obtained,
In step S48, the end user sends a message to the effect that the editing process is no longer necessary from the terminal 73 side to the video content editing support system 71 via the information network 72 as a confirmation command of the editing process completion. In response to this, the video composition device 89 converts the edited candidate video for which the confirmation of the editing process has been obtained into the format desired by the user,
The edited image is distributed from the information network 72 to the terminal 73 (step S50).

In the above series of procedures, the end user confirms a typical still image from a scene of a particular evaluation level according to the ease of viewing the video and decides how to edit the original image provided in advance. While inputting operation from the terminal 73,
All you have to do is confirm the sent editing candidate image and input how the editing process should be changed from the terminal as well. Various devices 81 to 89 from receiving the original image to delivering the edited video are all provided in the video content editing support system 71 as the center device, and the video content editing support system is also provided.
Since 71 and the terminal 73 are connected by an information network as a communication means, it is not necessary for the terminal 73 to have any software function necessary for editing the original image, and large-scale software is required for each terminal 73. There is no need to install it for each.

As described above, according to the present embodiment, the receiving device 81 as the receiving means for receiving the original image to be edited,
A video content analysis device 82 as a video content analysis means for analyzing the content of the original video, a content presentation device 83 as a content presentation means for extracting and presenting a specific scene from the analysis content of the video content analysis device 82, and Edit command receiving means (edit command receiving device 8 for receiving an edit command based on a specific scene)
8), edited candidate video presenting means (video composing device 89) for presenting the edited candidate video in which the original video is edited in accordance with the editing instruction, and original image editing process performed by the video composing device 89. When a change related to the editing process is accepted, a new edited candidate image by the changed editing process is presented, and an editing process change acceptance unit (editing command acceptance device 88), and when an editing process completion command is accepted, it is presented immediately before that. Edited video distribution means for converting the edited candidate video into a desired format and distributing the edited video (video composition device 89)
And is configured.

In this case, the content of the acquired original image is analyzed, and the specific scene extracted from it is presented to the end user. While confirming the specific scene presented, the end user sends out as an editing command what kind of editing should be performed on the original image, this time the original image is edited according to this editing command The candidate video is presented to the end user. When the editing process change from the end user is accepted, the new edited candidate image by this modified editing process is presented to the end user, and when the editing process completion command is accepted, the edited candidate image presented immediately before that is received. Is converted into a desired format and distributed as an edited video.

As described above, when the original video is edited on the end user side, not the original video itself, but a specific scene obtained by analyzing the content of the original video or an edited candidate video that has been edited. All you have to do is send the appropriate commands for editing while watching. Therefore, it is possible to save time and effort to evaluate the content of the video content (original video) itself with eyes and edit the video content freely in a short time at the will of the end user.

Such an operation effect is that when an original image to be edited is received, a specific scene extracted by the analysis content of the original image is presented, and an editing command based on this specific scene is accepted, this editing is performed. When the edited candidate video, which is the original video edited according to the instruction, is presented, and the edit processing change is accepted, a new edited candidate video is presented by the modified edit processing, and the edit processing completion instruction is accepted. The method can also be achieved by converting the edited candidate video presented immediately before that to a desired format and distributing it as the edited video.

Further, in the present embodiment, the original image is divided into a plurality of scenes, and the image content analysis device 82 is configured so as to judge the visibility of the image for each scene from a plurality of evaluation levels, The content presentation device 83 is configured to present a scene of an evaluation level that matches the requested condition as a specific scene.

In this case, the received original video image is divided into a plurality of scenes, and the visibility of the video image for each scene is judged from a plurality of evaluation levels. Then, the scene of the evaluation level that matches the condition requested by the end user is presented to the end user as a specific scene. Therefore, the end user can confirm the scene of the required evaluation level before editing without evaluating each scene in the original image.

In addition, such an effect is obtained by dividing the original image into a plurality of scenes when presenting a specific scene extracted based on the analysis contents of the original image, and making the image easy to see for each scene. It can be achieved by a method of judging from the evaluation levels and presenting the scene of the evaluation level that matches the requested condition as the specific scene.

Further, the content presentation device 83 of this embodiment is configured to present a typical still image in a specific scene. In this case, the end user only needs to confirm the representative still image of the required specific scene, and it is possible to reduce the amount of information presented as the specific scene.

The above-described effects can also be achieved by presenting a typical still image in the specific scene when presenting the specific scene.

Incidentally, according to the video content editing support system 71 in the present embodiment, various services can be provided under various usage environments. For example, in cooperation with a travel agency or the like, a video storage medium such as a video tape from a traveler who is a user is stored at an airport or a station, and an input device (not shown) installed there stores information network 72. The material video data (original image) is transferred to the receiving means 81 of the video content editing support system 71 via. The video content editing support system 71 on the remote editing center side that has acquired the original image analyzes the content with the image content analysis device 82 and presents it to the user through, for example, the Internet environment. This allows the user to perform a desired editing process from the terminal 73 and receive the final edited video in a desired format.

In addition, the video content editing support system 71 of this embodiment is used for athletic meet, graduation ceremony, wedding, performance,
It can also be used as an editing support system for shot original videos at various events such as speeches for elections.

The present invention is not limited to the above embodiments, but various modifications can be made within the scope of the gist of the present invention. The thresholds described above do not all have to be the same value, and an optimum value is set for each determination condition.

[0090]

According to the video content editing support device of the first aspect and the video content editing support method of the fourth aspect of the present invention, the video content can be freely edited in a short time at the will of the end user.

According to the video content editing support device of the second aspect and the video content editing support method of the fifth aspect of the present invention, it is possible for the end user to confirm the scene of the required evaluation level prior to editing.

According to the video content editing support device of claim 3 and the video content editing support method of claim 6, it is possible to reduce the amount of information presented as a specific scene.

[Brief description of drawings]

FIG. 1 is a schematic explanatory view showing an overall configuration of a video content editing support device showing an embodiment of the present invention.

FIG. 2 is a schematic explanatory view showing the overall configuration of the video content analysis device of the above.

FIG. 3 is a flowchart showing a more detailed configuration and procedure of the video content analysis device of the above.

FIG. 4 is a flowchart showing a more detailed configuration and procedure of the section feature amount extraction unit in FIG. 3 above.

FIG. 5 is a schematic diagram showing a clustering method of motion vector directions and a calculation method of an intra-frame motion vector histogram bias degree.

FIG. 6 is a schematic diagram showing a procedure for dividing a shot into a plurality of sections.

FIG. 7 is a schematic diagram showing a procedure for extracting a key frame from each of the divided sections.

FIG. 8 is a graph showing an average luminance and an average color difference in the center of each frame of the same.

FIG. 9 is a graph showing the degree of similarity with the start frame of the same as above for the entire frame and the center of the frame.

FIG. 10 is a schematic diagram showing the relationship between section detection using color information and camerawork detection using motion information.

FIG. 11 is a schematic diagram showing an example in which video contents are automatically structured based on the visibility of the same as above.

FIG. 12 is a flowchart showing a series of processing procedures performed by the same video content editing support device.

[Explanation of symbols]

81 Receiving device (receiving means) 82 Video content analyzing device (video content analyzing means) 83 Content presenting device (content presenting means) 88 Editing command device (editing command receiving means, editing process change receiving means) 89 Video composition device (editing) Candidate video presentation means, edited video distribution means)

Continued front page    (72) Inventor Hideyoshi Tominaga             3-4-1 Okubo, Shinjuku-ku, Tokyo Waseda Univ.             Faculty of Science and Engineering (72) Inventor Ryoyuki Kodate             3-4-1 Okubo, Shinjuku-ku, Tokyo Waseda Univ.             Faculty of Science and Engineering (72) Inventor Kentaro Dobashi             3-4-1 Okubo, Shinjuku-ku, Tokyo Waseda Univ.             Faculty of Science and Engineering (72) Inventor Ryohei Ogushi             3-4-1 Okubo, Shinjuku-ku, Tokyo Waseda Univ.             Faculty of Science and Engineering (72) Inventor Tsuyoshi Hanamura             2-4-12 Okubo, Shinjuku-ku, Tokyo Stocks             Company media glue F term (reference) 5B075 ND12                 5C052 AA03 AB04 CC11 DD02 DD04                       EE03                 5C053 FA07 FA14 FA22 FA23 FA29                       GB38 HA29 JA21 LA11 LA14                 5C064 BA07 BB10 BC18 BC23 BD02                       BD08

Claims (6)

[Claims] 1. Receiving means for receiving an original image to be edited, and image content analyzing means for analyzing the content of the original image,
Content presenting means for extracting and presenting a specific scene from the analysis content of the video content analyzing means, editing command receiving means for receiving an editing command based on the specific scene, and editing the original video according to the editing command Edited candidate video presenting means for presenting the processed edited candidate video, and edit, when accepting a change of the editing process, presents a new edited candidate video by the changed editing process by the edited candidate video presenting means A process change acceptance unit and an edited video distribution unit for converting the edited candidate video presented immediately before the editing change completion command to a desired format and distributing the edited video as an edited video. Characteristic video content editing support device. 2. The image content analyzing means divides the original image into a plurality of scenes, determines the visibility of the image for each scene from a plurality of evaluation levels, and presents the content. The video content editing support device according to claim 1, wherein the means presents a scene of an evaluation level that matches a requested condition as the specific scene. 3. The video content editing support device according to claim 1, wherein the content presenting means presents a representative still image in the specific scene. 4. An original image to be edited is received, and a specific scene extracted by the analysis content of the original image is presented,
When an edit command based on this specific scene is accepted, an edited candidate video in which the original video is edited according to the edit command is presented, and when a change in the edit process is accepted, a new one is created by the modified edit process. When the edited candidate video is presented and the editing processing completion command is received, the edited candidate video presented immediately before that is converted into a desired format,
A video content editing support method characterized by delivering as an edited video. 5. When presenting a specific scene extracted based on the analysis contents of the original image, the original image is divided into a plurality of scenes, and the viewability of the image for each scene is selected from a plurality of evaluation levels. 5. The video content editing support method according to claim 4, wherein a scene of an evaluation level that matches the requested condition is presented as the specific scene while the determination is made. 6. The video content editing support method according to claim 4, wherein a representative still image in the specific scene is presented when the specific scene is presented.