KR101029437B1 - Method and System for Detecting Duplicate Moving Picture Files - Google Patents

Abstract

Video duplication detection method according to an aspect of the present invention that can detect the duplication by analyzing the content of the video itself comprises the steps of detecting at least one key frame for each video; Calculating a feature value of the at least one key frame using the detected luminance value of each region of the at least one key frame; Calculating an entire feature value for each video by using the calculated feature values of the at least one key frame; And determining whether the video is entirely overlapped by comparing the calculated total feature values.

Video, duplicate, detection, feature value

Description

Method and system for detecting video duplication {Method and System for Detecting Duplicate Moving Picture Files}

The present invention relates to video overlap detection, and more particularly, to a video detection method and system capable of detecting a video having the same content.

Recently, with the development of digital cameras / camcorders, camera mobile phones, and portable video players (PMP, MP4P, etc.), the production and utilization of video contents has become easier. In addition, due to the rapid proliferation of UCC (User Created Contents) services, digital video files produced by users are uploaded through UCC specialized sites, P2P (peer to peer) file sharing sites, personal blogs or mini homepages. The uploaded video file is shared with other users by downloading or streaming service through the Internet.

Digital video has a much larger file size than text-based content. Therefore, efficient management of storage is very important for a producer or user of video content and a business that needs to mass-produce user-oriented video content in order to provide a UCC service.

For efficient storage management, duplicate recording of video files with the same contents should be prevented. This is because storage of resources is wasted when video files having the same contents are duplicated and recorded. In order to prevent the existing video recording, duplicate video detection was analyzed through text analysis, playback time analysis, and source analysis related to video. However, since the types of videos are diversified and users are intentionally changing the videos little by little, it is almost impossible to detect them in a simple manner. Therefore, it is necessary to develop a method for detecting duplication of video by analyzing the contents of the video itself.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and it is a technical object of the present invention to provide a video overlap detection method and system capable of detecting video duplication by analyzing the content of the video itself.

Another object of the present invention is to provide a video duplication detection method and system that can more accurately detect video duplication by determining whether two videos overlap.

According to an aspect of the present invention, there is provided a method for detecting a video overlap, including detecting at least one key frame for each video; Calculating a feature value of the at least one key frame using the detected luminance value of each region of the at least one key frame; Calculating an entire feature value for each video by using the calculated feature values of the at least one key frame; And determining whether the video is entirely overlapped by comparing the calculated total feature values.

At this time, the video overlap detection method according to an embodiment of the present invention includes the step of adding a key frame every reference time, when the at least one key frame is not detected in the step of detecting the key frame. It features.

In addition, the video overlap detection method according to an embodiment of the present invention, in the step of detecting the at least one key frame, characterized in that for detecting the at least one key frame using the cut detection.

In the video overlap detection method according to an embodiment of the present invention, in the detecting of the at least one key frame, the at least one key frame is detected using a difference value of a gray histogram. At least one key frame is characterized in that the difference value of the gray histogram is greater than the reference value.

Meanwhile, before the detecting of the key frame, the method may further include decoding the video.

The method of claim 1, wherein the at least one key frame is an eye frame.

Also, in the method for detecting duplicate video according to an embodiment of the present invention, in the step of calculating a feature value of the at least one key frame, each region of the at least one key frame is determined by a principal component analysis (PCA). The characteristic value of the at least one key frame is calculated from a part of the luminance value.

At this time, the video overlap detection method according to an embodiment of the present invention, if the video is not overlapping as a whole, generating a sequence for the feature value of the calculated at least one key frame; And determining whether the video is partially overlapped through string matching of the generated sequence.

In addition, the video overlap detection method according to an embodiment of the present invention, in the step of generating the sequence, characterized in that for generating the sequence using the reference size of the calculated feature value of the at least one key frame.

In addition, the video overlap detection method according to an aspect of the present invention for achieving the above object comprises the steps of detecting at least one key frame for each video; Calculating a feature value of the detected at least one key frame; Generating a sequence of feature values of the at least one key frame; And determining whether the video is partially overlapped through string matching of the generated sequence.

In the video overlap detection method according to an embodiment of the present invention, in the calculating of a feature value of the at least one key frame, the at least one key frame may be generated using an average value of luminance values of regions of the at least one key frame. A feature value of one key frame is calculated.

According to another aspect of the present invention, there is provided a video overlapping detection system, including: a key frame detector configured to detect at least one key frame for each video; A frame feature value calculator configured to calculate a feature value of the at least one key frame by using the detected luminance value of each region of the at least one key frame; An overall feature value calculator configured to calculate an overall feature value for each video using the calculated feature values of at least one key frame; And a total overlap determination unit that determines whether the video is overlapped as a whole by comparing the calculated total feature values.

According to the present invention, by analyzing the contents of the video itself, it is possible to detect whether or not the duplicated video intentionally modified.

In addition, the present invention has the effect that it is possible to detect the overlap more accurately by determining whether the video overlaps in full and partial overlap in two steps.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a schematic block diagram of a video overlap detection system according to an embodiment of the present invention. As shown, the video overlap detection system includes a decoder 110, a key frame detector 130, a frame feature value calculator 150, a full overlap detector 170, and a partial overlap detector 190.

On the other hand, in one embodiment of the present invention, if most of the video is black before decoding the video, it is determined as an abnormal video, it may not be determined whether the entire and partial overlap. For example, when a bin of the maximum gray histogram is 93% or more, the video may be classified as an abnormal video. In determining the abnormal video, the abnormal video may be detected from the byte characteristics of the video using the MD5 hash value of the video.

The decoding unit 110 decodes videos encoded in various formats. Here, the videos used in the present invention may be videos encoded using various formats such as FLV, WMV, AVI, MPG, MP4, H.264, and the like, and the present invention is not limited thereto. Encoded videos may also be used.

In addition to the above-described encoding schemes, the videos used in the present invention may be videos having different formats such as frame rate, bit rate, resolution, and the like. In addition, videos with content that differs in color or contrast, videos with logos, text overlaid, videos with extraneous frames with different content, videos with different lengths of the same content, May be used. The decoding unit of the video duplication detection system according to an embodiment of the present invention serves to decode video encoded in various formats.

Meanwhile, in an embodiment, the decoding unit 110 may normalize the decoded video in advance after decoding the video. For example, the size, frame rate, etc. of the video may be the same to make the video robust.

The key frame detector 130 detects at least one key frame for each video. In other words, a key frame, which is a frame that can well represent the characteristics of a video, is detected. For example, the at least one key frame may be an eye frame. An eye frame means a frame composed only of intra coding blocks in the MPEG compression method. Accordingly, when the at least one key frame is an eye frame, the at least one key frame may be a key frame that can be well represented as a feature of a video because it consists only of an intra coding block.

In one embodiment, the key frame detection unit 130 may detect at least one key frame using the cut detection. Here, cut detection, that is, scene detection means detecting a scene corresponding to a suddenly changed frame among several frames of the video, and the video detection system according to an embodiment of the present invention cuts a key frame into a cut detection. Use the frame detected by

In another exemplary embodiment, the key frame detector 130 may detect at least one key frame using a difference value of the gray histogram. In other words, the key frame is detected by using a difference value of the black and white distribution between adjacent frames.

The gray histogram is a vector form of the distribution of gray values constituting the image. In other words, the gray value is set to 128 or 64 gray values, and the percentage of pixels in the range occupies the entire image as a vector value. It means what is shown. For example, as illustrated in FIG. 2, a gray histogram of a video may be represented, and the key frame detector 130 may detect a key frame using a difference value of gray histograms between adjacent frames. In addition, the gray histogram illustrated in FIG. 2 may use 16 gray bins, and the reference value of the difference value of the gray histogram may be 0.15.

Meanwhile, in at least one key frame according to an embodiment of the present invention, a difference value of the gray histogram may be greater than or equal to a reference value. In other words, a frame in which the difference value of the black-and-white distribution map between adjacent frames is more than the reference value can be detected as a key frame in this invention. That is, when the difference in the black and white distribution is larger than the threshold, it means that the screen is changed abruptly, and thus such a frame is detected as a key frame.

In addition, in one embodiment, the key frame detector 130 may arbitrarily add a key frame in the middle when the key frame is not detected by the cut detection or only one key frame is detected. At this time, the arbitrarily added key frame may be inserted every reference time, for example, if the key frame is not detected by the cut detection, the key frame may be forcibly inserted every 5 seconds. This is to prepare for the case where no key frame is detected to calculate the overall feature value that can represent each video when the key frame is not detected.

In one embodiment, the key frame detection unit 130 may not be detected as a key frame when a flashback occurs with a time difference smaller than the reference time by setting a reference time between key frames when a flashback occurs. have. For example, assuming that the minimum time of the flashback is 0.2 seconds, when the flashback occurs within 0.2 seconds, even if a key frame is detected by the cut detection, it may not be used as the key frame.

The frame feature value calculator 150 calculates a feature value of the at least one key frame by using the luminance value of each region of the at least one key frame detected by the key frame detector 130.

In an exemplary embodiment, the frame feature value calculator 150 may calculate a feature value of a key frame by using an average value of luminance values of respective regions of the at least one key frame. For example, the video frame may be divided into N * N blocks, and the feature value of the key frame may be calculated by using an average value of luminance values for each area of each block.

In one embodiment, the frame characteristic value calculator 150 may be configured to generate at least a portion of the luminance value of each region of the at least one key frame by a principal component analysis (PCA). The feature value of one key frame can be calculated. Here, the PCA refers to a statistical method of reducing a high-dimensional signal that is difficult to deal with mainly to a lower level, thereby making it easier to handle. In other words, PCA is commonly used to reduce dimensions when vectors are present, because significant important values are placed first and non-important values are distributed later. As a result, a statistical value of luminance values for each region in the frame is obtained, the value is converted by using the PCA, and the feature value of the key frame is calculated using only the main values.

For example, as shown in FIG. 3, an average value of luminance values of regions of (N * N + M * M) blocks, which is a sum of the number of blocks of N * N and the number of blocks of M * M, is calculated. By using, the feature value of the frame can be calculated, and the dimension can be reduced by using the PCA. In FIG. 3, the number of reference bits is used to calculate a feature value of a frame by reducing the dimension by using the PCA. Based on the threshold value "0" for each value corresponding to the reference number of bits, It can be seen that the value greater than "0" is "1" and the value smaller than "0" is "0" to quantize the value.

The total overlap detection unit 170 determines whether the video is overlapped as a whole by using the feature values calculated by the frame feature value calculator 150. As shown in FIG. 172 and the total overlap determination unit 174.

The global feature value calculator 172 calculates the global feature value for each video by using the feature values of at least one key frame calculated by the frame feature value calculator 150. For example, as shown in FIG. 3, the entire feature value calculator 172 calculates an average value of feature values of a key frame calculated using luminance values for respective regions of (N * N + M * M) blocks. A global signature representing a feature of the entire video may be calculated for each video. In other words, if 100 key frames of the video are detected by the cut detection, the feature values of 100 key frames may be calculated, and the total feature values are average values of the feature values of the 100 key frames. It can exist as one value that represents.

The total overlap determination unit 174 determines whether the video is overlapped entirely by comparing the total feature values calculated by the preliminary feature value calculator 172. In other words, the overall feature value calculator 172 compares the feature values representing the feature for each video to determine whether each video is overlapped.

On the other hand, as shown in Figure 7, looking at an example in which the video overlaps as a whole, as shown in Figure 7 (a) may occur when the video is the same as the entire overlap. In this case, the size, format, frame rate, etc. of the video may be the same.

Referring to FIG. 1 again, the partial overlap detector 190 determines the feature value calculated by the frame feature value calculator 150 when it is determined that the video is not overlapped by the entire overlap detector 170. Determining whether or not the video is partially overlapped using, as illustrated in FIG. 5, includes a sequence generator 192 and a partial overlap determiner 194.

The sequence generator 192 generates a sequence of feature values of at least one key frame calculated by the frame feature value calculator 150.

In other words, the sequence generator 192 generates a sequence for the feature value of the key frame for each video, and the partial overlap determination unit 194, which will be described later, is used to determine the feature value of the key frame generated by the sequence generator 192. The sequence is compared with each other to determine whether the video is partially overlapped.

In one embodiment, the sequence generator 192 may generate the sequence using the reference size of the calculated feature value of at least one key frame. For example, a sequence for the feature value of the key frame may be generated by using five digits of the first digit of the feature value of the key frame of the video. Here, the sequence generator 192 may generate a sequence of key frames only when the number of key frames is greater than or equal to a reference number. For example, the sequence generator 192 may generate a sequence of key frames only when 10 or more key frames are detected. have.

The partial overlap determination unit 194 determines whether the video is partially overlapped through string matching of the sequence generated by the sequence generator 192.

For example, as shown in FIG. 6, a video consisting of an "A, B, F" sequence and a video consisting of "A, B, C, E" are shown as "A, B" through string matching of the sequence. Since the sequences are partially identical, it can be seen that the two videos are partially overlapped. In addition, the video consisting of the sequence "D, C, E" and the video consisting of the sequence "A, B, C, E" have the same sequence as "C, E", so the two videos are partially overlapped. .

On the other hand, as shown in Figure 7, looking at an example in which the video is partially overlapping, as shown in Figure 7 (b) and (c) may be partially overlapping video, in this case As in the case of (b), the entirety of one video may be completely overlapped with a portion of another video to belong to another video, and as shown in (c) of FIG. 7, only a part of two videos may overlap.

Hereinafter, a video overlap detection method according to an embodiment of the present invention will be described with reference to FIG. 8.

First, after decoding a video, at least one key frame is detected for each video (S710). In other words, a key frame, which is a frame that can well represent the characteristics of a video, is detected. In one embodiment, the at least one key frame may be an eye (I) frame.

On the other hand, in one embodiment of the present invention, if most of the video is black before decoding the video, it is determined as an abnormal video, it may not be determined whether the entire and partial overlap. For example, when a bin of the maximum gray histogram is 93% or more, the video may be classified as an abnormal video. In determining the abnormal video, the abnormal video may be detected from the byte characteristics of the video using the MD5 hash value of the video.

Here, the at least one key frame may be detected using cut detection, and in particular, may be detected using a difference value of the gray histogram. For example, when the difference value of the gray histogram is greater than or equal to the reference value, the key frame may be detected. Can be detected. That is, when the difference in the black and white distribution is larger than the threshold, it means that the screen is changed abruptly, and thus such a frame is detected as a key frame.

In one embodiment, if no key frame is detected by the cut detection, or if only one key frame is detected, the key frame may be arbitrarily added to determine whether the video overlaps. At this time, the arbitrarily added key frame may be inserted every reference time, for example, if the key frame is not detected by the cut detection, the key frame may be forcibly inserted every 5 seconds. This is to calculate the total feature value representing each video in case the key frame is not detected, since the overall feature value that can represent each video cannot be calculated when the key frame is not detected.

In addition, according to an embodiment, when a flashback occurs, a reference time between key frames may be determined, and when the flashback occurs with a time difference smaller than the reference time, it may not be detected as a key frame. For example, assuming that the minimum time of the flashback is 0.2 seconds, when the flashback occurs within 0.2 seconds, even if a key frame is detected by the cut detection, it may not be used as the key frame.

Next, a feature value of the at least one key frame is calculated using the detected luminance value of each region of the at least one key frame (S720). In an embodiment, the feature value of the at least one key frame may be calculated using the average value of the luminance values of the at least one key frame for each region.

In one embodiment, a feature value of the at least one key frame may be calculated using a PCA. Here, the PCA refers to a statistical method of reducing a high-dimensional signal that is difficult to deal with mainly to a lower level, thereby making it easier to handle. In other words, PCA is commonly used to reduce dimensions when vectors are present, because significant important values are placed first and non-important values are distributed later.

Next, an overall feature value is calculated for each video using the calculated feature values of the at least one key frame (S730), and it is determined whether the video is entirely overlapped by comparing the calculated feature values. (S740). In other words, if 100 key frames of the video are detected by the cut detection, the feature values of 100 key frames may be calculated, and the total feature values are average values of the feature values of the 100 key frames and represent the video. It becomes one value and it can be determined whether or not the video overlaps as a whole by comparing the entire feature value.

Next, when the video is not entirely overlapped in step S740, a sequence of the calculated feature values of the at least one key frame is generated (S750), and the video is partially generated through string matching of the generated sequence. It is determined whether or not to overlap (S760).

In an example embodiment, a sequence of the feature values of the key frame may be generated using the calculated reference size of the feature values of the at least one key frame. For example, a sequence for the feature value of the key frame may be generated by using five digits of the first digit of the feature value of the key frame of the video. Here, the sequence of key frames may be generated only when the number of key frames is greater than or equal to the reference number. For example, the sequence of key frames may be generated only when 10 or more key frames are detected.

Hereinafter, a video overlap detection method according to another embodiment of the present invention will be described with reference to FIG. 9.

First, after decoding a video, at least one key frame is detected for each video (S810). In other words, a key frame, which is a frame that can well represent the characteristics of a video, is detected. In one embodiment, the at least one key frame may be an eye (I) frame.

Here, the at least one key frame may be detected using cut detection, and in particular, may be detected using a difference value of the gray histogram. For example, when the difference value of the gray histogram is greater than or equal to the reference value, the key frame may be detected. Can be detected. That is, when the difference in the black and white distribution is larger than the threshold, it means that the screen is changed abruptly, and thus such a frame is detected as a key frame.

Next, a feature value of the detected at least one key frame is calculated (S820). In an embodiment, the feature value of the at least one key frame may be calculated using the average value of the luminance values of the at least one key frame for each region. In other words, if 100 key frames of the video are detected by the cut detection, the feature values of 100 key frames can be calculated, and the feature values of the 100 key frames use the average value of the luminance values for each key frame region. Is calculated.

In one embodiment, the feature value of the at least one key frame may be calculated from a portion of the luminance value for each region of the at least one key frame by Principal Component Analysis (PCA). Here, the PCA refers to a statistical method of reducing a high-dimensional signal that is difficult to deal with mainly to a lower level, thereby making it easier to handle. In other words, the statistical value of the luminance value for each region in the frame is obtained, and this value is converted using the PCA, and then the feature value of the key frame is calculated using only the main values.

Next, a sequence of the calculated feature values of at least one key frame is generated (S830), and it is determined whether the video is partially overlapped through string matching of the generated sequence (S840). In an example embodiment, a sequence of the feature values of the key frame may be generated using the calculated reference size of the feature values of the at least one key frame. For example, a sequence for the feature value of the key frame may be generated by using five digits of the first digit of the feature value of the key frame of the video.

On the other hand, the above-described video overlap detection method is implemented in the form of program instructions that can be executed by various computer means may be recorded in a computer-readable recording medium. In this case, the computer-readable recording medium may include program instructions, data files, data structures, and the like, alone or in combination. On the other hand, the program instructions recorded on the recording medium may be those specially designed and configured for the present invention or may be available to those skilled in the art of computer software.

The computer-readable recording medium includes a magnetic recording medium such as a magnetic medium such as a hard disk, a floppy disk and a magnetic tape, an optical medium such as a CD-ROM and a DVD, a magnetic disk such as a floppy disk, A magneto-optical media, and a hardware device specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. The recording medium may be a transmission medium such as an optical or metal wire, a waveguide, or the like including a carrier wave for transmitting a signal specifying a program command, a data structure, or the like.

In addition, program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

On the other hand, those skilled in the art will appreciate that the present invention can be implemented in other specific forms without changing the technical spirit or essential features.

Therefore, it is to be understood that the embodiments described above are exemplary in all respects and not restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

1 is a schematic block diagram of a video overlap detection system according to an embodiment of the present invention.

2 is a diagram illustrating an example of obtaining a difference value between a gray histogram and a gray histogram.

3 is a diagram illustrating an example of calculating a feature value of a key frame by using luminance values of regions of the key frame.

FIG. 4 is a block diagram illustrating a detailed configuration of the entire overlap detection unit illustrated in FIG. 1.

5 is a block diagram illustrating a detailed configuration of a partial overlap detection unit illustrated in FIG. 1.

6 is a diagram illustrating an example of string matching of a video.

7 is a diagram illustrating an example of full duplication and partial duplication of a video.

8 is a flowchart illustrating a video overlap detection method according to an embodiment of the present invention.

9 is a flowchart illustrating a video overlap detection method according to an embodiment of the present invention.

<Description of part about main part of drawing>

100: video overlap detection system 110: decoding unit

130: key frame detector 150: frame feature value calculator

170: total overlap detection unit 190: partial overlap detection unit

Claims (23)

Detecting at least one key frame for each video; Dividing the detected at least one key frame into a plurality of blocks, and calculating a feature value of the at least one key frame by using an average value of luminance values for each region of the plurality of blocks; Calculating an entire feature value for each video by using the calculated feature values of the at least one key frame; Determining whether the video overlaps as a whole by comparing the calculated total feature values; Generating a sequence of feature values of the calculated at least one key frame when the video is not overlapped entirely; And And determining whether the video is partially overlapped by string matching of the generated sequence. The method of claim 1, wherein the detecting of the key frame comprises: And if the at least one key frame is not detected, adding a key frame every reference time. The method of claim 1, wherein in detecting the at least one key frame: And detecting at least one key frame using cut detection. The method of claim 1, wherein in detecting the at least one key frame: And detecting at least one key frame using a difference value of a gray histogram. The method of claim 4, wherein the at least one key frame And a difference value of the gray histogram is equal to or greater than a reference value. The method of claim 1, wherein before detecting the key frame, Decoding the video further comprising the step of decoding the video. The method of claim 1, And the at least one key frame is an eye frame. The method of claim 1, wherein the calculating of the feature value of the at least one key frame comprises: And a feature value of the at least one key frame is calculated from a part of luminance values for each region of the at least one key frame by Principal Component Analysis (PCA). delete The method of claim 1, wherein in generating the sequence: And generating the sequence by using a reference size of the calculated feature values of at least one key frame. delete delete A computer-readable recording medium having recorded thereon a program for performing the method according to any one of claims 1 to 8 and 10. A key frame detector for detecting at least one key frame for each video; A frame feature value calculator for dividing the detected at least one key frame into a plurality of blocks and calculating a feature value of the at least one key frame by using an average value of luminance values for each region of the plurality of blocks; An overall feature value calculator configured to calculate an overall feature value for each video using the calculated feature values of at least one key frame; A total overlap determination unit that determines whether the video is overlapped as a whole by comparing the calculated total feature values; A sequence generator for generating a sequence of the calculated feature values of the at least one key frame when the video is not overlapped as a whole; And Partial overlap determination unit for determining whether the video is partially overlapped by the string matching of the generated sequence Video duplication detection apparatus comprising a. 15. The apparatus of claim 14, wherein the key frame detector If the at least one key frame is not detected, the video overlap detection device, characterized in that for adding a key frame every reference time. The method of claim 14, And the at least one key frame is detected using cut detection. The method of claim 14, And the at least one key frame is detected using a difference value of a gray histogram. 18. The apparatus of claim 17, wherein the at least one key frame is And a difference value of the gray histogram is equal to or greater than a reference value. The method of claim 14, And a decoding unit for decoding the video. The method of claim 14, And the at least one key frame is an eye frame. The method of claim 14, And a feature value of the at least one key frame is calculated from a part of luminance values for respective regions of the at least one key frame by a principal component analysis (PCA). delete The method of claim 14, And the sequence is generated using a reference size of the calculated feature value of the at least one key frame.

Images