KR102138142B1 - Method of detection of video double compression based on prediction of encoding parameters - Google Patents

Abstract

The present invention includes a first step (s10) for primary encoding and decoding, a second step (s20) for forgery-modulating the decoded image in the first step, and a third step (s30) for re-encoding the forged image. , A bitstream analysis step (s40) for analyzing the parameters used in the re-encoding in the third step, and a parameter used for compressing the parameters detected and analyzed in the bitstream analysis step and various pre-stored images and comparison judgment It includes a comparative analysis step (s50),
By comparing the encoding parameters of encoder A and encoder B, which have different characteristics depending on the manufacturer and developer, it can be predicted that the re-encoding (falsification) has occurred. It relates to a video forgery detection method through the prediction of the encoding parameters, characterized in that possible.

Description

Method of detection of video double compression based on prediction of encoding parameters}

The present invention relates to a method for detecting forgery of a video through prediction of encoding parameters, which enables statistical analysis of a bitstream of a compressed video to predict encoding parameters, and enables encoding of video forgery detection based on encoding parameters. It relates to a video forgery detection method through parameter prediction.

A video is a group of still photos (frames) gathered together over time. The data size of the video is calculated based on one frame.

For example, if you have a 512×512 video, based on one frame of this video, it consists of horizontal: 512, vertical: 512 dots (pixels), and 1 pixel is 8 bits (256 levels). If we consider only the ratio occupied by gray, it becomes Width×Height×Bit (512×512×8). And since it is composed of 3 colors, in the case of RGB, it becomes Width×Height×Bit×Color(512×512×8×3), which is 6291456bit (6291456(bit) == 6144(kbit) == 6(mbit )).

Therefore, even if a relatively small sized video of 512x512 is uncompressed, it becomes a large size of 6 mega frames per frame. Therefore, the video is compressed and used.

Three techniques are used for video compression. The first is a temporal correlation between frames, the second is a spatial correlation, and the third is a characteristic of a view sensitive to low-frequency components (DCT transformation).

The compressed video as described above uses the double quantization effect or predicts forgery and alteration based on a difference between adjacent frames.

First, in the prediction method using the double quantization effect, it is assumed that an input coefficient of a DCT (Discrete Cosine Transform) is u, and when it is quantized, Q1(u) = [u/q1].

Here, q1 is a quantization parameter and [x] means an integer close to x.

The result of this restoration is Q1-1(Q1(u)) = [u/q1] x q1, and when it is recompressed, Q2(u) = [([u/q1] x q1)/q2] Becomes.

In the above Q2(u), [([u/q1] x q1)/q2] has only a special number of combinations, and has a different value from [u/q2], so that forgery can be predicted.

In addition, the method based on the difference between adjacent frames has a characteristic that a difference value between adjacent frames is significantly different from other regions in a region that is forcibly inserted or deleted. Therefore, the inserted region is predicted through the frame difference between the decoded images.

However, in the conventional method, it is possible to detect forgery or modulation of digital image data itself, but there is a disadvantage in that it is not easy to find out exactly where the forgery is generated in the forged image data. Moreover, in the case of so-called'temporal forgery' by inverting the order of occurrence of events (i.e., time frame) with respect to video data sequentially connected in time, there is a technical limitation that an additional method must be additionally used to detect forgery and alteration. . Such temporal forgery has a problem of intentionally replacing the stored image data in a time frame in order to conceal the crime time.

Korea Patent Office Registration Patent Publication No. 10-1746074 Korea Patent Office Registration Patent Publication No. 10-0538414

In order to solve the above-described problems of the prior art, the present invention can be predicted that re-encoding (falsification) has been performed by comparing encoding parameters of encoder A and encoder B having different characteristics according to manufacturers and developers. It is to provide a video forgery detection method through prediction of encoding parameters that can be predicted through statistical analysis of a bitstream A, which is difficult to secure.

In order to achieve the above technical problem, in one aspect of the present invention, the first step (s10) for primary encoding and decoding, the second step (s20) for forging and forging the decoded image in the first step, and forged A third step (s30) for re-encoding the image, a bitstream analysis step (s40) for analyzing parameters used during re-encoding in the third step, and parameters and pre-stored parameters analyzed and detected in the bitstream analysis step Characterized in that it comprises a comparison analysis step (s50) for comparing and determining the parameters used when compressing various images.

In addition, the first step of encoding and decoding the first step (s10) is a step of decompressing the image, characterized in that the parameters used when compressing the image are stored in the first storage unit.

In addition, the first storage unit is characterized in that the encoder is classified and stored according to the manufacturer, and parameters used according to the classified encoder can be stored by the lower storage module.

In addition, the encoding parameter is characterized by including intra prediction of 16x16 or 4x4 macro blocks according to international standards.

In addition, when the image is forged in the second step of forgery (s20), the original bitstream image does not remain, and only the forged bitstream remains.

In addition, the encoder used in the third step (s30) to re-encode the forged image is different from the encoder used in the first step, so that the encoder algorithm of the original image cannot be identified.

In addition, in the bitstream analysis step (s40), after receiving video data through a bitstream, separating each channel, and encoding each channel in parallel to have different bit rates to analyze a real-time video stream having multiple bit rates. It is characterized by.

In addition, the bitstream analysis step (s40) is characterized in that by analyzing the re-encoded image to determine what parameters are used, and storing the detected data in the second storage unit.

In addition, the comparative analysis step (s50) compares and analyzes information about the parameters stored in the bitstream analysis step (S40) and the parameters stored in the first storage unit (10) to detect forgery and to use it in the image before forgery. It is characterized by predicting the encoding parameters.

In addition, in the bitstream analysis step (S40), forgery detection and encoding parameters can be predicted based on the statistical data.

In addition, the bitstream analysis step (S40) is characterized in that the statistics by the statistical data calculation module 30 based on the data in the first storage unit (10).

In addition, in the bitstream analysis step (S40), the original input image or bitstream A associated with bitstream B cannot be obtained by comparing and analyzing the encoding parameter obtained with the statistical characteristics of encoder A and the encoding parameter of bitstream B. It is characterized in that it is possible to predict that the video is encoded in a device having encoder A.

According to the embodiments of the present invention as described above, it can be predicted that re-encoding (falsification) is performed by comparing encoding parameters of encoder A and encoder B having different characteristics according to manufacturers and developers, and secured in a real environment Bitstream A, which is difficult to perform, has a predictable effect through statistical analysis.

1 is a conceptual diagram schematically illustrating a video forgery detection method,
2 is a diagram schematically showing the types of 16x16 Intra Prediction Mode included in the video compression standard,
3 is a diagram schematically showing the types of 4x4 Intra Prediction Mode included in the video compression standard,
4 is a block diagram schematically showing a video forgery detection method through prediction of encoding parameters according to an embodiment of the present invention,
5 is a diagram schematically showing a video forgery detection apparatus through encoding parameter prediction according to an embodiment of the present invention,
6 is a diagram schematically showing a method for predicting encoding parameters and detecting forgery and alteration based on statistical characteristics according to an embodiment of the present invention.

The present invention can be applied to various changes and can have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. In describing each drawing, similar reference numerals are used for similar components.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from other components. For example, the first component may be referred to as a second component without departing from the scope of the present invention, and similarly, the second component may be referred to as a first component. The term and/or includes a combination of a plurality of related described items or any one of a plurality of related described items.

When an element is said to be "connected" or "connected" to another component, it is understood that other components may be directly connected to or connected to the other component, but there may be other components in between. It should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that no other component exists in the middle.

The terms used herein are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms related to “comprises”, “haves”, and the like are intended to indicate that there are features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, and one or more other features. It should be understood that the existence or addition possibilities of fields or numbers, steps, operations, components, parts or combinations thereof are not excluded in advance.

Unless defined otherwise herein, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as meanings consistent with the contextual meaning of the related art, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined herein.

1 is a conceptual diagram schematically illustrating a video forgery and alteration method.

As shown in FIG. 1, in general, forgery of a video is a step of decoding a compressed video to be forged (s100), a video forgery step (s200) using a video editing tool, and a step of re-encoding the forged video. (s300).

At this time, the input video is re-encoded using a different encoder than the extruded encoder. For example, a video encoded with a car black box is re-encoded using a video editing software for PC. In this process, it is not possible to check the algorithm of the car black box encoder, and it is difficult to modify the video encoder of the video editing software for PC. Therefore, making the encoding parameters of the car black box video and the video encoding software for the PC the same is the same. It is quite difficult.

The step of decoding the video (s100) is a step of decoding the compressed video file to be reproduced again using an encoder program.

First, the video is a number of still photos (frames) collected on a time axis, and the data size of the video is calculated using horizontal*vertical*pixels*colors, which is a small size of 512*512 when uncompressed. 'S video also has a large size of 6 mega. In addition, when providing a video service, 24 frames per second must be guaranteed, which requires a large capacity of 144 megabytes per second.

Therefore, the encoding that compresses the image using a program called an encoder is performed and decoded using a decoder to reproduce it.

The encoding proceeds in accordance with the video compression standard. In the video compression international standard, only the functions to be supported by the decoder are defined, and there is no restriction on the algorithm development or implementation of the encoder. For example, the H.264 standard should support the following modes for intra prediction of 16x16 or 4x4 macro blocks.

First, 16x16 Intra Prediction Mode has four types shown in FIG. 2, and 4x4 Intra Prediction Mode has a total of nine types as shown in FIG.

However, the international standard does not define a method to select between 16x16 intra prediction mode and 4x4 intra prediction mode, which can be freely selected in the course of developing an encoder. There are various types of inter prediction block size (block decomposition), inter block skip, search area, reference frame, and QP change of rate control.

Using all the features and modes described in international standards, you can compress video efficiently. However, since video encoding requires a high amount of computation, if implemented in consideration of the efficiency of all functions and modes, the price of the product may increase or the computation speed may be slow, which may cause difficulty in real-time encoding. To solve this problem, many encoders use high-speed algorithms to selectively select functions and modes. For example, in order to reduce the amount of computation, 4x4 intra prediction is not performed in intra prediction, and only 16x16 intra prediction can be performed unconditionally.

In this case, the compression efficiency is somewhat reduced, but since 4x4 intra prediction is not performed, it is possible to significantly reduce the encoding time due to the computational effect.

In addition, only three of the four modes supported by 16x16 intra prediction can be considered to reduce the computation amount. It is possible to reduce the execution time by limiting the mode that consumes a lot of computation, and all four modes are considered in 16x16 intra prediction, but in order to reduce the computation, it is possible to check the three modes first by selectively prioritizing them. If not, the fourth mode can be considered.

As such, the standard defines the minimum functions and modes that the decoder must support, but the encoders can use these functions and modes selectively. These applications can be applied as is in all functions and modes. For example, B frame is supported in the high profile of H.264, but B frame may not be used.

There are many ways to speed up the function and mode selection algorithm or to limit the function and mode, as well as the methods described above, and it is a common technology for manufacturers and developers who develop video compression.

In the present invention, the function and mode finally selected are defined as encoding parameters by using a method for speeding up the function and mode selection algorithm or limiting the function and mode as described above.

As mentioned above, a method for determining a high-speed encoding parameter or selecting a limited encoding parameter is a common technique for video encoder manufacturers and developers.

However, different algorithms are developed and applied by manufacturers and developers who develop encoders. Therefore, different encoders compress video according to different encoding parameters.

Algorithms for determining or selectively limiting encoding parameters at high speed are determined in a typical encoder development process, but it is difficult to develop an efficient algorithm for each device and application. For example, embedded devices such as automobile black boxes and mobile phones require real-time encoding using low voltage and low memory, so this is an area that requires know-how in encoder development. Therefore, it is possible to check the parameters of the encoder through the encoded resultant bitstream, but it is a difficult problem to create an encoder that makes the same algorithm to generate the same. As a result, the encoding parameter acts like a human fingerprint, and it is possible to distinguish the type of encoder through comparison of encoding parameters.

Therefore, in the present invention, the image that has been encoded and decoded as described above is re-encoded using an encoder different from the previously encoded encoder after undergoing a forgery modulation process. At this time, the information on the parameters used for encoding is compared with the parameter information used for re-encoding to predict that the re-encoded image is forged.

The present invention, as shown in Figure 4, the first step (s10) for primary encoding and decoding, the second step (s20) for forging and forging the decoded image in the first step, and re-encoding the forged image The third step (s30), the bitstream analysis step (s40) for analyzing the parameters used during re-encoding in the third step, and the parameters detected and analyzed in the bitstream analysis step and various images stored in advance are compressed. It includes a comparison analysis step (s50) to determine the comparison with the parameters used when doing.

The first step of encoding and decoding (s10) is a step of decompressing an image, and parameters are used when compressing the image.

Assuming that the encoder used for encoding is A, encoder A performs encoding using parameters 1, 2, and 3, which can confirm the use of parameters 1,2,3 by analyzing the bitstream. The types of parameters identified by the bitstream are stored in the first storage unit 10 as described above.

The first storage unit 10 classifies the types of encoders, and classifies and stores parameters used according to the classified encoders. As described above, the stored information is provided at a later comparative analysis step (s50) to determine and predict whether the provided image is forged or forged.

Since the encoders use different parameters for each manufacturer and model, they are classified and stored in the first storage unit 10 by manufacturer or model, and encoded in the lower storage module 11 included in the first storage unit 10. It is possible to save the parameters used in.

In other words, when the encoder manufactured by the manufacturer A was used to encode the image in the first step (s10) of the primary encoding and decoding, the sub-storage module 11 uses parameters 1 and 2 used in the model a. Save 3.

If a b model other than a model is used, the sub-storage module 11 can store parameters 1,2 and 3 of the b model.

The second step of forgery (s20) is a step of forgery, if necessary, the compressed and decompressed image in the first step (s10) of the first encoding and decoding.

For example, a video captured by a specific user with a black box (encoder A) is against them and forged with PC video editing software (encoder B), and bitstream B is submitted as evidence.

As the above forged image does not remain in bitstream A, only bitstream B remains, so it is not possible to check whether the image is forged or not, so the present invention enables to check forgery forgery through a later comparative analysis step (s50). .

The third step (s30) of re-encoding the forged image is a step of re-encoding the image generated in the second step (s20) of forgery using a new encoder.

The encoder used in the first step and the encoder used in the third step are different from each other. For example, it refers to re-encoding an image encoded in a car black box using video editing software for PC.

When re-encoding as described above, it is impossible to check the algorithm of the car black box encoder, and it is difficult to modify the video encoder of the video editing software for the PC. Therefore, the encoding parameters of the car black box video and the video encoding of the video editing software for the PC are encoded. It can be seen that the parameters are not the same.

Therefore, the primary encoding and the re-encoding encoder means that different encoders are used.

The bitstream analysis step (s40) analyzes the re-encoded image to determine what parameters are used and detects them. The bitstream receives video data, separates them for each channel, and has different bit rates. Each channel is encoded in parallel to analyze a real-time video stream having a multi-bit rate, so that parameters used in the re-encoding step can be detected and stored in the second storage unit 20.

The data stored in the second storage unit 20 stores information on parameters used while using a model of manufacturer B used while re-encoding.

Since the information on the parameters stored in the first storage unit 10 and the information on the parameters stored in the second storage unit 20 may be the same or different, the first storage unit 10 in a later comparative analysis step (S50). ) To compare and analyze the data stored in the second storage unit 20.

In the comparison analysis step (s50), forgery detection is performed by comparing and analyzing the information calculated in the bitstream analysis step (S40) and the parameters stored in the first storage unit 10, and encoding used in the image before forgery modulation is performed. Make parameters predictable.

That is, although it is impossible to obtain the original input image or bitstream A related to the analyzed bitstream B by re-encoding, we know that the image is encoded on the device with encoder A, and we can secure various images encoded with encoder A. If possible, it is possible to check the statistical characteristics of the encoding parameters through repetitive experiments, thereby predicting the encoding parameters and detecting forgery and alteration through the comparative analysis step (S50).

Meanwhile, in the bitstream analysis step (S40), forgery detection and encoding parameters can be predicted based on the statistical data.

Therefore, it is necessary to calculate the statistical data, which is based on the data in the first storage unit 10, which can be made by the statistical data calculation module 30.

The statistical data calculation module 30 collects information about the encoder and statistically calculates information about the encoder to obtain information about parameters stored in the first storage unit 10 and the second storage unit 20. By comparing and analyzing, it is possible to detect forgery and detect information on parameters used in forgery (see FIG. 5).

The encoder is characterized in that the parameters used according to the manufacturer and model are different, and the data calculated by the statistical data calculation module 30 is statistically based on the data stored in the first storage unit 10.

Explained through specific examples,

manufacturer Model Whether to use 16x16 intra prediction mode One 2 3 4 A a o o o x o o o x

Table 1 above records the use of 16x16 intra prediction mode by encoding various images with encoder A and analyzing the resulting bitstream. The above table is marked with an o-table if the corresponding mode has been used even once, and is marked with an x if it has never been used. In the a model of manufacturer A, you can see that mode 4 was never used.

manufacturer Model Whether to use 16x16 intra prediction mode One 2 3 4 A b 385,332 30,440 95,570 474 327,809 83,071 90,668 243 c 789,581 59,558 58,677 53,564 454,226 36,377 186,536 83,552

Table 2 is a statistical result of recording the number of times each mode was used in 16x16 intra prediction mode by encoding various images with encoder A and analyzing the resulting bitstream.

From the above results, it can be seen that the b model and the c model of manufacturer A use all the 1,2,3,4 modes, unlike the a model above. However, the probability of using the 4th mode of the b model is significantly lower than that of the c model, and it can be seen that it has statistically different characteristics.

Since the comparison of these statistical characteristics is much dealt with in existing statistics, detailed description will be omitted in the present invention.

As described above, after the analysis of the statistical characteristics is completed, it is possible to compare the encoding parameters obtained with the statistical characteristics of encoder A and the encoding parameters of bitstream B. When the two statistical characteristics are different, it can be seen that they are encoded by different encoders, and it can be confirmed that bitstream B is not encoded by encoder A.

Therefore, it is possible to detect whether the encoding parameter calculated through the re-encoded bitstream B is forged, and as described above, the original input image or bitstream A related to bitstream B cannot be obtained, but the image is encoder A It has the feature of predicting the fact that it is encoded on a device with.

Accordingly, according to the present invention, as shown in FIG. 6, an image encoded using parameters 1, 2 and 3 of the encoder manufactured by manufacturer A is subjected to a forgery modulation process, and parameters 1-1 and 2 of encoder manufactured by manufacturer B are fabricated. When encoding by -1,3-1, the information on the parameters 1,2,3 used during the first encoding cannot be secured, so the statistical data calculation module 30 provides statistically calculated parameters through various images. The information is matched with the re-encoded parameters 1-1, 2-1, and 3-1 to determine whether forgery or alteration has occurred, and to predict which parameter is used for the first encoded parameter.

Although it has been described with reference to preferred embodiments of the present invention, those skilled in the art may variously modify and change the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that you can.

s10: a first step () for primary encoding and decoding, and
s20: second step of forgery and forgery
s30: third step of re-encoding
s40: Bitstream analysis step
s50: Comparative analysis step

Claims (12)

The first step (s10) for primary encoding and decoding, the second step (s20) for forgery-modulating the decoded image in the first step, the third step (s30) for re-encoding the forged image, and the first A bitstream analysis step (s40) for analyzing the parameters used when re-encoding in step 3, and a comparative analysis for comparing and determining the parameters that were analyzed and detected in the bitstream analysis step and the parameters used when compressing various pre-stored images. Step (s50),
The bitstream analysis step (S40) is statisticalized by the statistical data calculation module 30 based on the data in the first storage unit 10, and forgery detection and encoding parameters are predicted based on the statistical data. By making it possible,
The bitstream analysis step (S40) analyzes the re-encoded image to determine what parameters are used, stores the detected data in a second storage unit, and encoding parameters and bitstream B obtained as statistical characteristics of encoder A. By comparing and analyzing the encoding parameters of, the original input image or bitstream A related to bitstream B cannot be obtained, but it can be predicted that the image is encoded in a device having encoder A,
In the comparison analysis step (s50), forgery detection is performed by comparing and analyzing the information calculated in the bitstream analysis step (S40) and the parameters stored in the first storage unit 10, and encoding used in the image before forgery modulation is performed. By predicting the parameters,
The first storage unit classifies and stores an encoder according to a manufacturer, and stores parameters used according to the classified encoder by a lower storage module,
The second storage unit detects video forgery and alteration through prediction of encoding parameters, characterized in that it stores parameters used in the re-encoding step. According to claim 1,
The first step of encoding and decoding (s10) is a step of decompressing an image, and detecting forgery and alteration of a video through prediction of encoding parameters, characterized in that a parameter used when compressing the image is stored in a first storage unit. Way. delete According to claim 1,
The encoding parameter includes intra prediction of 16x16 or 4x4 macro blocks according to international standards. According to claim 1,
In the second step (s20) of forgery and forgery, when forgery the video, the original bitstream video does not remain, but only the forged video bitstream remains. According to claim 1,
The encoder used in the third step (s30) of re-encoding the forged video is different from the encoder used in the first step so that the encoder algorithm of the original video cannot be checked, and thus the video forgery through encoding parameter prediction is prevented. Detection method. According to claim 1,
The bitstream analysis step (s40) is characterized by analyzing the real-time video stream having multiple bit rates by encoding video channels in parallel to have different bit rates after receiving video data through the bit stream and separating each channel. Video forgery detection method by predicting encoding parameters. delete delete delete delete delete

Images