IT202100011090A1 - PROCESS OF INCORPORATING A WATERMARK INTO A VIDEO AND CORRESPONDING PROCESS OF EXTRACTING A WATERMARK FROM A VIDEO - Google Patents

Description

DESCRIPTION of the industrial invention entitled: ?Procedure for incorporating a watermark into a video and corresponding procedure for extracting a watermark from a video?

DESCRIPTION

Technical field

The present invention falls, in general, in the sector of copyright protection of multimedia content that can be published on the web. In particular, the invention relates to a process for incorporating a watermark into a video and to a corresponding process for extracting a watermark from a video.

Known technique

Watermarking processes are known in the sector which consist in the insertion of certain additional information within a digital content, typically multimedia. Such additional information may be extracted and later retrieved from the digital content, at the discretion of the access policy to such additional information. Typically, the additional information is not immediately visible, or at least interpretable, by a third party unrelated to the digital content.

The subjects who have access to this additional information are able to process it in order to obtain useful results for specific purposes, for example, to determine whether the digital content is been modified or to reconstruct the digital content following a degradation sometimes induced by appropriate attacks (correction codes).

The main goal of carrying out attacks against watermarking algorithms ? to destroy or degrade additional information incorporated in digital content.

There are attacks of different nature, for example, geometry attacks, noise attacks and coding attacks.

Noise attacks are attacks that obscure the image of digital content, creating confusion, and generate grain that replaces micro-sections of the original image. ? a type of annoying attack that significantly reduces the quality? of the image and is particularly visible.

Geometric attachments are capable of decisively modifying the image of the digital content, performing lateral, upper or lower cuts, or transformations such as rotations, or displacements, or even mirroring.

The coding attacks aim to significantly cut detailed information from an original digital content without perceptibly affecting the visual content, at least with a single processing (if successive coding operations are performed, the quality of the video is degraded).

The general objectives that lead to the use of watermarking techniques can be, among others, the protection of the digital copyright of a content (Digital copyright protection), the monitoring of the diffusion and reproduction of the digital content (Broadcast monitoring), the authentication of the digital content (Data authentication) and the identification of large files or data structures using truncated information (Digital fingerprinting).

The stages of watermarking processes are usually composed of watermark embedding, watermark extraction and watermark detection. These phases may possibly change according to the technique used.

The properties? that characterize digital watermarking are the perceptive transparency, the robustness, and the capacity.

The watermarking algorithm can? be classified according to various aspects, for example, the type of data that are incorporated, the extraction technique and so on? Street.

An exhaustive classification allows to recognize the various watermarking algorithms that are present in the literature. Basing the classification on the type of embedded data we get: Text watermark, Image watermark, Video watermark and Audio watermark.

Instead, basing the classification on the basis of the extraction technique we obtain: Blind or public watermarking, Non-blind or private watermarking and Semi-blind watermarking.

In public watermarking, isn't it? an original signal is needed during a watermark detection process. ? only a secret key is required. In public watermarking, users of the content are allowed to detect the watermark.

With private or non-blind watermarking, ? required an original signal for watermark detection. In private watermarking, users are not allowed to detect the watermark.

With semi-blind watermarking, sometimes more information is needed to correctly detect the watermark. Some algorithms need to access the original digital content immediately after adding the watermark, which is? called published watermarked signal.

Basing the classification on the basis of robustness characteristics, we obtain: Robust watermark, Fragile watermark and Semi-fragile watermark.

A watermark algorithm? robust if it can? survive after common signal processing operations such as filtering and lossy compression.

A brittle watermark algorithm should be able to detect after any change in the signal and also be able to identify the signal before the change. Is this type of watermark used for authenticity? or verification of the original content. Can? be useful when you want to demonstrate that the content ? been attacked.

A Semi-fragile watermark algorithm? sensitive to a certain degree of modification of a watermarked signal.

The techniques for carrying out the watermarking operation on digital video content can be divided into two macro-categories. Those that are done in the spatial domain and those that are done in the frequency domain.

In the spatial domain there are the Least significant bit modification (LSB modification) and the Correlation-based technique.

In the frequency/quantization domain there are:

- the Discrete Fourier Transform (?Discrete Fourier Transform?, DFT);

- the Discrete Cosine Transform (?Discrete Cosine Transform?, DCT);

- the Discrete Wavelet Transform (?Discrete Wavelet Transform?, DWT); And

- Contourlet Transformata (?Contourlet Transform?, CT).

For example, in figure 1 ? illustrated an example of decomposition through a 3-level DWT Discrete Wavelet Transform.

Some technologies coming from the video field are:

- Motion Vector: Movement vector of the image areas. It is used to optimize the encoding of videos, which need a quantity? considerable space to be stored ?raw? frame by frame on disk. The idea in this case? that of breaking down the video into main frames, also called i-frames, and dependent frames. Where the addiction? expressed as movement of image areas of the main frame or i-frames within the sequence. This allows you to save a considerable amount? of memory. Without this technique it would not be possible to manage this quantity? of data on very high resolution, high frame rate video;

- main frames, also called i-frames, and secondary frames: i-frames are the frames from which all the other ?inter-laced? (one or a few for scene change).

Speaking now of the Discrete Fourier Transform, DFT, it ? the discrete transform pi? used, in particular it is used to carry out the Fourier analysis in various practical applications. It decomposes a periodic signal into components more? simple in order to guarantee a better analysis.

This transform is used when you want to access the geometric characteristics of the image. Since the? Image is decomposed into its sinusoidal components, ? more easy to examine or process certain frequencies of the image, and thus influence the geometric structure in the spatial domain.

Speaking now of the Discrete Cosine Transform DCT, it ? a technique for converting a signal into elementary frequency components. It represents an image as the sum of sinusoids of various sizes and frequencies. His definition? the following:

The Discrete Cosine Transform DCT allows you to decompose an image into different frequency bands, making it easier to embed watermark information into image frequency bands. For embedding, mid-frequency bands are preferred. Increases robustness against attacks which, in many cases, can distort lower frequencies. high. Furthermore, by choosing the medium frequencies, the louder parts are avoided. visually important image, concentrated in the frequencies pi? low. To prevent the hidden information from being extracted directly from the transformed domain, watermarks are typically incorporated by changing the relationship of adjacent blocks of the original image's mid-frequency coefficients, rather than use an additive operation. Usually the digital watermark transformation to DCT is applied to the luminance signal. The DCT transform is based on dividing the original image into 8 x 8 blocks of pixels and applying the transform individually to each block. The mean frequency range from the DCT coefficients is extracted by applying a 2D mean frequency mask.

One of the main problems that digital content producers have to face today? illegal reproduction of content.

Digital content is usually recorded, or copied, and re-uploaded on the web, effectively illegally appropriating the rights to exploit the digital content.

In the light of the previous explanations regarding the technologies known and used in the watermarking sector, currently there are no watermarking algorithms sufficiently robust and light in computational terms, which allow their use on a large scale and which are able to systematically address and reliably the fight against piracy and misappropriation of copyrighted digital content.

Summary of the invention

An object of the present invention ? therefore that of providing a process for embedding a watermark in a video and a corresponding process for extracting a watermark from a video that are robust and lightweight at a computational level, which can be used to fight piracy and misappropriation of copyrighted digital content.

The present invention could be exploited in various usage scenarios, which exploit its peculiarities? and solve some problems such as the protection of content, the security of identities? of the authors, the scalability? of the service against the multitude of video reproduction portals already? spread on a global scale.

For practical reasons? Three possible scenarios have been developed:

1) Confederate platform scenario

2) film scenery

3) user scenario

In scenario 1, as shown in Figure 2, a private user or a company 2, which ? professional video content producer or not, pu? request the need to secure his work 3 by ensuring that he or someone on his behalf is recognized as the author, using some system to apply, maintain and recognize the copyright information. The private user or the company can therefore? making use of the process for incorporating a watermark according to the invention, by requesting the protection service from a platform 4 available on the net and easily accessible. The platform available on the net implements a system of incorporation and extraction in an individual and separate manner. When the private user or the company requests the copyright information embedding service, will he upload? on the platform the movie and it will come? modified - that is, will it come? information on possession -, in an imperceptible way to the human eye, so that the information resists eventual attacks of some kind, and in a way to extract at a later time what? That ? been imprinted. The edited video will be then released to the applicant so that he can then distribute it as he wishes, and to whomever he wishes. In such a scenario, an attacker 6 who ? managed to get his hands on video content not owned by him? 8, if he decides to upload the video on one of these portals of the confederate platform, he would be denied the actual upload because? the portal, upon acceptance would go to perform a check on the video in the upload request and would find traces on the paternity? not legitimate of the author. It, the video streaming site, to recognize possession or non-possession would call up an API, an application software interface, made available and accessible only to confederated actors, which recalls the process of extracting a watermark from a video, which checks if the video has some form of copyright. The video reproduction sites will therefore be able to act as a third party between the protection procedure and the actor requesting the service.

In scenario 2, the actor requesting the digital content protection service ? a cinematographic company or similar which, before releasing an outgoing film, imprints inside each of the copies distributed, through the watermark process of a video, an information which specifies which one? the destination cinema. In this way, if the film is used in violation of copyright, it would be possible to extract the information imprinted inside and trace which cinema allowed its illicit distribution. Through the watermark process of a video, specific information on the destination cinema is first engraved, before release, then, at a later time, when the video will be released. unduly cloned, sar? It is possible to extract that hidden information through the process of extracting the watermark from a video.

In scenario 3, who requests content protection? a common user. He has recorded a video and intends to share it with other entities or people but in a secure manner, making sure that no one can appropriate the content and impersonate it. For this reason, the user decides to give clear and concise information indicating who will be? the recipient, through the process of embedding a watermark in a video. As in the previous scenario, this way, if the video is distributed again under another name, the ordinary user could take that distributed copy and extracting his hidden information inside, through the process of extracting a watermark from a video, would be able to figure out who? that violated his trust. In this case, the process of embedding a video acts in its embedding part before distribution, and in its validation part, through the process of extracting a watermark from a video, after the infringement of ownership.

The above and other objects and advantages are achieved, according to one aspect of the invention, by a process for incorporating a watermark into a video and a corresponding process for extracting a watermark from a video having the characteristics defined in the respective independent claims. Preferred embodiments of the invention are defined in the dependent claims, the content of which? to be understood as an integral part of this description.

Brief description of the drawings

The functional and structural characteristics of some preferred embodiments of a process for incorporating a watermark into a video and of a process for extracting a watermark from a video according to the invention will now be described. Reference is made to the attached drawings, in which:

- figure 1 shows an example of decomposition through a 3-level DWT Discrete Wavelet Transform;

- figure 2 illustrates an exemplary scenario of use of the present invention;

Figure 3A illustrates a block algorithm of an exemplary embodiment of a method of embedding a watermark in a video according to the invention; figure 3B illustrates a block algorithm of an exemplary embodiment of a method of extracting a watermark in a video according to the invention;

- figure 4 illustrates an explanatory diagram of an exemplary incorporation of an optical information within a main frame, through the use of the cryptographic key;

figure 5 illustrates a block algorithm explaining a further exemplifying embodiment of a process for incorporating a watermark into a video according to the invention;

Figure 6 illustrates the arrangement of the coefficients in the frequency domain obtained from a main frame;

figure 7 illustrates an exemplary block algorithm in which a process of embedding a watermark in a video and the process of extracting watermarks from a video are applied to a video at the same time;

figure 8 illustrates an example showing the concept of subdividing an optical information;

figure 9 illustrates an example of partially compromised optical information and the same optical information following the replacement of some known regions thereof;

- figure 10 illustrates an example of an association table stored in a database; - figure 11 illustrates an example of pseudocode for a phase of embedding a watermark;

- Figure 12 illustrates an example of pseudocode for performing a DWT transform;

- figure 13 illustrates an example of a first part of pseudocode for an extraction of a watermark from a frame;

figure 14 illustrates an example of a second part of pseudocode for an extraction of a watermark from a frame, which continues the first part of figure 13; And

- Figure 15 illustrates an example of pseudocode for implementing a QR code improvement method.

Detailed description

Before explaining in detail a plurality? of embodiments of the invention, it should be clarified that the invention is not limited in its application to the details of construction and the configuration of the components presented in the following description or illustrated in the drawings. The invention? capable of taking other embodiments and of being implemented or practically accomplished in various ways. It is also to be understood that the phraseology and terminology are for descriptive purposes and are not to be understood as limiting. Using ?include? and ?understand? and their variations are to be understood as including the elements stated below and their equivalents, so? as well as additional elements and the equivalents of these.

In the sequel ? described a first embodiment of a process for incorporating a watermark (also called digital watermark in Italian) in a video according to the invention. This video includes a plurality? of frames arranged to be reproduced according to a predetermined reproduction order. Each frame of the plurality? of frames comprises an array of pixels. Each pixel of the matrix presents a chromatic information and a predetermined position in said matrix.

The process of incorporating a watermark includes the steps of:

a) select a frame of said plurality? of frames;

b) comparing the pixel matrix of said selected frame and a pixel matrix of a comparison frame, wherein the comparison frame ? a frame immediately following said selected frame according to said predetermined order of reproduction;

c) determining an image variation index of said frame selected in step a), by calculating the number of pixels of the pixel matrix of the comparison frame which have changed their chromatic information with respect to corresponding pixels of the pixel matrix of said selected frame, wherein said corresponding pixels are pixels having the same position in the pixel matrix of the comparison frame and in the pixel matrix of the selected frame;

d) repeat the steps a), b), and c) until when for all the frames of said plurality? of frames ? the respective image variation index was determined;

e) identify a plurality? of i-frame, through the selection of a predetermined number of frames of said plurality? of frame, which have the index of image variation pi? high;

With i-frame we mean the main frames on which the generation of all the other frames in the video decoding phase depends. Removing the iframes would consequently eliminate all the others.

In other words, the steps from a) to e) therefore concern the selection of a plurality? of i-frames that appear to be strategic, i.e. the selection of frames that contain the most information about a change of scene identifying the passage between one scene and the next, so that any cancellation of these frames would show the observer a change sudden. Video streams, in their various forms of encoding, mainly operate behind the fundamental idea that ? more It is more efficient to memorize the difference between the previous and the current frame than to repeat the information altogether. Net of this theoretical information about video coding systems, the i-frames selected by this procedure will be those that contain the greatest difference between two or more? consecutive frames.

As specified above, their selection has a strategic value for several reasons. First of all ? important to specify that the step of incorporation described later, will happen? right in the i-frame of said plurality? of iframes, so if an attacker were to discover that there is a hidden signal inside them and decide to eliminate these frames completely, the video stream would be greatly affected and an observer would see a video with sudden changes and a distorted history.

The process of embedding a watermark into a video also includes:

f) generate a cryptographic key;

In other words, a cryptographic key ? a secret that can be shared or generated with different methodologies. The purpose of the encryption key ? that of making the information present inside the i-frames accessible only to the owners (?owner?) or authorized persons. The encryption key can therefore being in turn an image (sequence of black and white pixels). Each image, in black and white, can be encoded in a sequence of zeros and ones.

The cryptographic key can be a unique parameter (with a minimum variability?) for a video or a group of videos that can? be defined on the basis of frames and the links between them. This parameter can? be the value of certain average frequency bands for each single frame (the same ones used for embedding) and its approximation to an integer value.

This process ? useful to the extent that, when you have to? extract the cryptographic key during the extraction phase of the watermark in order to extract an optical information from the iframe, there will be? an optimization in the search, as it will not be? It is necessary to try all the cryptographic keys present in a database of cryptographic keys (Key Store), otherwise it would be computationally impractical.

Pi? cryptographic keys can be used since pi? End service users are unlikely to want to share the same key. Furthermore, in the event of exfiltration (?exfiltration?) of sensitive data and the loss of the key, the security of all the videos protected up to that moment would be lost. A key management structure more? granular considerably improves this aspect.

The method of embedding a watermark in a video according to the present invention further comprises:

g) generating an optical information including copyright data of the video.

Preferably, the optical information ? a QR code and the data to ?hide? inside the i-frames they have been encoded in QR code format. This solution ? preferable given the need? to encode the content with images and the will? to preserve machine-readable information. Strings of information may be encoded within the QR code, treated for convenience, at least in this usage scenario, as structured data in JSON format. This information easily allows itself to be encrypted before embedding in the QR code to have an additional level of security (asymmetric encryption or PKI). Eventually the QR code will be? imprinted inside the i-frame through the cryptographic key and this process will bring? to the production of i-frames visually indistinguishable from the previous ones but with the previously presented content inside. There are obviously several parameters relating to the information that ? It is possible to encode within the video (quantity of characters, redundancy, correction codes and resistance to attacks/jams). The QR code processing phase corresponds to the encoding of the copyright information in the form of a QR code. The message you want to embed ? therefore encoded in the form of QR code and sar? then the QR code image to be incorporated. The choice of the QR code is explained by an innumerable series of reasons. First of all ? a standard format, well established, widely used, highly compatible on pi? platforms that offers advantages regarding its wide adoption. Secondly, the presence of the various versions of QR code guarantees a minimum of dynamism? between the amount of information to embed and that which can be embeded within a frame of a certain resolution. Third reason for using the QR code? his ability? error correction, which guarantees creating redundancy of? information encoded, and thus providing? an information loss range that stands at 30% for the version with the highest level of error correction. In other words, if the video was attacked with some technique and the QR code was defaced, the maximum tolerance of loss of information? 30%: above it, the copyright information contained within it would be lost.

Preferably, step f) generate a cryptographic key, pu? to include:

- generating a sequence of black and white pixels so as to create a certain pattern of pixels starting from a predetermined sequence of bits;

- associating the predetermined bit sequence to a key recognition index;

- storing the association of the key recognition index and of the predetermined bit sequence in an association table TA in a database.

For this reason, the basic idea? the generation of a random key and, starting from the parameter extracted in the "analysis and extraction of the i-frame" phase, create a key recognition index to be able to reach it in the phase of extracting the watermark from the video. The technique to be able to speed up access to this cryptographic key and which will be? discussed may be based on a design of a hash function that ? left to one of the techniques foreseen in the literature according to the required efficiency.

As can be seen for example in figure 5, the procedure for incorporating a watermark into a video also includes:

h) apply a discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames;

i) apply a one-level discrete cosine transform to the i-frames of said plurality? of i-frames;

l) applying a discrete wavelet transform at least at one level to the optical information including video copyright data;

m) applying a one-level discrete cosine transform to the optical information including video copyright data.

Additionally, the process of embedding a watermark into a video includes:

n) for each i-frame of said plurality? of i-frame, select a plurality? of frequency domain coefficients in a band of frequency domain coefficients as a function of cryptographic key, wherein the band of frequency domain coefficients includes frequency domain coefficients having a frequency between a first and a second frequency frequency.

For example, the first and second frequencies can be chosen according to the type of discrete cosine transform DCT, for example 8x8 blocks.

Preferably, ? Is it possible to embed the optical information including video copyright data in the 602 coefficients having an average frequency, i.e., those between the 604 coefficients in more? clear that indicate low-frequency components and the 606 coefficients in areas pi? dark that indicate those to pi? high frequency (see figure 6).

The choice compared to the mid-bands? motivated by an important consideration. The coefficients in the frequency domain of the frequencies pi? low contain most of the visual information, and human eyes are more? sensitive to changes in these frequencies relative to those in the high frequency range. On the other hand, the coefficients of the frequencies pi? high contain information about the edges of the figures and the details. Would a change at low frequencies result in a general degradation in quality? dell?image, while a change to the high frequencies would make more? vulnerable to some attacks such as data compression techniques, or low-pass filtering, or even under-sampling. The choice of medium frequencies ? therefore motivated by two fundamental parameters that characterize the watermark: perceptive transparency and robustness.

The process of embedding a watermark in a video also includes:

in each i-frame of said plurality? of iframes, embed at least a part of the optical information transformed into the frequency domain through the modification of the selected frequency domain coefficients.

In other words, steps n) and o) describe a generation step of the pseudorandom sequences, starting from a cryptographic key. The cryptographic key allows only their owners to encode and then decode the information contained within it. Furthermore, ? to note that the information itself can? be subject to encryption to ensure cos? two levels of security and the highest possible level of protection for the message contained within.

Figure 4 illustrates an explanatory diagram of an exemplary embedding of an optical information within an i-frame, through the use of the cryptographic key.

This step has the function of introducing additional information, copyright information, onto the original video, in a secure manner and minimizing the perceptual impact on the image. As described in steps h) to m) ? It is possible to apply the combination of discrete wavelet transform DWT and discrete cosine transform DCT techniques, as described later, which guarantee the addition of copyright information in a domain other than the spatial one, therefore in a domain not visible to the eye human. The embedding phase? seen as that of production of the protected frame.

Still further, the method of embedding a watermark into a video comprises: p) applying a one-level inverse discrete cosine transform to the i-frames of said plurality? of i-frames;

q) apply an inverse discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames.

Once the process of embedding a watermark in a finished video, you can then obtain a video, for example a multimedia movie with H264 encoding and mp4 file format, within which ? present the additional copyright information secured, and imperceptible to the human eye, capable of being kept inside indefinitely, and capable of resisting various possible attacks.

In figure 3A ? Illustrated is an algorithm block diagram of an exemplary embodiment of a method of embedding a watermark from a video, wherein, upon embedding of the optical information within the selected i-frames, ? It is possible to have a validation phase of the results 112, in which the correctness of the insertion of the optical information in the correct i-frames is verified, i.e. it is checked that, once the video has been modified by adding the information inside of copyright, this has the same i-frames as the original video, without which the decoding algorithm could not work.

To better specify, the selection of iframes can? be performed with automatic tools: open-source and easily accessible software libraries, which make available some specific functions. What check is performed? which, on equal terms? of selection algorithm, the i-frames before and after processing are the same.

Again with reference to figure 3A:

- block 102 generically corresponds to steps a) to d) of the process for incorporating a watermark into a video;

- block 104 generically corresponds to step e) of the process for incorporating a watermark into a video;

- block 106 generically corresponds to step f) of the process for incorporating a watermark into a video;

- block 108 generically corresponds to step g) of the process for incorporating a watermark into a video;

- block 110 generically corresponds to steps h) to q) of the process for incorporating a watermark into a video.

In figure 11 ? shows an example of pseudocode for a step of embedding a watermark.

In figure 12 ? illustrates an example of pseudocode for performing a discrete wavelet transform DWT.

Preferably, the discrete wavelet transforms applied to the i-frames of said plurality of i-frames and optical information including video copyright data are discrete 2-level wavelet transforms.

Preferably, the inverse discrete wavelet transform applied to said plurality? of iframes ? a 2-level inverse discrete wavelet transform.

Preferably, step o) in each iframe of said plurality? of i-frame, incorporate the optical information transformed in the frequency domain through the modification of the coefficients in the frequency domain selected, pu? understand the step of:

- convert each i-frame of said plurality? of grayscale i-frames.

Preferably, step g) generating an optical information including copyright data of the video, can to include:

- dividing the optical information into at least a first portion and a second portion.

Preferably, step o) in each iframe of said plurality? of i-frame, incorporate the optical information transformed in the frequency domain through the modification of the coefficients in the frequency domain selected, pu? to include:

- incorporate the first portion and the second portion in respective different i-frames of said plurality? of i-frames.

Figure 8 illustrates an example showing the concept of splitting an optical information.

The subdivision of the encoded message (the QR code) into n parts can? have a significant impact on the ability extraction of the copyright information at the time of extraction. In fact, by dividing the QR code, it is assumed that, if the video were subject to manipulation, ? It is possible that only some of these n parts are affected by disfigurement but in this way an entire QR of things would not be lost but only a portion of it. In the case of the QR code this takes on an additional value. In fact, the QR code allows you to decode the information encoded inside if the QR code has undergone deformations of less than 30%. This ? due to its ability intrinsic error correction. So going to reform the initial QR code by joining the n parts, if only some of them were not clear and usable, and these are proportionally less than 30%, then the content could still be decrypted. This improvement technique is implemented at the time of incorporation but also requires an extra step? at the time of?extraction as ? requested to reconstruct the QR code starting from the n parts. In Fig. 9, ? illustrated an example showing the concept of successive subdivision in order to obtain a result according to the level of security and the level of performance required.

The division represented here ? part of a well-known paradigm in the literature called divide-et-conquer.

In a further aspect, the process of embedding a watermark in a video can present a series of parameters or degrees of freedom? that can be modified and that have a direct impact on the properties? of the watermark (robustness, perceptive transparency). They are for example:

- the gain factor; this parameter? a multiplicative factor that serves to create greater ?distance? between the pseudo-random sequences that are incorporated into the blocks of the bands. Common application values are [0.5, 5], with extremes included;

- the Discrete Wavelet Transform (DWT) function; ? the core function that performs the DWT and there are various implementations of it. There are three possibilities: Haar, Daubechies, Dual-Tree Complex Wavelet Transform (DT-CWT);

- block size ?Block-size?; indicates the size of the block by which the sub-band is divided, and subsequently the DCT is applied to each block. Of the possible values are 4, 8;

- Steganographic key, ? the key that initializes the state of pseudorandom? of the random function; (?Also called seed?).

The invention also relates to a process for extracting a watermark from a video.

This watermark extraction process includes:

a) select a frame of said plurality? of frames;

b) comparing the pixel matrix of said selected frame and a pixel matrix of a comparison frame, wherein the comparison frame ? a frame immediately following said selected frame according to said predetermined order of reproduction;

c) determining an image variation index of said frame selected in step a), by calculating the number of pixels of the pixel matrix of the comparison frame which have changed their chromatic information with respect to corresponding pixels of the pixel matrix of said selected frame, wherein said corresponding pixels are pixels having the same position in the pixel matrix of the comparison frame and in the pixel matrix of the selected frame;

d) repeat the steps a), b) and c) until when for all the frames of said plurality? of frames ? the respective image variation index was determined;

e) identify a plurality? of i-frame, through the selection of a predetermined number of frames of said plurality? of frame, which have the index of image variation pi? high.

Once the i-frames have been determined, the process of extracting a watermark also includes:

f) identify a correct cryptographic key, among a plurality? of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video embedded in the i-frames of said plurality? of i-frames;

g) apply a discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames;

h) apply a one-level discrete cosine transform to the i-frames of said plurality? of i-frames;

i) obtain the optical information including copyright data of the video from the analysis of the coefficients in the frequency domain of each iframe of said plurality? of i-frames.

Preferably, the discrete wavelet transform applied to the i-frames of said plurality of i-frames ? a 2-level discrete wavelet transform.

In figure 13 ? illustrated an example of a first part of pseudocode for an extraction of a watermark from a frame and in figure 14 illustrated an example of a second part of pseudocode for an extraction of a watermark from a frame, which continues the first part of the figure 13.

In figure 3B ? illustrated in blocks an algorithm of an exemplary embodiment of a method of extracting a watermark from a video, wherein:

- block 202 generically corresponds to steps a) to d) of the process for extracting a watermark from a video;

- block 204 generically corresponds to step e) of the process for extracting a watermark from a video;

- block 206 generically corresponds to step f) of the process for extracting a watermark from a video;

- block 208 generically corresponds to steps g) to i) of the extraction of a watermark from a video.

Again with reference to figure 3B, following step i) obtaining the optical information including copyright data of the video from the analysis of the coefficients in the frequency domain of each frame of said set of i-frames, can? a phase 210 for processing the optical information is present. This phase ? useful for the correct functioning of the system under limit conditions such as for example in the presence of attacks. The ability to be able to process the optical information to have a certain measure of improvement guarantees an optical information more? limpid and clear, that is more? decodable, versus not having this stage. Here for the modalities? of processing of the algorithm and for the tests carried out, the types of processing can be essentially two: the improvement of the optical information, and the reunification of an optical information after its subdivision carried out during the process of incorporating a watermark in a video ( explained above for the incorporation procedure).

Figure 3B illustrates that after a possible processing of the optical information it can take place the extraction of copyright data from optical information, block 212.

In case instead of an improvement of the optical information, with reference to the figure 9, the procedure of extraction of a watermark from a video can? Also understand the step:

l) replacing predetermined areas of the obtained optical information including video copyright data with optical information structures known a priori for said predetermined areas.

The process of improving the QR code ? inserted in the extraction algorithm and allows to clarify (make clear) some 90 regions of the QR code. It is a question of clearly setting some parts of the QR code which are static and constant and which mainly serve the QR code readers to recognize the structure and initialize the decoding phase. In this way, what? as shown in Fig. 10, it improves the capacity? of recognition of the QR code why? the algorithms that decode the QR start from those regions to be sure that what they are examining is really a QR code. This technique should not be applied during the embedding phase but only during the watermark extraction phase, as it is a tool that improves the decoding of the message inside the QR code.

In figure 15 ? An example of pseudocode for the implementation of a QR code improvement procedure is illustrated.

A further consideration can? be done for the processing that is done afterwards, i.e. at the time of extraction of the QR codes. In that case, numerous versions of a piece of QR code are available, and a possible improvement considers only the best ones to reconstruct the best possible QR code. In other words, after the phase of extracting the QR code pieces from all the i-frames, a list of them is obtained. If we consider case n/2, the one used in this study, and name the sections of the QR code as BR (Bottom Right), BL (Bottom Left), TR (Top Right) and TL (Top Left), the finally, after the extraction phase we get K images of the BR section (where K is the number of i-frames), so? such as K images of section TR, TL, BL. The improvement algorithm performs permutations that allow you to find the best combination that guarantees the best extractable QR code.

Preferably, step f) identify a correct cryptographic key, among a plurality of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video incorporated in the iframes of said plurality? of i-frame, can? to include:

f1) selecting a first cryptographic key from an association table (TA) stored in a database and in which the predetermined bit sequence and a key recognition index are associated;

f2) attempt to extract the optical information including video copyright data from the iframes of said plurality? of i-frames, through the use of the first selected cryptographic key;

f3) if the recovery of the optical information in step f2) fails, discard the first cryptographic key selected in step g1) and restart from step f1) by selecting a different cryptographic key.

Figure 10 illustrates an example of an association table stored in a database.

Preferably, step f) identify a correct cryptographic key, among a plurality of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video incorporated in the iframes of said plurality? of i-frames, includes:

- obtain a key recognition index from an i-frame of said plurality? of i-frames;

- recovering a cryptographic key associated with this key recognition index from a database in which this cryptographic key and this key recognition index are stored.

Figure 7 illustrates an exemplary block algorithm in which a method of embedding a watermark into a video and the method of extracting watermarks from a video according to the invention are simultaneously applied to a video.

The benefit achieved? that of having developed procedures for incorporating a watermark into a video and for extracting a watermark from a video which are particularly robust against most of the common attacks, and which can be used to combat piracy and misappropriation of a copyrighted digital content.

Different aspects and embodiments of a method of embedding a watermark of a video and of a method of extracting a watermark from a video according to the invention have been described. It is understood that each embodiment can? be combined with any other embodiment. Furthermore, the invention is not limited to the embodiments described, but may? be varied within the scope defined by the appended claims.

Claims (10)

1. Process of embedding a watermark in a video including a plurality? of frames predisposed to be reproduced according to a predetermined order of reproduction, in which each frame of said plurality? of frame comprises an array of pixels, wherein each pixel of said array has color information and a predetermined position in said array; said procedure for incorporating a watermark comprising the steps of: a) select a frame of said plurality? of frames; b) comparing the pixel matrix of said selected frame and a pixel matrix of a comparison frame, wherein the comparison frame ? a frame immediately following said selected frame according to said predetermined order of reproduction; c) determining an image variation index of said frame selected in step a), by calculating the number of pixels of the pixel matrix of the comparison frame which have changed their chromatic information with respect to corresponding pixels of the pixel matrix of said selected frame, wherein said corresponding pixels are pixels having the same position in the pixel matrix of the comparison frame and in the pixel matrix of the selected frame; d) repeat the steps a), b), and c) until when for all the frames of said plurality? of frames ? the respective image variation index was determined; e) identify a plurality? of i-frame, through the selection of a predetermined number of frames of said plurality? of frame, which have the index of image variation pi? high; f) generate a cryptographic key; g) generating an optical information including copyright data of the video; h) apply a discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames; i) apply a one-level discrete cosine transform to the i-frames of said plurality? of i-frames; l) applying a discrete wavelet transform at least at one level to the optical information including video copyright data; m) applying a one-level discrete cosine transform to the optical information including video copyright data; n) for each i-frame of said plurality? of iframes, select a plurality? of frequency domain coefficients in a band of frequency domain coefficients as a function of cryptographic key, wherein the band of frequency domain coefficients includes frequency domain coefficients having a frequency between a first and a second frequency frequency; o) in each i-frame of said plurality? of iframes, incorporating at least a portion of the optical information transformed into the frequency domain through modification of the selected frequency domain coefficients; p) apply a one-level inverse discrete cosine transform to the i-frames of said plurality? of i-frames; q) apply an inverse discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames. 2. Process of embedding a watermark in a video according to claim 1, wherein: - the discrete wavelet transforms applied to the i-frames of said plurality? of i-frames and optical information including video copyright data are discrete 2-level wavelet transforms; And - the inverse discrete wavelet transform applied to said plurality? of i-frames ? a 2-level inverse discrete wavelet transform. 3. Process for embedding a watermark in a video according to claim 1 or 2, wherein step o) in each i-frame of said plurality? of iframe, embedding the transformed optical information in the frequency domain through the modification of the selected frequency domain coefficients, includes the step of: - convert each i-frame of said plurality? of grayscale i-frames. 4. Process of embedding a watermark in a video according to any one of the preceding claims, wherein step f) generating a cryptographic key, includes: - generating a sequence of black and white pixels so as to create a certain pattern of pixels starting from a predetermined sequence of bits; - associating the predetermined bit sequence to a key recognition index; - storing the association of the key recognition index and of the predetermined bit sequence in an association table (TA) in a database. 5. Process of embedding a watermark in a video according to any one of the preceding claims, wherein step g) generating an optical information including copyright data of the video, includes: - dividing the optical information into at least a first portion and a second portion; in which the step o) in each i-frame of said plurality? of i-frames, incorporating the transformed optical information in the frequency domain through the modification of the selected frequency domain coefficients, includes: - incorporate the first portion and the second portion in respective different i-frames of said plurality? of i-frames. 6. Process of extracting a watermark from a video including a plurality? of frames predisposed to be reproduced according to a predetermined order of reproduction, in which each frame of said plurality? of frame comprises an array of pixels, wherein each pixel of said array has color information and a predetermined position in said array; said watermark extraction process comprising the steps of: a) select a frame of said plurality? of frames; b) comparing the pixel matrix of said selected frame and a pixel matrix of a comparison frame, wherein the comparison frame ? a frame immediately following said selected frame according to said predetermined order of reproduction; c) determining an image variation index of said frame selected in step a), by calculating the number of pixels of the pixel matrix of the comparison frame which have changed their chromatic information with respect to corresponding pixels of the pixel matrix of said selected frame, wherein said corresponding pixels are pixels having the same position in the pixel matrix of the comparison frame and in the pixel matrix of the selected frame; d) repeat the steps a), b) and c) until when for all the frames of said plurality? of frames ? the respective image variation index was determined; e) identify a plurality? of i-frame, through the selection of a predetermined number of frames of said plurality? of frame, which have the index of image variation pi? high; f) identify a correct cryptographic key, among a plurality? of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video incorporated in the iframes of said plurality? of i-frames; g) apply a discrete wavelet transform at least at one level to the i-frames of said plurality? of i-frames; h) apply a one-level discrete cosine transform to the i-frames of said plurality? of i-frames; i) deriving the optical information including copyright data of the video from the analysis of the coefficients in the frequency domain of each i-frame of said plurality? of i-frames. 7. Process for extracting a watermark from a video according to claim 6, wherein the discrete wavelet transform applied to the i-frames of said plurality? of i-frames ? a two-level discrete wavelet transform. 8. Process for extracting a watermark from a video according to claim 6 or 7, wherein step f) identifying a correct cryptographic key, among a plurality of of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video incorporated in the iframes of said plurality? of i-frames, includes: f1) selecting a first cryptographic key from an association table (TA) stored in a database and in which the predetermined bit sequence and a key recognition index are associated; f2) attempt to extract the optical information including video copyright data from the i-frames of said plurality? of i-frames, through the use of the first selected cryptographic key; f3) if recovery of the optical information in step f2) fails, discard the first cryptographic key selected in step f1) and restart from step f1) by selecting a different cryptographic key. 9. Process for extracting a watermark from a video according to any one of claims 6 to 8, wherein step f) identifying a correct cryptographic key, among a plurality of of cryptographic keys, arranged to allow the extraction of an optical information including copyright data of the video incorporated in the iframes of said plurality? of i-frames, includes: - obtain a key recognition index from an i-frame of said plurality? of i-frames; - recovering a cryptographic key associated with this key recognition index from a database in which this cryptographic key and this key recognition index are stored. The process of extracting a watermark from a video according to any one of claims 6 to 9, further comprising: l) replacing predetermined areas of the obtained optical information including video copyright data with optical information structures known a priori for said predetermined areas.