METHOD FOR HIDDING DATA IN AN IMAGE OR VIDEO STREAM WITHIN A COMPRESSION CHAIN

Description

[0001] The present invention relates to a method for hiding data in an image or video stream within a compression chain.

[0002] It finds a particularly interesting application in the field of ENT medical videos and telemedicine, but it has a broader scope because it can be applied to any field in which a compression chain is used to set up blocks such as those described below.

[0003] Data concealment, also known as "data hiding," is a technique that allows information, in the form of a mark or message, to be hidden within a digital document. Different types of information can be concealed, such as text, a signature, code, an image, video, or audio.

[0004] This recent technique was initially designed in the 1990s to combat digital data piracy and protect intellectual property from identical copies, especially after the explosion in the use of digital media during transmission and storage via very open information networks that are sensitive to deliberate and / or involuntary attacks.

[0005] Nowadays, new issues have been discovered through different studies and applications.

[0006] In fact, in addition to copyright and authentication, data hiding can also be used for metadata transport, error correction or information redundancy reduction, useful for improving the performance of coding systems.

[0007] Since its inception, information hiding has been developed in several fields such as telecommunications, image / video coding, VOD (Video On Demand) services, medical imaging and computer security.

[0008] Data hiding techniques can be classified into several different categories: Covert channel: This is a communication channel established between two computers that uses the bandwidth of another channel invisibly to transmit data without the knowledge of the network administrator. This concept is very sensitive in the context of network security. Steganography: This is the concealment of one or more messages in another message invisibly without being able to detect it. It differs from cryptography, which aims to protect a message by making it incomprehensible to people prohibited from using it for security reasons such as protecting confidentiality, authenticity, and data integrity. Steganography can be used with text by hiding the message in letters (linguistic steganography) or in a sentence, or in a media medium (technical steganography) such as audio or images. It was widely used during the Second World War.Anonymization: This involves avoiding mentioning a person's name or publishing their private data to protect their privacy. Copyright marking: This involves concealing information in a document to protect copyright. Compared to the state of the art of steganography, copyright techniques can be visible or invisible. In this context, watermarking and fingerprinting are the two most well-known techniques based on this notion of data concealment.

[0009] For each application, the main constraints to be respected are: Capacity, Imperceptibility / Invisibility and Robustness.

[0010] There are different methods of hiding data in images and videos using different characteristics. For example, information can be hidden according to: the insertion domain: spatial, frequency (or spectral), multi-resolution or content-based. the brand insertion mode: additive scheme, substitutive scheme.

[0011] In the prior art, it is known to carry out the concealment in the compression chain.

[0012] Y. Liu, Z. Li, and X. Ma, "Reversible Data Hiding Scheme Based On H.264 / AVC without Distortion Drift," Journal of Software, vol. 7, April 2012, are known to describe a technique for hiding bits directly in the transformation and quantization module.

[0013] Document US6621933B2 is known, which describes a watermarking method applied to the MPEG2 compression standard. Watermarking is done in the compression chain between transformation and quantization.

[0014] Document US7006631B1 is known, which describes a method for hiding a digital tattoo by modifying the codes of the entropy encoder. The concealment is therefore directly implemented within the entropy encoder.

[0015] The present invention aims to provide a new method for concealing a maximum amount of data while preserving visual quality.

[0016] Another object of the invention is to solve the problem of low throughput in networks of limited capacity.

[0017] The invention is defined in all of the appended claims.

[0018] The invention can be applied to an image or a video stream. By block is meant a block as defined in the HEVC standard (otherwise called H.265) or a macroblock as defined in the AVC standard (otherwise called H.264) or any block generated by a DCT type transformation for example.

[0019] With the method according to the invention, data is hidden in real time in the compression chain. Applying this data to all or part of the blocks allows a compromise to be chosen between image quality and the amount of hidden data. This makes it possible to make the video or image intelligent according to the user's needs.

[0020] The invention consists of an original method for hiding as much information as possible in a video while maintaining acceptable visual quality. For this, a zonal approach was used to conceal data in frequencies to which the human eye is not very sensitive, while optimizing the compression time. This is greatly beneficial in certain applications such as low-speed transmission over a sensor network. For example, it is possible to transfer the signal of an electrocardiogram in real time without increasing the transmission rate or to carry out telemedicine everywhere in regions equipped with low-speed connections. This is a way of combating medical deserts.

[0021] The method according to the invention thus makes it possible to compress and conceal, at the same time, different types of multimedia data in a media carrier such as an image or video. The data can be concealed during compression, during a live transmission and / or storage of the video, without increasing the size / cost of data transfer.

[0022] Advantageously, the high frequency region can be located below the lower right diagonal of a square matrix of a block.

[0023] The transformation can be a DCT (Discrete Cosine Transform), DST (Discrete Sine Transform) or DWT (Discrete Wavelet Transform) using TU type blocks for “Transform Unit”.

[0024] Generally speaking, following a transformation, such as the DCT (Discrete Cosine Transform) or DST (Discrete Sine Transform), used in the latest compression standards such as H.264 and HEVC, the information is distributed in two different frequency zones: the upper left part, above the diagonal of a DCT block (the low frequencies) and the lower right part, below the diagonal (the medium and high frequencies).

[0025] Since the human eye is more sensitive to low frequencies than to high ones, the important information in the image is essentially translated by the transformed coefficients of the upper zone. The invention is thus clever in having an approach by hiding the data in the high frequencies, which makes it possible to limit the visual impact of the degradation of the image quality due to the insertion of data.

[0026] This zonal approach makes it possible to achieve significant capacities of up to 3 Mbits, for example, for an ENT endoscopic sequence of only 10s compressed to 5 Mbits / s with the HEVC standard.

[0027] Advantageously, the data may be concealed within a video stream, such data constituting a concealed stream.

[0028] The method according to the invention makes it possible, for example, to conceal a medical stream within another stream during a teletransmission session. This makes it possible to reduce the number of medical streams to be sent on a given platform.

[0029] This addresses the low-speed constraints encountered in particular in medical sector networks. Since the transmission of different streams remotely and in real time requires significant bandwidth, hiding one stream within another reduces the number of streams exchanged between different sites during a remote consultation session, for example.

[0030] According to a feature of the invention, the concealed data may constitute a digital tattoo. This is a concealed object for an application of copyright protection or confidential data.

[0031] According to an advantageous embodiment of the invention, the data concealment is only carried out on TU blocks of predetermined size. For example, the concealment can be limited to only TU blocks of size 4x4 or 8x8 or 16x16 or preferably 32x32, this depends on the nature of the images / videos processed.

[0032] According to the invention, the modification algorithm may comprise the following steps: if the bit to be hidden is equal to 1, the transformed and quantified coefficient must be non-zero, if the transformed and quantified coefficient is zero before hiding, it is modified to 1, otherwise it keeps its value, if the bit to be hidden is equal to 0, the transformed and quantified coefficient must be zero if it was different from 0 before hiding.

[0033] Despite the addition of operations due to this data hiding algorithm, it has been shown that the processing time used by the machine during compression decreases slightly compared to compression without hiding. This decrease can be explained by the increase in the number of "0" after the cancellation of significant values, following the application of the insertion algorithm. The more non-significant (zero) quantized transformed coefficients there are, the less processing the encoder performs during compression. This result confirms the simplicity and efficiency of the algorithm according to the invention.

[0034] Thus, the method according to the invention does not increase the consumption of CPU resources, regardless of the quantity of information inserted. This is compatible with real-time operation.

[0035] According to an advantageous characteristic of the invention, for a predefined concealment level k, k being between 1 and N-1 for a TU block of size NxN, the modification applies only to coefficients aligned along a line parallel to the diagonal, this line being the kth line from this diagonal.

[0036] In other words, in a TU block, it is possible to choose to apply concealment only to a part of the coefficients located below the diagonal in order to limit image distortion.

[0037] According to another advantageous characteristic of the invention, the method may comprise a step of applying a visual salience to the processed image, the modification being carried out on the most salient TU blocks or on the least salient TU blocks.

[0038] Visual saliency is a technique that allows us to define the salient objects in an image, the area of ​​concentration of the eyes during the viewing of a video, that is to say the objects having characteristics perceived as important by the visual system. This salient area, presenting relevant information located spatially in the image, can be useful to determine the ROI (“Region Of Interest”) regions in an image.

[0039] Depending on the type of video stream being processed, it may be advantageous to hide data only in the TU blocks of the salient parts of the image. This is the case, for example, in ENT endoscopic videos where the region of interest is limited. However, it may also be advantageous to hide data only in the TU blocks of the non-salient parts of the image. This is the case, for example, in videos where the interesting information is distributed over the entire image.

[0040] According to the invention, the most salient TU blocks may be those which have a saliency value greater than or equal to a predetermined threshold between 0 and 1.

[0041] To do this, we can generate a file giving a saliency coefficient for each TU block. Subsequently, it is easy to apply a saliency threshold equal to 0.4 for example (between 0 and 1). A TU block is then said to be salient if the average of all its quantified transformed coefficients is greater than 0.4. In this case, we assign it a value equal to 1. Otherwise, we assign it a 0.

[0042] According to another aspect of the invention, a data encoder is provided configured to implement the method described above. For example, this encoder can be configured to implement the H.264 or H.265 standard. Specifically, the data can be concealed according to the invention after a DCT / DST transformation and a quantization, but before a CABAC entropy coding.

[0043] The invention also relates to a data decoder configured to extract the hidden data according to the method of the invention.

[0044] Data can be extracted during video playback and can be combined to form another video stream.

[0045] Also provided is a computer program product comprising instructions which, when the program is executed by a computer, cause the latter to implement the steps of the method according to the invention.

[0046] Other advantages and characteristics of the invention will appear on examining the detailed description of a non-limiting embodiment, and the appended drawings, in which: [ Fig. 1 ] is a simplified schematic view of a compression chain or encoder according to the invention, [ Fig. 2 ] is a schematic view illustrating the method according to the invention, [ Fig. 3 ] is a schematic view of an 8x8 TU block comprising seven levels of transformed and quantized coefficients numbered from one to seven, [ Fig. 4 ] is a schematic view of three 4x4 TU block reading modes, and [ Fig. 5 ] is a view illustrating an endoscopic video sequence.

[0047] The embodiments which will be described below are in no way limiting; it will be possible in particular to implement variants of the invention comprising only a selection of characteristics described below isolated from the other characteristics described, if this selection of characteristics is sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art. This selection comprises at least one preferably functional characteristic without structural details, or with only a part of the structural details if this part alone is sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art.

[0048] In particular, all the variants and embodiments described are intended to be combined with each other in all combinations where there is no technical obstacle to this.

[0049] The various embodiments of the present invention comprise various steps. These steps can be implemented by instructions of a machine executable by means of a microprocessor for example.

[0050] Alternatively, these steps can be performed by specific integrated circuits including hard-wired logic to perform the steps, or by any combination of programmable components and custom components.

[0051] The present invention may also be provided in the form of a computer program product which may comprise a non-transitory computer storage medium containing instructions executable on a computing machine, which instructions may be used to program a computer (or other electronic device) to perform the method.

[0052] On the Figure 1We generally distinguish a compression chain 1 receiving a video sequence 2 as input. This chain operates in a closed loop insofar as the already coded data is available to process the current part. We speak of causal data which is obtained by a decoding loop integrated into the chain. The input video is processed image by image sequentially, each image being divided into slices representing all or part of an image. The slices are themselves divided into blocks in the H.265 standard or into macroblocks of size 16×16 pixels in the H.264 standard, the macroblocks can be partitioned into rectangles of smaller sizes. The blocks or macroblocks are processed according to a so-called "raster" path, starting at the top left of the image and ending at the bottom right.

[0053] According to the H.265 standard, each image undergoes a data structuring in module 3, this allows a separation between different blocks, CTU ("Coding Tree Unit") of sizes 16x16, 32x32 or 64x64. Each CTU can be divided into smaller blocks, called CU ("Coding Unit"). This defines a coding partitioning tree, called "quadtree". In fact, from the root, each node can be recursively divided into four square coding blocks CB ("Coding Block") of the same size. This difference allows for more performance at the computational level. Thus, a homogeneous area in an image can have blocks of size 64x64 and heterogeneous regions or parts of video with a lot of movement can be divided into smaller blocks, 4x4 for example, in order to bring more importance to the details of the video.

[0054] Other partitioning tree structures are nested on the leaves of the coding quadtree. These subdivisions are associated with the prediction and the residual, called the residual quadtree, before applying the frequency transformation and quantization of the residual coefficients.

[0055] Thus, each CU block can consist of several PU (Prediction Unit) and TU (Transform Unit) units depending on the prediction type (Intra or Inter) and the level of the quadtree. These units are responsible for the prediction decision mode and the application of the transform in each CU block.

[0056] Each CU block corresponds to a luminance block and 2 chrominance blocks.

[0057] The TU blocks then undergo a DCT transform and quantization within module 4. Loop 5 includes in particular a prediction step making it possible to considerably reduce the quantity of information to be transmitted, thereby improving the compression level. There are two types of pixel prediction, Intra and Inter, depending on whether spatial or temporal redundancies are exploited respectively. Intra prediction is similar to still image coding and exploits causal data within the current image and more particularly those close to the current set of pixels. Inter prediction performs motion compensation from a previously encoded image on which a motion estimation has been carried out. The nature of the blocks processed in the transformation and quantization module can thus be Intra or Inter. The present invention specifically targets Intra blocks, and in particular Intra luminance blocks.

[0058] Loop 5 therefore takes the outgoing residuals from the transformation and quantification module 4 and reinjects its result into the input of this transformation and quantification module 4 via a comparator receiving the structured images.

[0059] Thus, each of the blocks undergoes the steps of prediction generating a residual of pixels from the causal data, transformation and quantification of the residual. The residuals leaving the module 4 are coefficients arranged in TU blocks. According to the invention, the TU blocks coded in Intra of the Luminance component will undergo data concealment processing in the module 6. This data can be a digital marker stored in memory or a second video stream. The TU blocks then comprise transformed, quantified coefficients concealing data according to the invention. These TU blocks then feed directly or via a shaping module not shown an entropic coder 7 of the CABAC type (Context-adaptive binary arithmetic coding). The output is a compressed stream 8 including the second stream whose bits are concealed according to the invention.The entropy encoder usually also receives signaling information needed to decode the stream. The bitstream obtained at the output of the encoder can then be read by a decoder which renders a video by repeating the steps in reverse.

[0060] There Figure 1 a makes it possible to clearly visualize where the concealment function according to the invention is performed, which sets the invention apart from a large part of the techniques of the prior art.

[0061] On the Figure 2 we see in a little more detail the concealment process according to the invention. We distinguish an image 9 which is a luminance image coded in Intra with a resolution of 1920x1080.

[0062] The structuring makes it possible to obtain a CTU 10 block of size 64x64. This block is divided into different TU blocks of sizes 4x4, 8x8, 16x16 and 32x32. According to an exemplary embodiment of the invention, only or in addition to other blocks, the 32x32 TU block is selected on which the data concealment according to the invention will be applied. Data concealment is carried out in the transformed coefficients quantified by zonal approach from level N1 to level N31 as seen in block 11 of the Figure 2 .

[0063] We can define a level 1 of concealment allowing to hide the data in the 31 coefficient lines, parallel to the diagonal and located in the lower right corner: from N1 to N31. It is also possible to define another level 15 allowing to modify the 17 lines from N15 to N31.

[0064] On the Figure 3An 8x8 TU block is illustrated, illustrating the 7 levels of transformed and quantized coefficients numbered from 1 to 7. In this example, for an 8x8 block (N=8), we can define 7 (N-1) levels relative to the diagonal, in which we can modify the transformed and quantized coefficients. The coefficients relating to low frequencies are not represented. The diagonal is represented schematically with the reference N0. The high frequency coefficients are represented schematically with the references Ni, i ranging from 1 to 7. For reasons of simplicity and identification of the lines parallel to the diagonal, the coefficients of the same line have the same reference Ni.

[0065] For a chosen level k, the message can be hidden in all the coefficients of the Nk levels represented. Thus, at each level, a percentage of coefficients in the TU block can be modified. As shown in Table 3.1, hiding the information at level 2 for an 8x8 TU block corresponds to modifying approximately 33% of the coefficients in this block. [Tables 1] Level % of coefficients to modify 1 43,75 2 32,81 3 23,44 4 15,63 5 9,38 6 4,69 7 1,56

[0066] The second table indicates the percentage of coefficients to be modified to conceal information in the case of a 32x32 TU block. [Tables 2] Level % of coefficients to modify Level % of coefficients to modify 1 48,44 17 11,72 2 45,41 18 10,25 3 42,48 19 8,89 4 39,65 20 7,62 5 36,91 21 6,45 6 34,28 22 5,37 7 31,74 23 4,39 8 29,30 24 3,52 9 26,95 25 2,73 10 24,71 26 2,05 11 22,56 27 1,46 12 20,51 28 0,98 13 18,55 29 0,59 14 16,70 30 0,29 15 14,94 31 0,10 16 13,28

[0067] To further improve the efficiency of the method, it is planned in special cases to apply visual saliency to medical videos in order to select only a part of the TU blocks.

[0068] Visual saliency is used to define salient objects in an image, areas considered to display interesting elements. It is then possible to choose to apply concealment only to blocks that are part of the salient areas or to blocks that are part of the non-salient areas.

[0069] Salience is translated by a value ranging from 0 to 1.

[0070] For example, for certain ENT type medical images, like the image on the Figure 5 , the useful area is the central part. Thus, by applying the saliency technique, the inner circular part, which represents the medical information, is more salient than the all-black outer part.

[0071] The salient areas (inner circular part) contain more significant coefficients than the non-salient areas (black part).

[0072] It is possible to generate a file giving a saliency coefficient for each TU block. For example, if we set a saliency threshold equal to 0.4 (between 0 and 1), a TU block is said to be salient if the average of all its quantified transformed coefficients is greater than 0.4. In this case, we assign it a value equal to 1. Otherwise, we assign it a 0.

[0073] Based on this file, only salient blocks can be used to hide data. This allows for improved hiding capacity and quality.

[0074] But it may be appropriate to keep only non-salient blocks, for example for very busy images. The significant coefficients are found over the entire image and not on a specific part or area.

[0075] With the present invention, it is therefore possible to select all or part of the Intra coded TU blocks of the luminance component. The selected blocks then undergo a concealment treatment consisting of modifying all or part of the high frequency coefficients of these blocks.

[0076] Advantageously, this concealment processing is implemented by means of a concealment algorithm according to the invention so as to insert a large quantity of data in a simple manner. This algorithm can be as follows: [Tables 3] Input Transformed and quantified coefficient C(i,j) of the luminance component / Concealed bit Tb Output Coefficient after concealment C t (i,j) 1 BEGINNING 2 If Tb = 1 then 3 If C(i,j) = 0 then 4 C t (i,j) = 1 5 Otherwise C t (i,j) = C(i,j) 6 END 7 Otherwise (If Tb = 0) 8 C t (i,j) = 0 9 END 10 END

[0077] If the bit to be concealed Tb is equal to 1, the transformed and quantized coefficient C(i,j) must be non-zero. If it is zero before concealment, it is changed to 1, otherwise it keeps its value. This manipulation minimizes the distortion that can occur following concealment.

[0078] Otherwise (Tb=0), the algorithm cancels the coefficient C(i,j) if it was different from 0 before the concealment.

[0079] This algorithm can for example be integrated into an x265 real-time encoder.

[0080] Before caching data, the coefficient reading mode for each TU block can be determined so that it is taken into account during hiding. This reading mode depends on the block size and the prediction mode applied to generate the coefficients. Three reading modes are defined in the HEVC standard: Vertical reading: TU size 4x4 and 8x8 for intra predictions {6, 7, 8, 9, 10, 12, 13, 14} Horizontal reading: TU size 4x4 and 8x8 for intra predictions {22, 23, 24, 25, 26, 27, 28, 29, 30} Diagonal reading: TU size 4x4 and 8x8 for intra predictions {0, 1, 2, 3, 4, 5, 15, 16, 17, 18, 19, 20, 21, 31, 32, 33, 34, 35} and TU size 16x16 and TU size 32x32 for all intra predictions.

[0081] These three reading modes are presented on the Figure 4 in the example of a 4x4 TU block.

[0082] Knowledge of the reading mode is advantageous for the operation of the algorithm at the time of insertion and extraction of concealed messages. This makes it possible to check whether the coefficient being analyzed is in the selected area before applying concealment. This verification is done by the position and order of each coefficient in the associated TU block. For example, a coefficient 12 in an 8x8 block in vertical reading mode is not equal to the coefficient 12 in an 8x8 block that follows the horizontal reading mode. The same principle for each coefficient except the first coefficient and the last coefficient of the block.

[0083] Following the same steps as for insertion, the message extraction algorithm can be as follows: [Tables 4] Input Coefficient after concealment C t (i,j) Output Hidden Bit Tb 1 BEGINNING 2 If C t (i,j) = 0 then 3 Tb = 0 4 Otherwise (If C t (i,j) <> 0) 5 Tb = 1 6 END 7 END

[0084] Thus, if the modified coefficient C t (i,j) is zero, the hidden bit Tb is zero. Otherwise, it is equal to 1. This algorithm simplifies the processing at the decoder level, which should not cause significant additional processing time during video decoding.

[0085] For example, for medical video sequences, with an associated playback mode for x265 compression, at a bit rate of 5 Mbits / s with a standard configuration of 5 intra-coded I-frames for 10s, the results are displayed in the following two tables. Table 5 illustrates the distribution of TU blocks according to the playback mode. Table 6 illustrates the distribution of TU blocks according to their size. [Tables 5] TU block reading mode Sequence 2 (%) Diagonal 99,9 Horizontal 0,1 Vertical 0 [Tables 6] TU block size Sequence 2 (%) 4x4 0,1 8x8 0,4 16x16 4,4 32x32 95,1

[0086] These results show that the majority of TU blocks are large 32x32 (95.1%) and with diagonal reading (99.9%), for this type of medical video.

[0087] To simplify the method according to the invention, the concealment can be applied only to 32x32 TU blocks with the diagonal reading mode.

[0088] Furthermore, it was found that despite the addition of operations due to the algorithm according to the invention, the processing time used during compression decreases slightly compared to compression without concealment. This decrease can be explained by the increase in the number of "0" after the cancellation of significant values, following the application of the method according to the invention. The more non-significant (zero) quantized transformed coefficients there are, the less processing the encoder performs during compression. This result confirms the simplicity and efficiency of the concealment method according to the invention. Thus, this method does not increase the consumption of CPU resources, regardless of the amount of information inserted, this being an advantageous parameter for real-time applications.

[0089] The present invention is a new data hiding approach allowing the transport of additional metadata at constant cost. To do this, the data are inserted into the TU blocks after quantization and before CABAC entropy encoding, more precisely in the lower right zone with respect to the diagonal of the intra-coded TU blocks and associated with the luminance component. This zonal approach makes it possible to achieve significant capacities of up to 3 Mbits for an ENT endoscopic sequence of only 10s.

[0090] Furthermore, it is possible to define different insertion levels, which makes the technique configurable and scalable, thus allowing capacity to be managed according to quality.

[0091] The visual saliency technique allows to reduce the number of blocks to be processed while improving the capacity (quantity of hidden data) and the quality (visual aspect).

Claims

1. Method for concealing data in an image or a video stream inside a compression chain, this method being implemented by computer and comprising the following steps: - a phase of structuring and processing during which at least one image is structured in blocks containing coefficients, - a phase of transformation and quantization of the blocks so as to generate transformed and quantized coefficients, - an entropy encoding intended to encode the transformed and quantized coefficients, characterized in that it also comprises a data concealment step during which bits from said data are concealed by modifying transformed and quantized coefficients, situated in a zone of high frequencies of at least a part of said blocks which are relative to the luminance component of the video stream and which are intra-coded, in that this modification is carried out after the quantization but before the entropy encoding; and in that for a predefined concealment level k, k being comprised between 1 and N-1 for a block of size N x N, the modification applies only to the coefficients aligned on a row parallel to the diagonal, this row being the kth row from this diagonal.

2. Method according to claim 1, characterized in that data concealment is only carried out on blocks of predetermined size.

3. Method according to claim 1 or 2, characterized in that the modification comprises the following steps: - if the bit to be concealed is equal to 1, the transformed and quantized coefficient must be non-zero; if the transformed and quantized coefficient is zero before concealment, it is modified to 1, otherwise it retains its value, - if the bit to be concealed is equal to 0, the transformed and quantized coefficient is equal to 0.

4. Method according to any one of the preceding claims, characterized in that the zone of high-frequencies is the zone situated under the lower right part, below a square matrix of a block.

5. Method according to any one of the preceding claims, characterized in that it comprises a step of applying a visual prominence to the processed image, the modification being carried out on the most prominent blocks or on the least prominent blocks.

6. Method according to claim 5, characterized in that the most prominent blocks are those having a prominence value greater than or equal to a predetermined threshold comprised between 0 and 1.

7. Method according to any one of the preceding claims, characterized in that the data are concealed in a video stream, these data constituting a concealed stream.

8. Method according to any one of claims 1 to 6, characterized in that the concealed data constitute a watermark.

9. Method according to any one of the preceding claims, characterized in that the transform is a discrete cosine transform (DCT), discrete sine transform (DST), or discrete wavelet transform (DWT) using blocks of the transform unit (TU) type.

10. Data encoder configured to implement the method according to claim 1.

11. Encoder according to claim 10, characterized in that it is configured to implement the H.264 or H.265 standard.

12. Computer program product comprising instructions which, when the program is executed by a computer, control the latter to implement the steps of the method according to claim 1.