Adaptive and progressive system and method for the secure distribution of wavelet-coded still images

The present invention relates to the field of processing digital wavelet coded images.

It is proposed in the present invention to provide a system for visually scrambling and gradually and adaptively restoring the original content of a digital still image encoded in wavelets.

The general problem is to provide a method capable of securely transmitting digital data corresponding to high quality images in any digital format, resulting from wavelet coding, live or delayed to a display screen and / or to be recorded on the hard disk or any other backup device belonging to a box connecting the remote transmission network to a screen of the monitor or television screen type, while preserving the visual quality but avoiding any fraudulent use such as the possibility of making pirated copies of digitally coded images. Conventional encryption techniques generally consist in combining (according to operations of the addition or subtraction type) with the original data values ​​generated in a pseudo-random manner and from an initialization key. The simple possession of this key thus allows the complete decryption of the encrypted data, these containing essentially all of the original information.

In the European patent application referenced <patcit id="pcit0001" dnum="EP1011269A1"><text>EP 1011269A1</text></patcit> and entitled “System for processing an information signal”, a method for encrypting an information signal is described which can be applied to the case of still images. The method consists in adding pseudo-random noise to the original uncompressed signal so as to obtain a new signal. The signal thus encrypted is then compressed using appropriate standard algorithms and then transmitted. The key is transmitted securely to the future user of the encrypted signal. This known method can be applied in the case of images coded according to the JPEG standard.

In this document of the prior art, no reference is made to the case of wavelet-coded images. In addition, possession of the key completely conditions the decryption of the transmitted signal.

In the article <nplcit id="ncit0001" npl-type="s"><text>"An integrated approach to encrypting scalable video", Eskicioglu et al., Proceedings of the 2002 IEEE International Conference on Multimedia and Expo, Lausanne, Switzerland</text></nplcit>, the authors describe a method for generating, managing and updating encryption keys used to protect a multi-broadcast bitstream encoding a video sequence and having scalability properties (multi-layers). The flow protection system described is a system in which each layer is encrypted with a different secret key, this key being known by a group of users and being able to change periodically over time and according to the number of users . The proposed system is therefore based on conventional selective encryption technologies using one or more keys: all the data is present in the protected stream and they alone condition access to the content, therefore this prior art does not resolve the problem of high security, object of the present invention.

In the article <nplcit id="ncit0002" npl-type="s"><text>"Protecting VoD the EasierWay", Griwodz et al., Proceedings of the ACM Multimedia, September 1998</text></nplcit>, the authors describe a distribution process, via broadband networks or temporary servers and a secure point-to-point connection, of protected multimedia content whose access is controlled and traced. The initial stream encoding the original audiovisual content is deliberately corrupted by a predetermined modification of certain bytes within the stream and a signal allowing its reconstruction is only transmitted later to the client when viewing the content: a key is to first communicated to the client which will allow him to recalculate the location of the corrupted bytes within the stream. Then a signal containing the original bytes is sent to it after encryption in order to reconstruct the initial flow. The reconstruction of the flow is thus conditioned by a key and therefore the method described in this document of the prior art does not provide the high level of security proposed in the present invention.

<patcit id="pcit0002" dnum="WO2004062287A"><text>WO2004 / 062287</text></patcit> relates to a process for the secure distribution of images and is part of the state of the art as a consequence of Article 54 (3) EPC.

The present invention relates more particularly to a device capable of securely transmitting a set of fixed digital images of high visual quality to a display and / or to be recorded in the memory of the backup device of a box connecting the network. of teletransmission to the display, while preserving the visual quality but avoiding the possibility that these images may be copied illegally.

The invention relates to a method for the secure distribution of digital still images according to a nominal format resulting from wavelet coding, represented by a binary stream consisting of at least one packet (relating to the organization of the binary sequence) containing at least one block grouping simple elements (for example coefficients) coded digitally according to a mode specified inside the stream concerned and used by all decoders capable of restoring or decoding it in order to be able to display it correctly. This process includes:<ul id="ul0001" list-style="bullet" compact="compact"><li>a preparatory step consisting in modifying at least one of said simple elements,</li><li>a transmission step<ul id="ul0002" list-style="dash" compact="compact"><li>a main flow conforming to the nominal format, consisting of the blocks and packets modified during the preparatory stage and</li><li>by a separate channel from said main stream of complementary digital information making it possible to reconstruct the original stream from the calculation on the recipient equipment, as a function of said main stream and of said complementary information. Said complementary information is defined as a set consisting of data (for example coefficients describing the original digital stream or extracts from the original stream) and functions (for example, the substitution or permutation function). A function is defined as containing at least one instruction relating data and operators. Said additional information describes the operations to be carried out to restore the original flow from said modified flow.</li></ul></li></ul>

In the present invention, the concept of "stream" is defined as a structured binary sequence, consisting of simple and ordered elements representing in coded form data and meeting a given audiovisual standard or norm.

The fact of having removed part of the original data from the original stream during the generation of the modified main stream does not allow the restitution of said original stream from only the data of said modified main stream. The modified main flow is then called "secure flow". “Secure distribution” is a distribution of secure flows.

In the present invention, the term “scrambling” is understood to mean the modification of a digital bit stream according to appropriate methods so that this stream remains in conformity with the standard according to which it was generated, while making it decodable and displayable on a screen visual information coded by this flow, but altered from the point of view of human visual perception.

In the present invention, the term “descrambling” is understood to mean the process of restitution by appropriate methods of the initial bit stream, the bit stream restored after descrambling being identical, therefore without loss, to the initial bit stream.

We define the notion of “granular scalability” from the English expression “granular scalability”. We define scalability as the property that characterizes an encoder capable of encoding or a decoder capable of decoding an ordered set of bit streams so as to produce or reconstruct a so-called multi-layer sequence. We define the granularity as the quantity of information likely to be transmitted by each layer of a flow resulting from an encoding by layers, the flow being then also qualified as “granular”. We then define the qualitative, spatial and resolution scalabilities. A flow has a “qualitative scalability” if it is organized according to an ordered structure of successive sublayers whose addition allows the improvement of the visual quality of the image.

A flow has a “spatial scalability” if it is organized according to an ordered structure of data blocks coding locally spatial information in the image.

A stream has a “scalability in resolution” if it is organized according to an ordered structure of data blocks coding information making it possible to decode the image at a fixed level of resolution.

Resolution scalability is defined as the possibility of decoding the image according to several levels of resolution from a single bit stream representing the image encoded by wavelets.

A stream has a “spectral scalability” if it is organized according to an ordered structure of data blocks coding information making it possible to decode a multi-component image according to a fixed component.

The present invention proposes the protection of the digital image coded in wavelets based entirely on the structure of the “bit stream” (bitstream), protection which consists in modifying targeted parts of the bit stream (relating to modeling by wavelets) and its characteristics. Certain true values ​​are extracted from the bit stream and are stored as additional information, and, in their place, are put random or calculated values ​​or permuted values, and this for the entire digital stream. Thus, the scrambler adds “decoys” to the decoder, which receives as input a bit stream completely conforms to the original digital format, but from which the decoded and displayed image is not acceptable from the point of view of human visual perception. The scrambling module performs a bit stream analysis and selects the locations of the bit stream where it introduces disturbances. A perturbation is defined as a change (for example change of the value, inversion of the sign, saturation, thresholding), or a substitution by a random or calculated value, or a permutation. The scrambling - descrambling process carried out is without loss of quality for the original image. Advantageously, the scrambling operation is also carried out with a decoding - partial encoding of the bit stream (bitstream) representing the encoded image.

Advantageously, the present invention allows the protection of a digital image coded in a stream having a property of spatial scalability.

Advantageously, the present invention allows the protection of a digital image coded in a stream having a scalability property in resolution.

Advantageously, the present invention allows the protection of a digital image coded in a stream having a quality scalability property.

Advantageously, the present invention allows the protection of a digital image coded in a stream having a spectral scalability property.

Advantageously, the scrambling operation is also carried out with a complete prior decoding of the stream of bits representing the encoded image then a re-encoding before modification into a stream having scalability properties.

Unlike most encryption systems already known to those skilled in the art, the principle described below ensures a high level of protection while reducing the volume of information necessary for the decoding performed by progressive and adaptive.

The protection, carried out in accordance with the invention, is based on the principle of deletion and / or replacement of the information encoding the original visual signal by any method, namely: substitution, modification, permutation or displacement of the information. This protection is also based on knowledge of the structure of the stream at the output of the visual encoder: scrambling depends on the structure of said digital stream. The original stream is reconstituted on the recipient equipment from the modified main stream already present or available (for example on a CD or DVD) or sent in real time to the recipient equipment and additional information sent in real time at visualization including data and functions executed using digital routines (set of instructions).

Knowing the manner in which the wavelet encoding of the image is carried out by the wavelet coder and / or the given standard or standard, it is always possible to extract from the bit stream ( bitstream) the main parameters which describe it and which are sent to the decoder.

Many scrambling systems have an immediate effect, i.e. either the initial flow is completely scrambled or the initial flow is not at all scrambled, and the same is true for descrambling systems for visual content . With rigid systems of this type, it is difficult to satisfy the management of user rights and the quality of service of client-server multi-user, multi-application and multi-service systems, that is to say adapting services. according to the different profiles of users and their rights.

The term “profile” of the user is understood to mean a digital file comprising descriptors and information specific to the user, for example his cultural preferences and his social and cultural characteristics, his usage habits such as the frequency of use. display means, the average duration of viewing the scrambled and / or descrambled still images, the frequency of viewing a scrambled and / or descrambled sequence, or any other behavioral characteristic with regard to the use of fixed images and successions of fixed images. This profile is formalized by a digital file or a digital table usable by IT means and is located in the server and / or the client's set-top box.

The term “hardware profile” of the user is understood to mean a digital file comprising descriptors and information specific to the user's viewing equipment, for example the resolution of his viewing screen, the computing power of the still image decoder. or any other physical characteristic with regard to the use of fixed images or a succession of fixed images. This profile is formalized by a digital file or a digital table usable by IT means and is located in the server and / or the client's set-top box.

The international patent application <patcit id="pcit0003" dnum="WO9842098A"><text>WO-98/42098</text></patcit> describes a method for the secure distribution of digital still images in the form of a stream comprising data sequences each containing part of the image information.

The method described in this document comprises a step of modifying the original stream by modifying at least part of said data sequences, as well as a step of transmitting the modified stream and a step of reconstruction using a decoder on the recipient equipment, as defined by the preamble of claim 1 below.

A subsidiary step of data encryption before modification of the flow is also carried out in the method described in this document.

The reconstruction of the image is done using a digital profile established for a user, the profile being used to encrypt a decryption key of the content of the additional information, which does not allow an adaptive and progressive reconstruction of the image. The user's digital profile is therefore only used as part of a cryptographic security measure.

Requirement <patcit id="pcit0004" dnum="FR2003002915W"><text>PCT / FR2003 / 002915</text></patcit>, published under number <patcit id="pcit0005" dnum="WO2004032478A"><text>WO 2004/032478</text></patcit> belongs to the state of the art under Article 54 (3) EPC. It describes a method for the secure distribution of a video stream with adaptive and progressive descrambling. The encodings described in this application are of the DCT type, so that the images of the video stream are not encoded in wavelets.

The document <patcit id="pcit0006" dnum="WO9842098A"><text>WO 98/42098</text></patcit> describes a process for the distribution (for example of still images) in the form of a modified main stream and of additional information allowing the reconstruction.

The present invention intends to remedy the drawbacks of the prior art by proposing an adaptive and progressive descrambling system for the content viewed according to the profile and the rights of the users.

In the present invention, an adaptive and progressive descrambling of the displayed content is applied according to the profile and the rights of each user. The server sends only the parts of said additional information, which has a structure characterized by “granular scalability” to provide the user with more or less scrambled content according to certain criteria, profiles and rights. The digital wavelet encoded streams have the properties of spatial, qualitative and resolution granular scalability.

The granularity of said additional information relates to the degree of scrambling. For example, still images are completely blurred, once for all users. Then, the server sends all or part of said additional information, so that the image or the succession of fixed images appears more or less confused to the user. The content sent of said additional information and the content displayed on the user display screen are a function of each client and the server manages and sends in real time at the time of viewing by each user.

In its most general sense, the present invention relates to a method as defined by claim 1.

Advantageously, said modification produces a modified main flow and additional information allowing the reconstruction of the original flow by a decoder, the method comprising a step of transmitting the modified flow, and further comprising a step of transmitting to the equipment receiving a subset of said additional modification information, said subset being determined as a function of information coming from a hardware profile of the recipient.

In addition, said original stream is coded according to a wavelet coding method.

Advantageously, said original stream has a scalability property in resolution.

Advantageously, said original stream has a property of spatial scalability.

Advantageously, said original flow has a qualitative scalability property.

Advantageously, said original flux has a spectral scalability property.

In a variant, the modified main stream is available on the recipient equipment prior to the transmission of the additional information on the recipient equipment.

According to a variant, part of the modified main stream is available on the recipient equipment prior to the transmission of the additional information on the recipient equipment.

In another variant, the modified main stream and the additional information are transmitted together in real time.

Advantageously, the determination of said subset of said additional information is based on the scalability properties of said original stream.

Advantageously, the determination of said subset of said additional information is based on the granular scalability properties of said additional information.

In addition, the quantity of information contained in said subset corresponds to a level of scalability determined according to the profile of the recipient.

Advantageously, the type of information contained in said subset corresponds to a level of scalability determined according to the profile of the recipient.

Advantageously, said additional information comprises at least one digital routine capable of executing a function.

Advantageously, said functions transmitted to each recipient are personalized for each recipient according to the session.

According to a variant, said additional information is encrypted beforehand for each recipient as a function of the session.

According to a variant, said additional information is subdivided into at least two sub-parts.

According to one embodiment, said sub-portions of the additional information are distributed by different media.

According to another embodiment, said sub-parts of the additional information are distributed by the same media.

In a variant, all or part of the additional information is transmitted on a physical vector.

In another variant, the additional information is transmitted online.

Advantageously, the type of information contained in said subset is updated according to the behavior of said recipient during connection to the server, or according to his habits or according to data communicated by a third party.

Advantageously, the amount of information contained in said subset is updated as a function of the behavior of said recipient during connection to the server, or as a function of their habits or as a function of data communicated by a third party.

In addition, the method comprises a prior step of analog / digital conversion in a structured format, the method being applied to an analog signal.

According to a variant, a prior step transcodes the digital stream from any format to a format having scalability properties.

Advantageously, said still images constitute a succession of images fixed over time.

According to one embodiment, said modification of said data sequences is different for at least two images of said succession of images.

According to another embodiment, said modification of said data sequences of an image of said succession of images includes modification of said data sequences of previous images in the temporal order of the succession based on the properties of spatial and qualitative scalability of wavelet transformations.

Advantageously, the granular scalability of said additional information made up of said subsets is based on the qualitative, spatial and resolution scalability of the flows originating from a wavelet transformation of the images.

In addition, the process is without loss of quality.

According to a particular variant embodiment, the invention also applies to the processing of still images, for example of images compressed according to the JPEG2000 standard. In this case, it relates to a method for the secure distribution of digital still images providing that during the reconstruction of said original stream an indelible and imperceptible trace is inserted into said original stream, this trace carrying an unambiguous identifier.

According to a variant, an indelible and imperceptible trace is inserted into the image after reconstruction and decoding of said original stream, this trace carrying an unambiguous identifier.

According to an exemplary embodiment, said indelible and imperceptible trace is detectable by suitable software analyzing the reconstituted content.

Preferably, said unambiguous identifier authenticates the user.

According to a variant, said unambiguous identifier authenticates the equipment on which the algorithm for reconstructing the original stream was executed.

According to another variant, said unambiguous identifier identifies the session opened by the user during which the reconstruction of the original stream is executed.

Advantageously, the scrambling session and the descrambling session are carried out under the control of a secure server playing the role of trusted third party.

According to a particular mode of implementation, said session is identified by a secure server, making available a register comprising for each session information on the session number, the user identifier or the identifier of the user equipment, the identifier of the content subject of the session and of a date-time group.

According to another mode of implementation, a digital signature is calculated from the reconstituted stream, the inserted trace generates a unique and different signature for each reconstituted stream and this signature is stored on a secure server playing the role of trusted third party.

Preferably, the stream reconstituted by descrambling has the same visual quality as the original stream and exists in exploitable form only if it bears said trace.

Advantageously, the stream reconstituted by descrambling exists in exploitable form only if the digital signature extracted during an authenticity checking step is identical with the signature stored on the secure server playing the role of trusted third party.

Advantageously, the invention is applied to an audiovisual digital stream originating from a proprietary standard or standard.

The invention also relates to a system for the secure distribution of digital still images comprising a server comprising means for broadcasting a modified stream, and a plurality of devices provided with a descrambling circuit, the server further comprising means for recording of the digital profile of each recipient and a means of analyzing the profile of each recipient of a modified flow, said means controlling the nature of the additional information transmitted to each of said recipients.

Finally, the invention relates to a system for the secure distribution of digital still images, the level (quality, quantity, type) of the additional information being determined for each recipient as a function of the state of their profile when viewing main stream.

The invention will be better understood with the aid of the description, given below for purely explanatory purposes, of an embodiment of the invention, with reference to the appended figure:<ul id="ul0003" list-style="dash" compact="compact"><li>the figure illustrates a particular embodiment of the client-server system according to the invention.</li></ul>

Digital images are obtained using compression technologies based on wavelets (for example still images in standard JPEG-2000, MPEG-4, JJ2000, JASPER, Kakadu, Moving JPEG-2000) wavelets is an iterative scheme, that is to say the repetition of the same filtering operation at increasingly lower resolutions, and which generates flows characterized by spatial, qualitative and resolution scalability. The original flow is reconstructed on the destination equipment from the modified main flow and additional information. The additional information is divided into subsets, and depending on the profile of the user, a subset, several subsets or all of the additional information is sent for partial or total descrambling of the images.

Advantageously, the additional information sent to the user is encrypted before it is sent using a specific key for each user.

The quantity of information contained in said subset is defined as the number of data and / or functions belonging to the additional information sent to the recipient during the connection.

The type of information contained in said subset corresponds to a level of spatial, qualitative and resolution scalability determined according to the profile of the recipient. We define as "type" the nature of the data and / or functions belonging to the additional information sent to the recipient during the connection to the server. For example, the type of data relates to the habits of the recipient (full subscription, partial subscription, à la carte payment, connection time, duration of connection, regularity of connection and payments), his environment (lives in a large city, what the weather is like at the moment) and its characteristics (age, sex, religion, community).

Advantageously, the additional information consists of a succession of subsets each corresponding to a level of scalability defined in the original flow.

Said additional information is composed of at least one function, and the functions are personalized for each recipient with respect to the connection session. A session is defined from the connection time, the duration, the type of said first stream viewed and the connected elements (recipients, servers).

Said additional information is subdivided into at least two sub-parts, each of the sub-parts being able to be distributed by different media, or by the same media. For example, in the case of distribution of additional information by several media, we can ensure more complex management of the rights of recipients.

Several exemplary embodiments are described below.

The wavelet transform of an image (two-dimensional spatial signal) consists in applying to the original image a succession of high-pass and low-pass filters developed from the characteristics of the analysis wavelets. The synthesis operation, which consists in reconstructing the image from the set or a subset of the wavelet coefficients generated by the transform, obeys a reverse filtering scheme.

The application of a wavelet transform step on a digital image (which may be composed of one or more matrices of real or integer values) is equivalent to a filtering operation on the rows and columns of one or more value matrices , followed by a dyadic decrease in size (halved). It therefore generates at each step 4 new matrices of wavelet coefficients, called sub-bands and whose width and height are equal to half the width and height of the transformed matrix (dyadic progression). Let an image I of width L and height H and resolution R. The application of a step of the wavelet transform therefore generates 4 matrices of wavelet coefficients of dimension (L / 2, H / 2): the sub- LL tape_R-1, result of a horizontal (rows) and vertical (columns) low pass filtering on image I, the LH sub-band_R-1, result of a horizontal low pass and vertical high pass filter, the HL sub-band_R-1, result of a horizontal high pass and vertical low pass filtering and the HH sub-band_R-1, result of a horizontal and vertical high pass filtering.

We consider the wavelet transform at R levels (equivalent to R steps) of an image. A wavelet transform at R levels is associated with R + 1 resolution levels, numbered from R to 0, with R and 0 corresponding respectively to the finest resolution levels (initial image) and the coarsest (approximate image). Each sub-band resulting from the decomposition into wavelets of the image I is identified by its orientation (LL or HL or LH or HH) and its corresponding level of resolution (between 0 and R-1).

The original image can be considered as the LL tape_R. At each level i of the wavelet decomposition (except the last i = 0), the LL sub-band_i is thus broken down into 4 new LL sub-bands_i-1, HL_i-1, LH_i-1 and HH_i-1 and whose size is halved compared to LL_i. The process is iterated until the LL subband₀ has been obtained. Thus, for a wavelet transform at R levels, 3R + 1 sub-bands of wavelet coefficients are generated: LL₀, HL₀, LH₀, HH₀, HL₁, LH₁, HH₁, ..., HL_R-1, LH_R-1, HH_R-1.

The reconstruction of the image (synthesis) from the 3R + 1 sub-bands of coefficients consists in applying an inverse filtering operation on these wavelet coefficients followed by a dyadic increase in size. A progressive reconstruction of the image according to different levels of resolution can thus be carried out. For example, by adding in the synthesis operation to the reconstructed image of resolution r-1 the 3 sub-bands of wavelet coefficients HL_r-1, LH_r-1, HH_r-1 a new resolution image r is obtained.

The unique LL wavelet coefficient sub-band₀ is an approximation of the original LL image_R whose resolution is 2^R times lower than the original image.

The 3R HL wavelet coefficient sub-bands_r-1, LH_r-1, HH_r-1 (<i>r</i>∈[1,<i>R</i>]) correspond to details in the image, extracted at resolution level r-1. The larger r is, the more the wavelet coefficients of these sub-bands are characteristic of increasingly fine (small) details in the original image.

The wavelet coefficients resulting from the wavelet transform of an image are the spatially local characteristics of frequency information. The more r decreases, the more the spatial area characterized by a single wavelet coefficient increases (multiplication by a factor of 4 at each step).

A wavelet transform at R levels of an image generates a so-called “image” of approximation, of resolution 2^R lower and 3R "images" say details at different resolutions (0 to R).

Consequently, a bit stream offering granular scalability can be represented in the following form: {B₀, B₁, ..., B_Ntot}. Each B_i represents a subset of bits, the bit stream can then be described as a series of subsets B_i of binary symbols. Thus, a bit stream resulting from a wavelet coding has the property of "qualitative granular scalability" if and only if:<ul id="ul0004" list-style="dash" compact="compact"><li>The decoding of n (n &lt;Ntot where the bit stream is described as a sequence of Ntot subsets Bi) bits subsets B₀, B₁, ..., Bn implies a decoded image I_d quality Q_not, the quality being measured with respect to the original image I according to a predefined metric M calculated from subjective and / or objective elements, that is to say Q_not= M (I_d(or).</li><li>When m (m &lt;n) subsets B_i are decoded, {B₀, B₁, ..., B_m}, then Q_m &lt;Q_not.</li><li>When p (p&gt; n, p &lt;N) subsets B_i are decoded, {B₀, B₁, ..., B_m, ..., B_not, ..., B_p}, then Q_not &lt;Q_p.</li><li>When the Ntot subsets B_i are decoded, Q_Ntot is maximum and Q_Ntot ≥ Q_i for 0 &lt;i ≤ Ntot.</li></ul>

Likewise, a bit stream resulting from a wavelet coding has the property of “scalability in resolution” if and only if:<ul id="ul0005" list-style="dash" compact="compact"><li>Decoding of n (n &lt;Ntot) subsets of bits {B₀, B₁, ..., B_not} implies an image I_d resolution R_not,</li><li>When m (m &lt;n) subsets B_i are decoded, {B₀, B₁, ..., B_m}, then R_m &lt;R_not.</li><li>When p (p&gt; n, p &lt;Ntot) subsets B_i are decoded, {B₀, B₁, ..., B_m, ..., B_not, ..., B_p}, then R_not &lt;R_p.</li><li>When the Ntot subsets B_i are decoded, R_Ntot is maximum and R_Ntot ≥ R_i for 0 &lt;i ≤ Ntot.</li></ul>

For example, an image has been scrambled by modifying (modification of noise addition / substitution type, thresholding, permutation) a subset of N wavelet coefficients relating to one or more spectral components of the image and / or belonging to one or more regions of interest in the original image or to the whole of the image and / or relative at different resolution levels in the wavelet decomposition (from 0 to R-1) and / or belonging to one or more sub-bands (among LL, HL, LH and HH).

Adaptive and progressive descrambling of the image consists in gradually descrambling the image in several stages: first we replace n₀ (0 &lt;n₀ &lt;Ntot) wavelet coefficients modified by their original values, then n₁ (0 &lt;n₁ &lt;Ntot) and so on until n_P (0 &lt;n_P &lt;Ntot) such as: <maths id="math0001" num=""><math display="block"><msub><mi mathvariant="normal">not</mi><mn>0</mn></msub><mo>+</mo><msub><mi mathvariant="normal">not</mi><mn>1</mn></msub><mo>+</mo><mn>...</mn><mo>+</mo><msub><mi mathvariant="normal">not</mi><mi mathvariant="normal">P</mi></msub><mo>=</mo><mi>Ntot</mi><mn>.</mn></math><img file="EP1588561B2_D0001.tif" /></maths>

Depending on the scrambling method used and the client profile, descrambling is adapted to the behavior of the client when connecting to the server.

Let us describe an example of progressive descrambling. In this example, the modified wavelet Ntot coefficients belong to the sub-bands HL, LH and HH corresponding to 4 different resolution levels (ie r, r + 1, r + 2, r + 3). It consists of first replacing the n₀ wavelet coefficients belonging to the HL sub-bands_r, LH_r and HH_r, then the n₁ wavelet coefficients belonging to the HL sub-bands_{r + 1}, LH_{r + 1} and HH_{r + 1}, then the n₂ wavelet coefficients belonging to the HL sub-bands_{r + 2}, LH_{r + 2} and HH_{r + 2} and finally the n₃ coefficients belonging to the HL sub-bands_{r + 3}, LH_{r + 3} and HH_{r + 3}. The first descrambling step (replacement of the n₀ coefficients) attenuates in resolution and extent the effects of the initial scrambling (suppression of scrambling of the resolution details r) but the details belonging to higher resolution levels (r + 1, r + 2 and r + 3) are always degraded. The following steps increasingly mitigate scrambling and ultimately achieve complete descrambling. This example is purely illustrative and should not be considered as limiting. The number of resolution levels affected by the initial scrambling can be between 1 and R + 1. Depending on this number, the maximum number of descrambling steps can therefore be between 1 and R + 1.

Another variant is to send n_i coefficients belonging to the HL sub-bands_{r + i}, LH_{r + i} and HH_{r + i} in several sub-stages, thus increasing the number of stages of progressive descrambling.

Another means of progressive descrambling consists in restoring the original wavelet coefficients relating to one of the C spectral components of the image, then to two of the C components and so on until the complete restitution of the original wavelet coefficients relating to the C components. We can then speak of progressive spectral descrambling. Depending on the number of components initially scrambled (between 1 and C), the number of descrambling steps also varies between 1 and C.

Progressive descrambling of a scrambled image consists in restoring the original wavelet coefficients belonging to a predefined spatial area in the image while maintaining total scrambling of the rest of the image. Let (L, H) be the dimension of the original image and (w, h) the dimension of the area of ​​interest that we want to descramble on the original image. In this example, we assume that this area is located in the center of the image, but this area can be defined anywhere in the original image.

Let r be the level of resolution of the wavelet coefficients sub-band that must be restored. Then the following formulas indicate the intervals [is, js] and [ie, je] of the indices of the wavelet coefficients to be restored in the sub-band of resolution r considered:<maths id="math0002" num=""><math display="block"><mi>is</mi><mo>=</mo><mi>nlr</mi><mo>/</mo><mn mathvariant="normal">2</mn><mo>-</mo><mfenced separators=""><mi mathvariant="normal">l</mi><mo>/</mo><mi>js</mi><mo>=</mo><mi>ncr</mi><mo>/</mo><mn mathvariant="normal">2</mn><mo>-</mo><mfenced separators=""><mi mathvariant="normal">h</mi><mo>/</mo><msup><mn mathvariant="normal">2</mn><mrow><mi mathvariant="normal">r</mi><mo>+</mo><mn mathvariant="normal">1</mn></mrow></msup></mfenced><mo>,</mo><msup><mn mathvariant="normal">2</mn><mrow><mi mathvariant="normal">r</mi><mo>+</mo><mn mathvariant="normal">1</mn></mrow></msup></mfenced><mo>,</mo></math><img file="EP1588561B2_D0002.tif" /></maths><maths id="math0003" num=""><math display="block"><mi>ie</mi><mo>=</mo><mi>nlr</mi><mo>/</mo><mn mathvariant="normal">2</mn><mo>+</mo><mfenced separators=""><mi mathvariant="normal">l</mi><mo>/</mo><msup><mn mathvariant="normal">2</mn><mrow><mi mathvariant="normal">r</mi><mo>+</mo><mn mathvariant="normal">1</mn></mrow></msup></mfenced><mo>,</mo><mspace width="1em" /><mi>I</mi><mo>=</mo><mi>ncr</mi><mo>/</mo><mn mathvariant="normal">2</mn><mo>+</mo><mfenced separators=""><mi mathvariant="normal">h</mi><mo>/</mo><msup><mn mathvariant="normal">2</mn><mrow><mi mathvariant="normal">r</mi><mo>+</mo><mn mathvariant="normal">1</mn></mrow></msup></mfenced><mo>,</mo></math><img file="EP1588561B2_D0003.tif" /></maths>where (nlr, ncr) are respectively the number of rows and columns of the matrix of the wavelet coefficients in the sub-band considered. The progressive descrambling of the zone of interest is carried out by successively restoring the original wavelet coefficients of each sub-band for each resolution: for example LL₀, then HL₁, LH₁, HH₁, then HL₂, LH₂, HH₂, and so on until HL_R-1 LH_R-1, HH_R-1.

According to another embodiment, progressive descrambling consists in restoring first the original wavelet coefficients belonging to the LL sub-band, then the original wavelet coefficients belonging to all the LH sub-bands, then the original wavelet coefficients belonging to all the sub-bands -HL bands and finally the original wavelet coefficients belonging to all the HH sub-bands. Thus, the scrambling of the details is gradually reduced according to their orientations. Advantageously, the order of the types of sub-bands for which the wavelet coefficients are restored can be modified.

The invention will be better understood on reading an exemplary embodiment concerning a stream in JPEG-2000 format. In this example, the invention consists in modifying the value of certain fields, in particular the information necessary for a decoder for the reconstruction of the original stream.

In the appended drawing, the figure represents a particular preferred embodiment of the client-server system according to the invention.

The original stream (11) can be directly in digital form (111) or in analog form (101). In the latter case, the analog stream (11) is converted by an encoder not shown into a digital stream (111). In the following text, we will note (1) the digital input stream corresponding to the still image. The JPEG-2000 stream that we want to secure (1) is sent to an analysis and scrambling system (121) which generates a modified main stream (122) in the same JPEG-2000 format, format identical to the stream d 'input (1) outside that the values ​​of some elements of the stream have been replaced by values ​​different from the original ones, and is placed in an output buffer. The complementary information (123), of any format and organized in granular scalability layers, contains information relating to the elements of the images which have been modified, replaced, substituted or moved, as well as their value or location in the original flow and has several subsets, relating to its granular scalability property.

The stream in JPEG-2000 format (122) is transmitted, via a telecommunication network (4) of the radio type, cable, satellite, etc., to the terminal (8) of the user, and more precisely in his memory or on his hard drive (85). When the user wishes to display still images present in his terminal, the terminal (8) requests the display of the still images present in his memory or on his hard disk (85). The server (12) checks the rights of this user for this request. For this, the server can use the data of this user contained in a database linked to the server (12) and / or use a smart card-based system (82) linked to the synthesis system (87). are then possible.

If the user does not have all the rights necessary to view the image, in this case, the JPEG-2000 stream (122) generated by the scrambling system (121) present in the memory or on the hard disk (85 ) is sent to the synthesis system (87), via a read buffer memory (83). The synthesis system (87) does not modify it and transmits it identically to a conventional JPEG-2000 player (81) and its content, visually degraded by the scrambling system (121), is displayed on the screen display (6). The user of the terminal (8) therefore sees a scrambled image.

Alternatively, the user has the rights to view the image. In this case, the synthesis system sends a viewing request to the server (12) containing the information necessary (123) for recovering the original image (101). The server (12) then sends via telecommunication networks such as analog or digital telephone line, DSL (Digital Subscriber Line) or BLR (Local Radio Loop), via DAB (Digital Audio Broadcasting) networks or via digital mobile telecommunications networks. (GSM, GPRS, UMTS) (5) at least one subset of the additional information (123) allowing the reconstruction of the image at the terminal (8), which stores said subset in a buffer memory (86) . The synthesis system (87) then restores, in the scrambled JPEG-2000 stream which it reads in the read buffer memory (83), the modified fields of which it knows the positions as well as the original values ​​thanks the content of the additional information read from the image descrambling buffer (86). The amount of information contained in the additional information (123) and which is sent to the descrambling system is specific, adaptive and progressive for each user and depends on his rights, for example single or multiple use, right to make one or more private copies, delay or advance payment. To determine the quantity of information of the additional information (123) to be sent to the terminal (8), the server (12) consults the rights of the user beforehand.

According to one embodiment, a progressive descrambling of an image for which the Ntot modified wavelet coefficients belong to the sub-bands HL, LH and HH corresponding to 4 different levels of resolution (i.e. r, r + 1 , r + 2, r + 3) consists in first replacing the n₀ wavelet coefficients belonging to the HL sub-bands_r, LH_r and HH_r, then the n₁ wavelet coefficients belonging to the HL sub-bands_{r + 1}, LH_{r + 1} and HH_{r + 1}, then the n₂ wavelet coefficients belonging to the HL sub-bands_{r + 2}, LH_{r + 2} and HH_{r + 2} and finally the n₃ coefficients belonging to the HL sub-bands_{r + 3}, LH_{r + 3} and HH_{r + 3}. The first descrambling step (replacement of the n₀ coefficients) attenuates in resolution and extent the effects of the initial scrambling (suppression of scrambling of the resolution details r) but the details belonging to higher resolution levels (r + 1, r + 2 and r + 3) are always degraded. The following steps increasingly mitigate scrambling and ultimately achieve complete descrambling. Depending on the number of resolution levels R chosen to scramble the flow, the number of resolution levels affected by the initial scrambling is between 1 and R + 1. Depending on this number, the number of descrambling steps is therefore between 1 and R + 1. In this exemplary embodiment, the sending of a subset of the additional information containing the n₀ coefficients are performed when the user connects, selects and downloads the image they want. The selected image is then displayed on its partially descrambled screen, because it is calculated only from the n₀ coefficients transmitted to the user's terminal. If the user decides to view the image at a higher resolution, the server offers the user to pay a predetermined amount. If the user pays immediately by a conventional means of remote payment (bank card, etc.), the server sends a second subset of the additional information containing the n₁ coefficients. During the payment transaction, the subsets of the additional information concerning the following descrambling stages are sent and more and more attenuate the scrambling to finally achieve complete descrambling, the displayed image being identical to the original image . If the customer does not agree to pay immediately, the coefficients will be sent gradually according to the arrival of the payment. At each transaction, the server records the behavior of the user and updates the profile of the latter in a database according to said behavior.

The content sent of said additional information (123) and the content displayed on the client viewing screen are a function of each client and the server manages and sends in real time the sending of said subsets at the time of viewing for each user, for example based on the price the customer is willing to pay. Consider that we have still images stored on the server with 10 different resolutions from R = 1 to R = 10, R = 10 being the maximum resolution. If a customer is used to ordering images of medium resolution, his subscription corresponds to descrambling obtained with R = 5. If he wishes to obtain a higher resolution, therefore for example descrambling for R = 7, he must change the type of payment or subscription. He can then, if he wishes, and for example by a new payment, obtain the resolution R = 8, then R = 9 and finally R = 10. All these operations are managed by the server (12) according to the behavior of each user and through the use of a database linked to the server (12).

Similarly, another client needing high resolution images takes the subscription corresponding to the maximum resolution for R = 10. If said client has a delay in payment, the server automatically sends them descrambled images for R = 6 by example, to remind him to regularize his payment.

As we have just described, the level (quality, quantity, type) of the additional information is determined according to each recipient, according to the state of his profile at the time of the transmission of the main stream and a part to the less of said profile is stored on recipient equipment. For example, in the appended drawing, part of the user profile is recorded on the smart card (82) linked to the synthesis system (87), such as for example the digital data concerning the frequency of connections or the regularity payments. This same data and / or the rest of the profile may or may be located on the server (12). The rest of the profile can contain, for example, the type of images the user prefers.

In an alternative embodiment, the recipient's profile is updated. The update also depends on the connection time to the server (behavioral data), namely whether the client connects regularly (referring to his habits). Similarly, the recipient's profile can be updated according to data retrieved from a consumer database already existing on a server and relating to this client.

According to another exemplary embodiment, the server transmits all or part of the additional information to the user for a few seconds of the display of the image, then, over time, transmits fewer and fewer subsets additional information. Thus, the descrambling of the image is less and less complete, thus giving the effect to the user that the image displayed on his screen becomes less and less understandable, therefore more and more scrambled. This feature prompts the user to purchase the rights to see the image fully descrambled, since they have partially seen the content.

According to another embodiment, all or part of the additional information (123) is transmitted to the user on a physical vector such as a memory card or a smart card (82).

According to another embodiment, only part of the modified main stream is available on the recipient equipment: if the characteristics of the display screen (6) only allow a limited number of resolutions to be displayed (R = 1 at R = 5), the user only needs to recover part of the modified main stream (122). Depending on the profile and the rights of the user, part of the additional information enabling him to view the image only at resolutions R = 1 to R = 5 will be sent by the server (12) to the user terminal (8) .

The example described below represents another preferred embodiment of progressive and adaptive descrambling for digital images from the JPEG-2000 standard.

The analysis and scrambling module (121) generates the additional information (123) and sends it to the destination equipment via the network (5). Advantageously, the network (5) comprises a secure server on which the additional information (123) is stored. The module (121) also generates the modified main stream (122) in the JPEG-2000 format which is transmitted on the hard disk (85) of the client's destination equipment by the network (4). Advantageously, the network (4) comprises a multimedia server on which the modified main stream (122) is stored. The scrambling module (121) also inserts into the metadata of the modified main stream the identifier of the additional information corresponding to said modified main stream and the physical address of the secure server (5) on which said additional information is stored.

The additional information is characterized by the presence of said subsets corresponding to several scalability layers, for example four in number. The first subset contains all the additional information relating to the high quality of the original image. The second subset contains part of the additional information relating to acceptable quality. The third subset contains part of the additional information, relating to a low quality, the image still visible, but unusable. The fourth subset contains just the part of the additional information corresponding to a minimum quality, the image is barely visible.

During the reconstitution of the original stream, the descrambling module (87) inserts into the reconstructed stream an indelible and imperceptible trace by the human eye, this trace carrying an unambiguous identifier. The trace inserted in the stream is detectable by suitable software which has the capacity to analyze the reconstituted content. The insertion of said trace and descrambling takes place sequentially and gradually, so that a flow which has been scrambled by the analysis and scrambling system (121) and then descrambled by the module ( 8), exists in exploitable form only if it includes said unambiguous trace. During descrambling and sequential insertion of the trace, the reconstruction is done in such a way that the trace is gradually inserted during the progressive descrambling. At the end of descrambling, protection is ensured by the trace which is substituted for the protection obtained by scrambling.

Advantageously, an indelible and imperceptible trace by the human eye is inserted into the image and after decoding of the reconstituted stream.

Advantageously, a stream protected with the scrambling module (12) and descrambled with this variant of the descrambling module (87) always carries protection: either invisible, resulting from the insertion of the identifying trace, or visible, from adaptive and progressive descrambling.

Advantageously, said unambiguous identifier carried by the trace relates to the identification of the interactive session opened by the user. The role of said unambiguous identifier is the authentication of the user, of the recipient equipment on which the image is reconstructed and / or of the session opened by the user.

According to one embodiment, during the descrambling step, the flow is reconstituted and an indelible and imperceptible trace is inserted. A digital signature of the reconstituted stream is then calculated, this signature being unique and different for each reconstituted stream thanks to the insertion of the trace, and this signature is stored on a secure server playing the role of trusted third party.

Advantageously, the reconstituted stream exists in exploitable form only if the signature extracted during an authenticity checking step is identical to the signature stored on the secure server during the reconstitution.

The session is characterized by a number assigned by the secure server (5), which acts as a trusted third party between the user and the adaptation of the parameters characterizing the type of descrambling. The secure server (5) assigns a specific number to each session, which is saved in a register. The information included in the register is the session number, constructed from the user identifier or the identifier of his equipment (8), the identifier of the content of the image which is the subject of the session. and an ISO standard date-time group.

The identification of the content of the image is carried out by the descrambling module (87) which recovers from the metadata of the main stream modified the identity of the additional information relating to the main stream modified and the physical address of the network. (5) where the secure server is located on which said additional information is stored.

The reconstruction of the original image is carried out in several stages. When establishing the session, the descrambling module (87) reads in the metadata of the modified main stream, the identifier of the additional information and the URL of the secure server (5). The server (5) firstly sends said fourth subset of the additional information which produces an image of minimum quality, the image being unusable and poorly visible. This step serves to confirm the identity of the secure server (5). The second step consists in sending to the descrambling module said third subset of additional information which makes the image visible, but still unusable. This step is necessary for the client to decide if he wants to obtain rights to exploit the image, by seeing a preview of its content. Depending on the customer's wish (8) to obtain an acceptable or maximum image quality, said second subset (corresponding to an acceptable quality) or said first subset (corresponding to maximum quality) is sent to it. additional information, subject to payment corresponding to the quality required. Regardless of the subset used for descrambling, an unambiguous trace is always present in the reconstituted flow. Said trace is intended for the protection of intellectual property, in accordance with the WIPO (World Intellectual Property Organization) treaty of December 1996 stipulating that the data intended for the protection of intellectual property may be digital data.

The embodiments described above have the value of examples and do not constitute a limitation of the present invention.