FR2792153A1 - Determination method for a partition in order to insert a watermark into a digital image - Google Patents

Abstract

The method involves modulating coefficients (X) of the regions (R) representing a digital image (I). The partitioning into regions is effected by an adaptive partitioning as a function of a criterion of detectability of an information bit inserted on each region. The method involves generation of a centered pseudo-random sequence equal in size to the cardinal number (N) of the set of coefficients, formed from centered pseudo-random sub-sequences. The set of coefficients are modulated by the centered pseudo-random sequence in order to insert the same information bit on the set of coefficients. Finally the method involves checking, over each region, of a criterion of detectability of the information bit inserted by modulation. Independent claims are included for a method of decoding a watermarking signal in a digital image, for a partition determination device and for a decoding device.

Description

The present invention relates to a method for determining a partition.

  a digital image to insert a marking signal and a

  method of decoding an associated marking signal.

  It correlatively relates to a device for determining a partition of a digital image for inserting a marking signal and a

  device for decoding a marking signal.

  The present invention relates generally to the technical field of marking (in English watermarking) or tattooing of images.

  digital, and more particularly still images.

  Marking digital data helps protect that data

  for example by associating copyright information with them.

  In its general principle, marking consists in inserting an indelible mark in digital data, assimilated to the encoding of a

  additional information in the data.

  The decoding of this additional information makes it possible to verify

  copyright information that has been inserted.

  This inserted mark must therefore be both imperceptible, robust to certain distortions applied to the digital image and reliable detection. Conventionally, a usual technique for inserting a marking signal into a digital image consists in using a linear modulation model in which at least a subset of coefficients representative of the digital image is modulated according to this linear model. in

  using a weighting coefficient.

  By noting X = {Xi, 1 <i <N} a set of coefficients representative of a digital image and w = {wj, 1 <j <P} a marking signal of size P <N, the insertion formula linear is: X'j = Xj (i) + aj wj in which cj is a weighting coefficient, also called

modulation amplitude.

  The detection of the marking signal then consists in detecting whether one has

  whether or not to insert a modulation sequence into a set of coefficients.

  This detection is made without using the original marked image and is based on a correlation calculation or on a standardized statistical test which allows

  theoretically calculate a correct probability of detection.

  Such an insertion technique allows, by the insertion of a marking signal, to insert a single bit of information since the response of the detector

is binary.

  To insert a greater number of bits of information into the digital image, in particular when a C-bit code indicating for example the name or address of the owner or author of the image is desired, it it is necessary to repeat the insertion process described above as many times

  only bits of information to insert.

  In other words, one must choose C subsets of coefficients and operate the modulation of these subsets by choosing C marking signals. Preferably, subsets of distinct coefficients are chosen so that the modulations are not superimposed on each other,

  which could disturb detection or cause annoying visual effects.

  It is therefore a question of choosing a partition of the coefficients representative of the digital image into C distinct subsets, carriers

each of a bit of information.

  Many known methods use a technique of inserting a given marking signal by spread spectrum. The disadvantage of these methods, described for example in the article entitled "Secure spread spectrum watermarking for multimedia" by l.J. COX et al, in Proc. ICIP, pages 243-246, Sept. 1996 and in the article entitled "Digital watermarking of raw and compressed video" by F. HARTUNG et ai, in Proc. SPIE 2952: Digital compression technologies and systems for video Communication, pages 205-213, October 1996, is that they use an arbitrary partitioning of the image into blocks of fixed size, without guarantee of the

  detectability of the modulations carried out on each of the blocks.

  The object of the present invention is to propose a method for determining a partition of a digital image which is adapted to the image, ensuring the detectability of the bits inserted for an insertion technique by modulation. The present invention thus relates to a method of determining a partition into distinct regions of a digital image for inserting a marking signal by modulation of coefficients of said regions representative of a

digital image.

  According to the invention, this partition into regions is carried out by an adaptive partitioning as a function of a bit detectability criterion

  of information inserted on each region.

  An adaptive partitioning is thus carried out which has the advantage of adapting to the content of the image signal and of guaranteeing the detectability of the marking signal inserted, unlike the choice of an arbitrary partitioning in

blocks fixed.

  This spatial adaptability is necessary because the content of the images

is not spatially homogeneous.

  Indeed, for a given modulation amplitude, the greater the length of the sequence of the inserted marking signal, and consequently the greater the size of the region and therefore the number of modulated coefficients,

the more reliable the detection.

  We show in fact that when a correlation calculation is used to operate the detection, the response of the detector is proportional to p1 / 2 o P is

  the length of the marking signal sequence.

  Furthermore, the response of the detector is also dependent on the image signal itself. The greater the variance of the representative coefficients of

  the larger the image, the lower the detector response.

  In conclusion, for a given modulation amplitude and a given image, there is a minimum length of the marking signal necessary to ensure a given probability of detection, this minimum length further depending on the local content of the image signal in the region in

  which is inserted the marking signal.

  The adaptive partitioning carried out by the method according to the invention makes it possible to guarantee the detectability of the marking signal inserted on

  the different regions determined by this partition.

  Preferably, the detectability criterion is calculated by using a maximum value of the weighting coefficient ensuring the imperceptibility of the marking signal for the modulation of all the representative coefficients

of said digital image.

  Indeed, the greater the modulation amplitude, that is to say the value of the weighting coefficient, and the more reliable the detection in the sense that a better probability of detection can be obtained. However, the greater this amplitude of modulation, the more this modulation risks being

visible in the marked image.

  It is therefore judicious to choose a modulation amplitude equal to a maximum amplitude, generally noted JND (in English Just Noticeable Difference), beyond which the modulation becomes visibly perceptible.

in the picture.

  It can be shown that this maximum amplitude also depends on the length of the sequence of the marking signal and that it decreases with this length. By choosing a maximum value of the weighting coefficient ensuring the imperceptibility of the marking signal for the modulation of all the coefficients representative of said digital image, it is guaranteed that the value of the weighting coefficient on each region of the partition, of size less than of the coefficients representing the image, will be much less than the maximum amplitude of

  modulation which guarantees the imperceptibility of the modulation on this region.

  The partitioning thus carried out also ensures invisibility

  of the modulation performed on each of the regions obtained by the partition.

  According to an advantageous characteristic of the invention, the detectability criterion is calculated by using, for each coefficient to be modulated, a weighting coefficient according to a law of the form ai = ki. ov, o ki is a modulation factor depending on the coefficients located in the vicinity of the coefficient considered on said region and av is equal to said maximum value

of the weighting coefficient.

  One can thus exploit the possibilities of visual "masking" of the modulation by the image signal itself. Typically, the fact that the higher the activity of the image signal is used, for example in the sense of a variance

  local coefficients important, and less the modulation will be visible.

  It is thus possible to locally increase the weighting coefficient and therefore the modulation amplitude, for the benefit of detectability, without harming

  the imperceptibility of the inserted modulation.

  According to another preferred characteristic of the invention, the method for determining a partition comprises a preliminary step of comparing the size of each region of the partition with a minimum size corresponding to the minimum size of a statistically significant sample for the detection

  of a marking signal inserted in said region.

  Indeed, due to the statistical calculations made during the detection of the marking signal, in particular when using standardized hypothesis tests, a minimum bound must be considered for the size of

  each region of the partition which guarantees the validity of the detection tests.

  According to a particularly practical embodiment of the invention, for each region of the partition, the set of coefficients of said region is modulated by inserting the marking signal, a detectability amplitude is calculated from said detectability criterion and one validates said detectability criterion by comparing said detectability amplitude with a value of

predetermined threshold.

  The amplitude of detectability can be the result of a correlation calculation between the set of modulated coefficients and the marking signal or the result of a standardized statistical test such as that described in the article entitled "A method for signature casting on digital images "by I. PITAS, in Proc. ICIP, pages 215-218, September 1996. According to a preferred characteristic of the invention, a partition is made by iterative division of the digital image and, for each region at one rank of said partition, a partition of a directly higher rank of said region if and only if there are at least two sub-regions of said region

  region for which said detectability criterion is validated.

  Thus, when an attempt is made to divide a region into sub-regions, this partition of a higher rank is only validated if the number of regions in which a marking signal can be effectively increased

  actually inserted respecting the detectability criterion.

  Otherwise, it is preferable to keep the region before division, of size

  higher, in order to increase the detectability of the marking signal.

  According to a preferred embodiment of the invention, for each region in one row of said partition, the value of a capacity of the digital image is incremented when the partition of directly higher rank is not validated, said capacity being equal to the cardinal of distinct regions obtained by

  said partition on which the detectability criterion is validated.

  It is thus possible to calculate the capacity of the digital image which thus corresponds to the number of bits of information that can be inserted into the digital image. Thanks to the adaptive partitioning carried out by the invention, by reiterating the partitioning of each region as long as the detectability criterion is respected, the maximum capacity of the digital image is obtained which corresponds to the watermarking capacity of this image by the technique d insertion by modulation. This watermarking capacity can be defined as the maximum number of bits of information that can be inserted into the image so

  imperceptible with a guarantee of detectability at the decoding level.

  According to another embodiment of the invention, for each region at one rank of said partition, the value of a capacity is incremented when the partition of directly higher rank is not validated, said value of capacity is compared to a fixed capacity value and the partition is reiterated for a region having the highest detectability amplitude among all the other regions to be treated when said capacity value is lower

  at said fixed capacity value.

  This partitioning method makes it possible in this embodiment to find the best possible partition in terms of detectability when a number of predetermined information bits must be inserted, equal to said value.

of fixed capacity.

  By selecting during the partition the regions for which the detection amplitude is the highest, the present invention makes it possible to determine the partition guaranteeing a priori the best detectability at the decoding of the marking signals making it possible to insert a number

  predetermined information bits.

  According to another embodiment of the invention, a partition is performed by iterative fusion of the digital image and, for each region of said partition for which said detectability criterion is not validated, one

  merges said region with another region of said partition.

  Preferably, in this embodiment, said other region of the partition is if possible a region for which said detectability criterion

is not valid.

  It is thus possible to increase the capacity of the digital image by bringing together regions in which the detectability criterion is not verified. The present invention also relates to a method of decoding a marking signal in a digital image, inserted by modulation of coefficients representative of said image in distinct regions forming

  a partition of said digital image.

  According to the invention, this decoding method includes a step of determining a partition into distinct regions of the digital image to be decoded by performing an adaptive partitioning as a function of a detectability criterion of an information bit obtained by demodulation of

coefficients on each region.

  Thus, this decoding method makes it possible to find at the level of the decoder, the adaptive partition produced for the insertion of a marking signal. In fact, when the partition has been performed in an adaptive manner as described above by a method for determining a partition in accordance with the first aspect of the invention, this partition which depends on the

  image content is not known to the decoder.

  The decoding method according to the invention then consists in testing all of the admissible partitions and in checking the detection of a signal of

  marking on each of the regions associated with this partition.

  According to a preferred characteristic of the invention, the decoding method comprises a preliminary step of comparing the size of each region of the partition with a minimum size corresponding to the minimum size of a statistically significant sample for the detection of a signal of

  marking inserted in said region.

  According to a practical embodiment of the decoding method according to the invention, for each region of the partition, a detection amplitude is calculated from said detectability criterion and said detectability criterion is validated by comparing said detection amplitude to a value of threshold

predetermined for decoding.

  The decoding process thus makes it possible to find the partition

  adaptive performed during the insertion of the marking signal.

  According to an advantageous characteristic of this embodiment of the invention, the predetermined threshold value for decoding is less than said predetermined threshold value used during the determination process

of a partition.

  Indeed, during decoding, the detection test is carried out on coefficients having been a priori modulated and possibly noisy. These coefficients are therefore different from the original coefficients, and the result of the decoding detectability test will be different from that carried out to determine the partition. By choosing different thresholds for decoding and coding, one can take into account the trend of variation of the detection amplitude due to an increased variance of the coefficients which have been modulated and possibly noisy. Correlatively, the present invention also relates to a device for determining a partition of a digital image and a decoding device adapted respectively to implement the method for determining a partition and the decoding method in accordance with

the invention.

  The present invention also relates to a computer, a digital image processing apparatus, a digital printer, a digital photographic camera and a digital camera adapted to implement the partition determination method and the decoding method in accordance with the invention and / or comprising a device for determining a

  partition or a decoding device according to the invention.

  These partition determination and decoding devices, this computer, this digital image processing device, this digital printer, this digital photographic camera and this digital camera have characteristics and advantages similar to those described with reference to the methods. determining a partition and decoding that they

implement.

  The present invention also relates to an information storage means, readable by a computer or by a microprocessor, integrated or not in a partition determination or decoding device, possibly removable, which stores a program implementing the method determining a partition and / or decoding in accordance with the invention. Other features and advantages of the invention will become apparent

  in the description below of an embodiment of the invention.

  In the accompanying drawings, given by way of nonlimiting examples - FIG. 1 is an algorithm of the method for determining a partition in accordance with a first embodiment of the invention; - Figure 2 is an example of partitioning a region by division; - Figure 3 is an algorithm of the partition determination method according to a second embodiment of the invention; - Figure 4 is an algorithm of the decoding method according to a first embodiment of the invention; FIG. 5 is an algorithm illustrating the application of the method for determining a partition from FIG. 1 to a digital image; - Figure 6 schematically illustrates a spectral decomposition step implemented in the algorithm of Figure 5; FIG. 7 is an algorithm illustrating the application of the decoding method of FIG. 4 to a digital image; - Figure 8 is a block diagram illustrating an alternative embodiment of the partition determination method according to a third embodiment of the invention; and - Figure 9 is a block diagram illustrating a device adapted to implement the method of determining a partition and / or decoding

according to the invention.

  We will first describe, with reference to FIG. 1, a method of determining a partition into distinct regions of a digital image which makes it possible to maximize the number of regions obtained in order to maximize the

  number of information bits inserted in the image.

  We consider a representation of the digital image by a set of coefficients, either in the spatial domain or in a transformed domain, the coefficients being in the latter case hybrid, that is to say located both in the spatial domain and in the frequency domain. A

  such representation of the image is for example obtained by using a sub-

  band resulting from a spatio-frequency decomposition of the image, for example

  a discrete wavelet decomposition.

  The image is partitioned to insert a marking signal

  by modulating the coefficients of each region.

  According to the invention, this partition into regions is carried out by an adaptive partitioning according to a criterion of detectability of a bit

  of information inserted on each region.

  The number of regions obtained under constraint of the detectability criterion makes it possible to define a maximum capacity of the digital image. This maximum capacity corresponds to the number of bits of information which it is possible to insert into the digital image while respecting the criterion of detectability. Initially, at a step E100, this capacitance C is initialized at 0 and we start from an initial partition which is limited here to a single region R of size P. Of course, the initial partition could be different and in particular already include several regions distinct of possibly different sizes. It is assumed that the size P of this initial region R is greater than a minimum size denoted Pstat which corresponds to the minimum size of a sample of statistically significant coefficients for the detection of a marking signal inserted in these coefficients. Typically, this size

minimum may be set to 100.

  This initial region R can be constituted by the set of coefficients of the representation considered or a subset of this representation. A test step E101 then makes it possible to test the detectability of the insertion of a marking signal on the initial region R. This detectability test can be carried out by effective modulation and detection operations of the coefficients of the initial region A. In practice, we copy all the coefficients of the initial region R

  in a working memory so as not to mark the image directly.

  We then modulate all the coefficients of the initial region R into

inserting the marking signal.

  Conventionally, we proceed as follows to perform this modulation. The initial region R consists of a set of coefficients: X: {Xi, i = 1, ... P} We commonly use a linear modulation of the type: Xi * = Xi - XCiWi, with i = 1, ..., P o W: {Wi, i = 1, ..., P} is the marking signal generally chosen as a pseudo-random sequence of known distribution and zero mean. The value of the + sign of the modulation depends on the binary value to be inserted: for example, the sign corresponds to the value 0 and the + sign

corresponds to the value 1.

  One can choose as an example a pseudo-random sequence W which

  follows a uniform law on the interval [-1, 1].

  Of course, any pseudo-random marking signal, of known distribution and of zero mean may be suitable. The most common distributions for the marking signal W are, in addition to the aforementioned uniform distribution over [-1, 1], the binary distribution t-1,1 and the normalized Gaussian distribution.

centered N (0, 1).

  The modulation can optionally be secured using a secret key characterizing for example the seed enabling the pseudo-random sequence to be reproduced W. The term ai is the modulation amplitude, or weighting coefficient,

applied to the coefficient Xi.

  A constant weighting coefficient can be used for all the coefficients of the region R, such that ai = av for all i, the value of the weighting coefficient av being equal to a maximum value ensuring the imperceptibility of the marking signal for the modulation of l set of coefficients

  representative of said digital image.

  This maximum value av (called in English Just Noticeable Difference or amplitude JND) decreases in fact with the length of the sequence of the marking signal W, and consequently with the number of coefficients

modular.

  Since the capacitance C will be maximized when all the coefficients of the representation of the digital image are used for modulation, it is advisable to determine a JND amplitude of modulation for a number of modular coefficients equal to the cardinal of the set of

  coefficients representative of the image.

  This maximum value cav or amplitude JND corresponds to the maximum modulation amplitude usable in the above-mentioned linear insertion model beyond which an observer is able to visually detect a

  change in the marked reconstructed image.

  To determine this JND amplitude, it is possible to use a visibility model which makes it possible to predict the visibility of a marking operation as a function of different parameters such as the representation of the signal through a spatio-frequency transformation S used, the sub-band. considered for insertion, the type of distribution of the W signal sequence and the length P of the W sequence. A simple model, developed by WATSON and described in the article entitled "Visibility of wavelet quantization noise", AB WATSON et ai, IEEE Trans. on Image Process, 6 (8), 1164-1175, 1997, makes it possible to predict the visibility of a set of modulated coefficients from the visibility measurement of a single modulated coefficient. We can advantageously refer to this document

  for the detailed description of this model.

  A function is thus used which depends on the length P of the marking signal W, on the type of transformation S used and on the sub-band (base) considered for the insertion, but independent of all the coefficients to be marked X. This calculation function used can be written a, (P, S, w) = ab., (S) p Yp (E [JI "]) // 3 O 0base (S) is a basic value, dependent on the transformation used S and of the sub-band (base) considered for the insertion, of the maximum weighting coefficient ensuring imperceptibility during a modulation of a single coefficient of this sub-band, 3 is strictly greater than 2, and

  E [lw p] is the mathematical expectation of the function wll.

  The basic values Cbase (T) can be measured once and for all, for each sub-band of coefficients in the wavelet decomposition, from a single psychovisual measurement, and can be stored in a table of visual amplitudes. 13 is the exponent of a Minkowsky sum and can be chosen equal to 5

for example.

  The mathematical expectation E [JlQ] corresponds to an estimate of the

mean of the function lk.

  This visibility model does not take into account the image I itself to mark and is independent of it. This is equivalent to considering that image I is uniform. This is a "worst case" model since the presence of the signal

  allows you to visually hide the modulation itself.

  The amplitude JND can thus be calculated for a length P of the marking signal W equal to the cardinal of the set of modular coefficients of

the sub-band considered.

  However, it is wise to exploit the fact that the image signal itself

  even allows you to hide the modulation.

  Also, in this example, for each coefficient X to be modulated, a weighting coefficient is used according to a law of the form cai = ki. cav, where ki is a modulation factor depending on the coefficients located in the vicinity of the coefficient considered in the region R and av is equal to the value

  maximum JND of the weighting coefficient.

  Thus, each coefficient will be modulated according to the local content, which locally increases the amplitude of the modulation for the benefit of

detectability.

  To test detectability in test step E101, a detectability amplitude is then calculated from the detectability criterion and this detectability criterion is validated by comparing the detectability amplitude with a value of

predetermined threshold noted Tc.

  One generally operates a correlation calculation C (X *, W) between the set of modulated coefficients X * and the marking signal W and one decides whether there has actually been watermarking of the image when the result of the calculation of

  correlation C (X *, W) is greater than the predetermined threshold value Tc.

  One can also use a standardized statistical test for detection such as that described in the article entitled "A method for signature casting on digital images" by I. PITAS, in Proc. ICIP, pages 215-218, September 1996. We then characterize the detection in terms of probability. We can thus choose a threshold value Tc corresponding to a probability rate

  fixed detection, for example 99.95%.

  Thus, for a given region R defining a set of coefficients X of size P and with a fixed weighting coefficient aX, one can calculate an amplitude of detectability and compare it with a threshold value Tc so as to validate or not the criterion detectability on this region R. The detectability criterion is in theory a function dependent on the size P of the region R, the detection threshold Tc related to the desired probability of detection, the marking signal W, the variance 62x of the coefficients X to be modulated on the region R and of the weighting coefficient used a for the

  modulation of the coefficients X according to the linear model described above.

  Thus, for example, for a detection by standardized hypothesis test and a value of the weighting coefficient a constant and equal to cxv for all the coefficients Xi of the region R, a minimum length Pmin (X) is defined by the function: Pmin (X) = 2Tc) j [hc2 + ca2 + dl (aa 'oa, b, c and d can be assimilated to constants. The signal being non-stationary, that is to say that the variance a2x being a function of space, it is very necessary to adapt the modulation length P generated by the

  marking signal to the image signal itself.

  At decoding, all the coefficients X being modulated, this function also depends on the variance a2w of the inserted marking signal W. In addition, it is necessary to provide for any distortions which may be suffered by the image after modulation, due for example to a

image compression.

  When this noise can be considered as additive and decorrelated from the coefficients, one can possibly use a model depending on the variance O2q of this additive noise modeling possible distortions

  applied to said coefficients after modulation.

  Thus, at a given weighting coefficient ax, making it possible to ensure the invisibility of the inserted marking signal, a detection threshold Td fixed as a function of the probability of detection desired at decoding and a set of selected coefficients X, also corresponds to a length minimum modulation Pmin (X) which ensures the detectability of the inserted marking signal. During decoding, for detection by standardized hypothesis test, this minimum length Pmin (X) is then defined by the function:

) 22 2

  Pmin (X) = (- 2Td [b (+ a + r) + ca2 + d] aa

  o a, b, c and d can be compared to constants.

  It is therefore necessary if the same detectability test is used during coding and decoding, to choose a threshold value Tc at coding greater than that Td used during decoding if we want to be able to find the same

partition when decoding the image.

  In practice, the threshold value Td is fixed for the decoding that it is desired to use, that is to say for example the desired probability of detection rate when the detection is carried out by a standardized hypothesis test, and we chooses a threshold Tc with higher coding, in order to take into account the variance of the marking signal W and the variance of an additive noise fixed a priori so as to

  anticipate the influence of post-processing undergone by the image.

  This condition thus makes it possible to find the same detectability criterion which constrains the partition of the digital image without having to transmit to the decoder the partition produced during the insertion of the marking signal W. In this exemplary embodiment, a partition is produced by iterative division of the digital image. It is therefore considered in step E101 that the detectability criterion is verified when the detection amplitude

  calculated is greater than or equal to the prefixed threshold Tc.

  If the detectability criterion is not verified, then it is very likely that the detectability criterion is also not verified on sub-regions of the initial region R of smaller size since the probability of detection generally decreases with the size of the regions. A stop step E102 of the method for determining a partition makes it possible to stop the process of

iterative division of the image.

  As insertion is not considered possible in the initial region R, the

  capacitance C of the image is therefore zero.

  If on the other hand, the detectability criterion is verified, that is to say here that the detection amplitude is greater than or equal to the threshold value Tc, it is possible that it is also verified for sub-regions of the initial region R. It is therefore judicious to try to divide the region R as long as the detectability criterion is checked in order to increase the capacity C. There are of course several possible means for operating the spatial division of the region R into subregions Rj, of sizes Pj, with j = 1 ..., K. We choose in this example a division into a quaternary tree (in

  English quadtree) as shown in Figure 2.

  This division consists of dividing each region into four sub-regions

of identical size.

  Thus, by way of example in FIG. 2, a digital image of initial size N has been partitioned into two sub-regions R5 and R6 of sizes N / 4, seven sub-regions R1, R2, R3, R4, R7, R8, R9 of sizes N / 16 and four sub-regions

RIO, R 1, R12, R13 of sizes N / 64.

  Thus, in the division step E103, the region R is divided into subregions Rj, j = 1, ..., K of sizes Pj. In the case of a division into a quaternary tree, K =

4 and Pj = P / 4.

  We consider in step E104 a first sub-region of index j (for example j = 1) and we initialize a potential gain in capacity value, noted

Gc, at-1.

  The partition determination method then comprises a prior step of comparison E105 of the size Pj of the region Rj with the minimum size Pstat corresponding to the minimum size of a statistically significant sample for the detection of a marking signal inserted in said region

as previously described.

  If the size Pj of the region Rj is less than this minimum size Pstat, in step E108 the index J = J + 1 is incremented to process the following subregion Rj + i. Otherwise, we check the detectability criterion on the subregion Rj of the

  same manner as previously described with reference to step E101.

  In practice, we insert the marking signal on the coefficients of the subregion Rj, we calculate a detection amplitude and compare this

  amplitude of detectability at the threshold value Tc.

  The detection amplitude can be a correlation value between the modulated signal X * and the carrier of the marking signal W or the result of a

standardized hypothesis test.

  If the detectability criterion is verified, the gain value is increased

  capacity potential Gc of a unit in step E107.

  Then, in step E108, the index j of the sub-regions is incremented and it is verified in the test step E109 that this index j is less than or equal to the number of sub-regions K. If so, we repeat the steps E105 to E109 on the next sub-region. Otherwise, when the index j = K + 1, it is checked in a test step E110 if the potential gain value of capacity Gc is strictly positive, that is to say if there are at least two sub-regions Rj of the initial region R for which

  the detectability criterion is validated.

  If not, the higher rank partition performed on the initial region R is not validated and this region R is preserved. This is the case when the detectability criterion has not been validated for any of the subregions Ri

  or only for a single subregion Rj.

  In the latter case, it is preferable to keep the region R for the insertion of the marking signal as soon as it is of size P greater than the size of the single validated subregion Rj. In step El 12, the value of the capacitance C is then increased by

  unit and we keep the region R of size P in the partition.

  An effective insertion step E113 by modulation of the coefficients of the region R can then be implemented in the usual way using the

  linear insertion model described above.

  An El 14 elimination step then makes it possible to eliminate the region R

of a stack of regions to be treated.

  Conversely, if in step El 110, the value of the potential gain in capacity Gc is strictly positive, in all of the subregions Rj for which the detectability criterion is validated in the battery

of R regions to be processed.

  A return step El15 makes it possible to repeat all the steps

  E103 to E1 13 to process the regions of the stack recursively.

  At the end of the recursive partitioning method thus described, a capacitance value Cm is obtained corresponding to the maximum capacity of the image for the insertion of a marking signal while guaranteeing a correct detection of this signal. This maximum capacity Cm is equal to the cardinal of the distinct regions obtained by said partition on which the detectability criterion is validated. Furthermore, the insertion step proper El113 can be omitted during a first "pass" which then only makes it possible to determine the

  maximum capacity Cm of the image and validated regions.

  Once the maximum capacity Cm has been calculated and the corresponding partition stored, it is then possible for the user to determine the message he wishes to insert, corresponding to a number of information bits equal to the maximum capacity Cm, and to then insert this message in the usual way by modulating the coefficients of the different regions

  obtained by the partition determined beforehand.

  We will now describe, with reference to FIG. 3, a second embodiment of the invention in which it is wished to insert into a picture I a message of fixed capacity Cs. Of course, this fixed capacity Cs must have a value lower than the maximum capacity Cm of the digital image to be marked. As illustrated in FIG. 3, the method of determining a partition in this embodiment comprises steps E100 to El 13 identical to those of the method of determining a partition of the embodiment of the

  figure 1 and that it is not necessary to describe again here.

  As before, in step El112, the value of the

  capacity C when the partition of directly higher rank is not validated.

  Then, the method includes a comparison step E116 in which this value of the capacitance C is compared with the fixed capacitance value Cs. In practice, it is tested in step El116 if the capacity value C is equal

  at the fixed capacity value Cs.

  If so, a stop step E117 makes it possible to stop the iterative partitioning of the image as soon as there is a number of

  regions equal to the number of information bits Cs that one wishes to insert.

  Otherwise, we delete, in a deletion step E114, the region

  processed from the stack of regions R to be treated formed in step E111.

  The method for determining a partition further comprises in this embodiment a sorting step El118 which makes it possible to sort the regions R to be processed from the stack by value of detectability amplitude calculated on these regions R. Thus, a step return El115 makes it possible to reiterate the partition from the dividing step E103 for a region R of the stack having an amplitude of

* highest detectability among all other regions to be treated.

  A sorting variant could consist in sorting the regions to be treated in decreasing order of size so as to prioritize the regions of more

  large size which a priori ensure better detectability.

  Of course, one could perform in the same way a partition by iterative fusion of the digital image, for example by also using

  a quaternary tree partitioning structure.

  The initial partition is then made up of regions of smaller admissible sizes, greater than or equal to the minimum size defined above.

Pstat-

  For each region of the partition for which the detectability criterion is not validated, this region is merged with another region of

the partition.

  This other region of the partition is if possible a region for which

  the detectability criterion is not validated in order to increase the capacity.

  Alternatively, this other region may be a region satisfying the

  detectability criterion in order to make detection more reliable.

  There are multiple possibilities for choosing the regions to merge.

  The detection amplitude calculated on each of the regions can be taken as the fusion criterion and the regions associated with the

  lowest detection amplitudes.

  In the particular case of a fusion in a partitioning structure in a quaternary tree, the fusion is structurally constrained and consists in deciding whether four adjacent blocks, for example R1, R2, R3 and R4 on

  Figure 2, must be merged or not.

  We will now describe, with reference to FIG. 4, the associated decoding method which makes it possible to find the partition during the extraction of the

  marking signal inserted in a digital image.

  Indeed, due to the adaptation of the partition to the digital image, this

  partition will not be known a priori at the level of the decoder.

  The decoding process then generally consists in testing all of the admissible partitions. We use the same iterative segmentation technique as that used when determining the partition

  for the insertion of the marking signal.

  We will thus describe an exemplary embodiment in the case of a segmentation by division by blocks into a quaternary tree, symmetrical to that used in the process of determining a partition of the mode of

  embodiment described with reference to FIG. 1.

  At an initialization step E200, there is an initial partition limited to a single region R of size P. This region consists of all the coefficients of the same representation of the digital image as that used

  when inserting the marking signal.

  A test step E201 makes it possible to verify that the size P of the initial region R is much greater than the minimum size Pstat as defined above. In the affirmative, a step E203 is checked for a detectability criterion of the information bit inserted on the region R, obtained by demodulation of the coefficients on this region R. In practice, a detection amplitude is calculated for the set coefficients of the region R and the detectability criterion is validated by comparing this detection amplitude to a predetermined threshold value Td for the

decoding.

  We can operate by a standardized statistical test and we compare the

  result at the predetermined threshold value Td.

  As explained above, it is necessary, if the same detectability test is used during coding and decoding, to choose a threshold value Tc at coding greater than that Td used during decoding if we want to be able to find the same partition when decoding the image even when

  it was affected by post-processing.

  This condition thus makes it possible to find the same detectability criterion which constrains the partition of the digital image without having to transmit to the decoder the partition produced during the insertion of the marking signal W. If the detectability criterion is not verified in step E203, a division is made in step E204 of the region R into subregions Rj of sizes Pj with j = 1, ..., K. We then use the same partitioning structure as that used coding, that is to say in this example a division of the region R into four subregions Rj of identical sizes. Then, in a step of adding E205, the subregions Rj are added in a stack of regions R to be processed and the decoding process from the test step E201 is repeated recursively in a return step E202.

each region R of the stack.

  When in the step of calculating the detectability criterion E203, this is verified for a region R, the information bit inserted in this region R is extracted in the extraction step E206 as an element of the message inserted in the digital image I. This extraction is carried out in a conventional manner by calculating a detection amplitude. The result of the test gives an absolute amplitude of detection to compare with the threshold value Td of

  decoding, the sign giving the value 0 or 1 of the information bit inserted.

  Note that this extraction is already partly carried out during the detection step E203 so that the detection steps E203 and extraction

  E206 are structurally coupled.

  A deletion step E207 is implemented to eliminate the region R from the stack of regions to be treated constituted in the addition step E205 and the return step E202 then makes it possible to implement the method of

  recursively decoding on each region R of the stack to be processed.

  We will now describe with reference to FIGS. 5 and 6 a practical application of the method for determining a partition of a digital image.

  in a particular embodiment of the invention.

  In this application, the insertion of the marking signal is carried out by a spread spectrum insertion technique, by modulation of coefficients of a space-frequency representation of the image, obtained by a spatio-frequency transformation S of l digital image I. We use here a discrete wavelet decomposition S represented schematically in FIG. 6. This spatio-frequency decomposition S implemented in the transformation step E300 makes it possible to split the image into Si frequency sub-bands with j = 1, ..., M. This decomposition into wavelets makes it possible to obtain hybrid coefficients, that is to say spectral coefficients also located in space, here in the plane of the image. A diagram of a standard wavelet decomposition of an image is illustrated in FIG. 6 and the principle of such a decomposition is recalled below.

multi-resolution spectral.

  Image I consists of a series of digital samples. Image 1 is for example represented by a series of bytes, each byte value representing a pixel of image 1, which can be a black and white image, at 256

shades of grey.

  The multi-resolution spectral decomposition means consist of a sub-band decomposition circuit, or analysis circuit, formed of a set of analysis filters, respectively associated with decimators in pairs. This decomposition circuit filters the image signal I in two directions, into sub-bands of low frequencies and high spatial frequencies. The circuit includes several successive analysis blocks to decompose the image I

  in sub-bands according to several levels of resolution.

  By way of example, image I is broken down here into sub-bands at a

  level of decomposition equal to 4.

  A first analysis block receives the image signal I and filters it through two digital low-pass and high-pass filters, respectively, in a first direction, for example horizontal. After passing through decimators in pairs, the resulting filtered signals are in turn filtered by two filters, low pass and high pass, in a second direction, for example vertical. Each signal is passed through a decimator in pairs again. We then obtain at the output of this first analysis block, four sub-bands LL1, LH1, HL1 and HH, with the highest resolution in the

decomposition.

  The sub-band LL1 comprises the low-frequency components in the two directions of the image signal I. The sub-band LH1 comprises the low-frequency components in a first direction and of high frequency in a second direction of the image signal I. The HL1 sub-band comprises the high frequency components in the first direction and

  the low frequency components in the second direction. Finally, the sub-

  HH1 band includes the high frequency components in both directions. A second analysis block in turn filters the low frequency sub-band LL1 to provide in the same way four sub-bands LL2, LH2 HL2 and HH2 of intermediate resolution level in the decomposition. A third analysis block then filters the low frequency sub-band LL2 to provide four sub-bands LL3, LH3, HL3 and HH3. Finally, in this example, the LL3 subband is in turn analyzed by a third analysis block to provide four resolution LL4, LH4, HL4 and HH4 subbands.

  weaker in this decomposition.

  This gives 13 sub-bands and four levels of resolution. Well

  heard, the number of levels of resolution, and therefore of sub-

  bands, can be chosen differently.

  Of course other types of spectral transformation could be used such as the discrete Fourier transformation, the transformation in

  discrete cosine or the Fourier-Mellin transformation.

  We generally obtain Sj sub-bands with J = 1, ..., M and M

equal here to 13.

  Then, for each sub-band Sj of index j considered in step E301, a partition is performed in step E302 by implementing the method of determining a partition described previously with reference to FIG. 1

on the Sj sub-band.

  We thus deduce a capacity Cj for each sub-band Sj.

  An E303 insertion step by modulating the coefficients allows

  then insert Cj bits of information into the Cj validated regions.

  We then consider in steps E304 and E305 the following sub-bands as long as all the sub-bands Sj have not been processed and we reiterate on each sub-band If the steps E302 and E303 partition and insertion

  of a capacity marking signal Cj.

  Then, an inverse transformation S-1 is applied to the image to obtain the marked image, of total capacity C equal to the sum of the capacities Cj calculated for each of the sub-bands Sj. Thus, each sub-band of the representation is considered independently for the adaptive partitioning which is operated at the same time

coding and decoding.

  As illustrated in FIG. 7, the decoding method in this embodiment is implemented in a similar manner using the decoding method described with reference to FIG. 4, the initial region R corresponding here

  each time to a sub-band of the decomposition of the image.

  A step E400, identical to step E300, makes it possible to split the image I into Sj sub-bands with j = 1, ... M. We consider a first sub-band Sj at step E401. The order of processing of the sub-bands is identical to that used during coding in order to

  extract the inserted message in order.

  Adaptive partitioning is implemented at the partition step E402 in order to calculate the capacity Cj of the sub-band Sj. Then, at the extraction step E403, the Cj information bits are extracted by demodulation of the

  coefficients of the Cj regions of the partition.

  One verifies in the usual way in the test steps E404 and E405 that all the sub-bands have been processed and in the negative, the decoding steps E402 and E403 (partition and extraction) are repeated on the subbands

remaining.

  A message of C bits is thus obtained at the output of the decoder, C being

  equal to the sum of the capacities Cj ce each sub-band Sj.

  In a variant of this method illustrated in FIG. 8, it is possible to perform partitioning on a post-processed image signal in order to obtain a priori robustness of the marking signal with possible post-processing undergone

by image.

  The principle of this variant consists in carrying out a detectability test in a test step E309 which is no longer coupled to the partition steps.

  E302 and insertion E303 described above.

  After the insertion of the marking signal in a region, an inverse transformation S ′ is first carried out in step E306 to find the marked image. Then, in a step E307, post-treatments are applied, such as

  than digital image compression.

  We then reiterate in a step E308 the transformation S for

  find the marked coefficients possibly noised by the post-

  processing performed on the image in step E307, and the criterion of

  detectability as previously described.

  The steps E300 to E309 are repeated on another admissible partition of the image, here for example by dividing the region considered R into four, and the detectability test is carried out in step E309 on the

four sub

regions.

  This partition is only retained if it represents a gain in capacity, that is to say in this example of division into a quaternary tree, if the criterion of

  detectability is validated on at least two sub-regions.

  This variant of realization makes it possible to take into account directly

  post-processing which will be applied to the digital image.

  The present invention also relates to a partition determination device and a decoding device which are suitable for implementing the partition determination and decoding methods.

described previously.

  These partition determination and decoding devices can be implemented in a computer 50 as illustrated in FIG. 5,

  independently or on the same computer 50.

  The partition determination device comprises partitioning means 500, 502, 503 adapted to perform adaptive partitioning as a function of a detectability criterion of an information bit inserted in each region. These partitioning means use, during coding, a maximum value cxv of the weighting coefficient ensuring the imperceptibility of the marking signal for the modulation of all the coefficients representative of said digital image, possibly affected by a modulation factor k depending on each coefficient. as previously described. This device for determining a partition also comprises means 500, 502, 503 of comparing the size P of each region R of the partition with a minimum size Pstat corresponding to the minimum size of a statistically significant sample for detection. a marking signal

inserted in said region.

  The partitioning means comprise means of modulation 500, 502, 503 of the set of coefficients of each region by inserting the marking signal, calculation means 500, 502, 503 of a detectability amplitude from the criterion of detectability and validation means 500, 502, 503 of the detectability criterion adapted to compare

  the detectability amplitude at a predetermined threshold value Tc.

  The partitioning means 500, 502, 503 are adapted to perform a partition by iterative division of the digital image and for each region in one row of said partition, the validation means 500, 502, 503 are adapted to validate a partition d '' a directly higher rank in the region if and only if there are at least two sub-regions of the region for which

  the detectability criterion is validated.

  This device further comprises means 500, 502, 503 of incrementing the value of a capacity C of the digital image adapted, for each region to one row of the partition, to increment the value of capacity C when the partition of directly higher rank is not validated, the capacity C being equal to the cardinal of the distinct regions obtained by the partition on which the

  detectability criterion is validated.

  In another embodiment, this device for determining a partition comprises incrementing means 500, 502, 503 adapted, for each region R to a rank of the partition, to increment the value of a capacity C when the partition of directly higher rank is not validated, comparison means 500, 502, 503 adapted to compare the value of the capacity C with a fixed capacity value Cs and reiteration means 500, 502, 503 adapted to reiterate the partition for a region having the highest detectability amplitude among all the other regions to be treated when the capacity value C is less than the fixed capacity value Cs. In a third embodiment of the invention, the partitioning means 500, 502, 503 are adapted to perform a partition by iterative fusion of the digital image and are adapted, for each region of the partition for which the detectability criterion is not validated, to merge the region with another region of the partition, and preferably with another

  region of the partition for which the detectability criterion is not validated.

  In a completely analogous manner, the device for decoding a marking signal in a digital image comprises partitioning means 500, 502, 503 into regions distinct from the digital image to be decoded adapted to perform adaptive partitioning as a function of '' a criterion of detectability of an information bit obtained by demodulation of the coefficients

on each region.

  It further comprises comparison means 500, 502, 503 of the size P of each region R of the partition with a minimum size Pstt corresponding to the minimum size of a statistically significant sample.

  for the detection of a marking signal inserted in the region.

  It comprises means of calculation 500, 502, 503 of a detection amplitude from the detectability criterion and validation means 500, 502, 503 of the detectability criterion adapted to compare the detection amplitude

  at a predetermined threshold value Td for decoding.

  This predetermined threshold value Td for decoding is less than the predetermined threshold value Tc used during the determination process

  of a partition described above.

  All the means of the device for determining a partition and of the decoding device are incorporated in a microprocessor 500 of the computer 50, a read-only memory 502 being adapted to memorize a program for determining a partition and / or decoding the signal. marking and a random access memory 503 comprising registers adapted to memorize

  variables changed during program execution.

  The microprocessor 500 is integrated into the computer 50 which can be connected to various peripherals, for example, a digital camera 507 or a microphone 511 via an input / output card 506 in order to

  receive and store documents.

  The digital camera 507 allows in particular to provide

  images to be authenticated by insertion of a marking signal.

  This computer 50 includes a communication interface 512 connected to a communication network 513 to optionally receive

images to mark.

  The computer 50 further comprises document storage means, such as a hard disk 508, or is adapted to cooperate by means of a floppy disk drive 509 with document storage means

  removable such as 510 floppy disks.

  These fixed or removable storage means may also include the code of the insertion process according to the invention, which, once read

  by the microprocessor 500, will be stored in the hard disk 508.

  As a variant, the program allowing the insertion device to implement the invention may be stored in the memory

  still life 502 (ROM or Read Only Memory in English).

  In the second variant, the program can be received to be stored as described above via the network of

communication 513.

  The computer 50 also has a screen 504 allowing for example to serve as an interface with an operator using the keyboard 514 or

any other way.

  The central unit 500 (CPU) will execute the instructions relating to the implementation of the invention. During power-up, the programs and methods relating to the invention stored in a non-volatile memory, for example read-only memory 502, are transferred to the random access memory 503 (RAM or Random Access Memory in English) which will then contain the executable code of the invention as well as the variables necessary for the implementation

work of the invention.

  This random access memory 503 comprises various registers for storing the variables necessary for the execution of the program, and in particular a register for storing the coefficients of the regions R which are modulated or demodulated temporarily in order to verify the detectability criterion, the minimum value

  Pstat, the threshold value Tc and Td respectively for coding and decoding.

  A communication bus 501 allows communication between the different sub-elements of the computer 50 or linked to it. The representation of the bus 501 is not limiting and in particular the microprocessor 500 is capable of communicating instructions to any sub-element directly

  or through another sub-element.

  Of course, many modifications can be made

  to the examples described above without departing from the scope of the invention.

  Thus any other type of partitioning than those described

previously can be used.

  In particular, it is possible to use various techniques of iterative segmentation by division / fusion, to which belongs the block partitioning in the quaternary tree described above, but also a partitioning in graph. In general, a distinction is made between so-called bottom-up methods (bottom-up in English) and top-down methods (in English top-down) according to

  that the iterative means of partitioning is fusion or division.

  In addition, it is possible to combine two partitioning techniques by division and fusion. This is the case when one chooses for example an intermediate initial partition, between the two extremes that are on the one hand a single region and on the other hand a partition in smaller regions

  admissible, authorizing both division and merger operations.

  One can also implement the method of determining a partition in accordance with the invention to maximize the capacity of the digital image by using several different segmentation techniques and then choose the partition which provides the greatest capacity C for the digital image.

Claims (46)

  1. Method for determining a partition into distinct regions of a digital image (I) for inserting a marking signal (W) by modulating coefficients (X) of said regions (R) representative of a digital image (I) , characterized in that the partition into regions is carried out by an adaptive partitioning according to a criterion of detectability of a bit   of information inserted on each region (R).   2 A method for determining a partition according to claim 1, characterized in that the detectability criterion is calculated using a maximum value ((xv) of the weighting coefficient ensuring the imperceptibility of the marking signal for the modulation of the set of   coefficients representative of said digital image.   3. Method for determining a partition according to claim 2, characterized in that the detectability criterion is calculated by using, for each coefficient (Xi) to be modulated, a weighting coefficient according to a law of the form oi = ki. ov, o ki is a modulation factor depending on the coefficients located in the vicinity of the coefficient considered (Xi) on said region and cv is equal to said maximum value of the coefficient weighter.   4. Method for determining a partition conforming to one of the   Claims 1 to 3, characterized in that it includes a prior step of   comparison (E105) of the size (P) of each region (R) of the partition with a minimum size (Pstat) corresponding to the minimum size of a statistically significant sample for the detection of an inserted marking signal in said region.   5. Method for determining a partition conforming to one of the   Claims 1 to 4, characterized in that for each region (R) of the   partition, we modulate all the coefficients of said region by inserting the marking signal (W), we calculate a detectability amplitude (T) from said detectability criterion and we validate said detectability criterion by comparing said detectability amplitude (T) at a threshold value predetermined (Tc).   6. Method for determining a partition according to claim 5, characterized in that a partition is made by iterative division (E103) of the digital image and in that, for each region at one rank of said partition, we validate a partition of a directly higher rank in said region if and only if there are at least two sub-regions of said region   region for which said detectability criterion is validated.   7. A method of determining a partition according to claim 6, characterized in that for each region in a row of said partition, the value of a capacitance (C) of the digital image is incremented when the partition of directly higher rank is not validated, said capacity (C) being equal to the cardinal of the distinct regions obtained by   said partition on which the detectability criterion is validated.   8. A method of determining a partition according to claim 6, characterized in that for each region at one rank of said partition, the value of a capacity (C) is incremented (E112) when the partition of rank directly higher is not validated, we compare (El 15) said capacity value (C) with a fixed capacity value (Cs) and we reiterate the partition for a region having the highest detectability amplitude among all the other regions to be processed when said capacity value is   lower than said fixed capacity value.   9. A method of determining a partition according to claim 5, characterized in that a partition is performed by iterative fusion of the digital image and in that, for each region of said partition for which said detectability criterion n is not validated, we merge   said region with another region of said partition.   10. A method of determining a partition according to claim 9, characterized in that said other region of the partition is so   possible a region for which said detectability criterion is not validated.   11. Method for decoding a marking signal in a digital image, inserted by modulation of coefficients (X) representative of said image in distinct regions forming a partition of said digital image, characterized in that a partition is determined by distinct regions of the digital image to be decoded by performing an adaptive partitioning as a function of a criterion of detectability of an information bit obtained by demodulating the coefficients on each region (R). 12. A decoding method according to claim 11, characterized in that it comprises a preliminary step of comparison (E201) of the size (P) of each region of the partition with a minimum size (Pstat) corresponding to the minimum size a statistically significant sample for detection   of a marking signal inserted in said region.   13. A decoding method according to one of claims 11 or   12, characterized in that for each region (R) of the partition, a detection amplitude (T) is calculated (E203) from said detectability criterion and said detectability criterion is validated (E203) by comparing said amplitude of   detection (T) at a predetermined threshold value (Td) for decoding.   14. A decoding method according to claim 13, characterized in that the predetermined threshold value (Td) for decoding is less than said predetermined threshold value (Tc) used during the method of   determination of a partition according to one of claims 6 to 11.   15. Device for determining a partition into regions distinct from a digital image for inserting a marking signal by modulation of coefficients of said regions representative of a digital image, characterized in that it comprises partitioning means (500, 502, 503) adapted to perform an adaptive partitioning according to a detectability criterion   of a bit of information inserted on each region.   16. Device for determining a partition according to claim 15, characterized in that the detectability criterion is calculated by using a maximum value of the weighting coefficient ensuring the imperceptibility of the marking signal for the modulation of all of the   coefficients representative of said digital image.   17. Device for determining a partition according to claim 16, characterized in that the detectability criterion is calculated by using, for each coefficient to be modulated, a weighting coefficient according to a law of the form oi = ki.oxv, o ki is a modulation factor depending on the coefficients located in the vicinity of the coefficient considered on said region   and av is equal to said maximum value of the weighting coefficient.   18. Device for determining a partition in accordance with one of the   Claims 15 to 17, characterized in that it further comprises means   comparison (500, 502, 503) of the size of each region of the partition with a minimum size corresponding to the minimum size of a statistically significant sample for the detection of an inserted marking signal in said region.   19. Device for determining a partition conforming to one of the   claims 15 to 18, characterized in that the partitioning means   (500, 502, 503) comprise means of modulation (500, 502, 503) of the set of coefficients of each region by inserting the marking signal, means of calculation (500, 502, 503) of an amplitude detectability based on said detectability criterion and validation means (500, 502, 503) of said detectability criterion adapted to compare said   amplitude of detectability at a predetermined threshold value.   20. A device for determining a partition according to claim 19, characterized in that the partitioning means (500, 502, 503) are adapted to perform a partition by iterative division of the digital image and in that, for each region has a rank of said partition, the validation means are adapted to validate a partition of a rank directly higher in said region if and only if there are at least two sub-regions   of said region for which said detectability criterion is validated.   21. A device for determining a partition according to claim 20, characterized in that it further comprises means for incrementing (500, 502, 503) the value of a suitable capacity of the digital image, for each region at one rank of said partition, to increment said capacity value when the partition of directly higher rank is not validated, said capacity being equal to the cardinal of the distinct regions obtained by said partition on which the detectability criterion is validated . 22. Device for determining a partition according to claim 20, characterized in that it further comprises incrementing means (500, 502, 503) adapted, for each region to a row of said partition, to be incremented the value of a capacity when the partition of directly higher rank is not validated, comparison means (500, 502, 503) adapted to compare said capacity value with a fixed capacity value and reiteration means ( 500, 502, 503) adapted to reiterate the partition for a region having the highest detectability amplitude among all the other regions to be treated when said capacity value is   lower than said fixed capacity value.   23. Device for determining a partition according to claim 19, characterized in that the partitioning means (500, 502, 503) are adapted to perform a partition by iterative fusion of the digital image and are adapted, for each region of said partition for which said detectability criterion is not validated, to merge said region with a other region of the said partition.   24. A device for determining a partition according to claim 23, characterized in that said other region of the partition is so   possible a region for which said detectability criterion is not validated.   25. Device for determining a partition in accordance with one of the   claims 15 to 24, characterized in that it is incorporated in a   microprocessor (500), a read only memory (502) being adapted to memorize a program for determining a partition and a random access memory (503) comprising registers adapted to memorize variables modified during the execution of said program.   26. Device for decoding a marking signal in a digital image, inserted by modulation of coefficients representative of said image in distinct regions forming a partition of said digital image, characterized in that it comprises partitioning means (500 , 502, 503) into distinct regions of the digital image to be decoded adapted to perform adaptive partitioning as a function of a bit detectability criterion   of information obtained by demodulating the coefficients on each region.   27. A decoding device according to claim 26, characterized in that it further comprises means of comparison (500, 502, 503) of the size of each region of the partition with a minimum size corresponding to the minimum size d '' a statistically significant sample for detection   of a marking signal inserted in said region.   28. Decoding device according to one of claims 26 or   27, characterized in that it comprises means for calculating (500, 502, 503) a detection amplitude from said detectability criterion and validation means (500, 502, 503) for said detectability criterion adapted to comparing said detection amplitude to a predetermined threshold value for decoding.   29. Decoding device according to claim 28, characterized in that the predetermined threshold value for decoding is less than said predetermined threshold value used during the determination process   a partition according to one of claims 5 to 10.   30. Decoding device according to one of claims 26 to   29, characterized in that it is incorporated in a microprocessor (500), a read only memory (502) being adapted to store a program for decoding a marking signal and a random access memory (503) comprising registers adapted to store variables modified during the execution of said program. 31. Computer, characterized in that it comprises means suitable for implementing a method for determining a partition in accordance with one of claims 1 to 10.   32. Computer, characterized in that it comprises means suitable for implementing the method of decoding a conforming marking signal   to one of claims 11 to 14.   33. Computer, characterized in that it incorporates a device for   determination of a partition according to one of claims 15 to 25.   34. Computer, characterized in that it incorporates a device for   decoding of a marking signal according to one of claims 26 to   30. 35. Apparatus for processing a digital image, characterized in that it comprises means suitable for implementing a method of   determination of a partition according to one of claims 1 to 10.   36. Apparatus for processing a digital image, characterized in that it comprises means suitable for implementing the method of   decoding of a marking signal according to one of claims 11 to 14.   37. Apparatus for processing a digital image, characterized in that it incorporates a device for determining a partition conforming to one from claims 15 to 25.   38. Apparatus for processing a digital image, characterized in that it incorporates a device for decoding a marking signal conforming to one of claims 26 to 30.   39. Digital printer, characterized in that it comprises means suitable for implementing a method for determining a partition   according to one of claims 1 to 10.   40. Digital printer, characterized in that it comprises means suitable for implementing the method of decoding a signal   marking according to one of claims 11 to 14.   41. Digital printer, characterized in that it incorporates a   partition determination device according to one of the claims 15 to 25.   42. Digital printer, characterized in that it incorporates a device for decoding a marking signal conforming to one of the claims 26 to 30.   43. Digital camera, characterized in that it comprises means suitable for implementing a method of   determination of a partition according to one of claims 1 to 10.   44. Digital camera, characterized in that it comprises means suitable for implementing the decoding method   of a marking signal according to one of claims 11 to 14.   45. Digital camera, characterized in that it incorporates a device for determining a partition conforming to one of the claims 15 to 25.   46. Digital camera, characterized in that it incorporates a device for decoding a marking signal conforming to one from claims 26 to 30.   47. Digital camera, characterized in that it comprises means suitable for implementing a method for determining a partition   according to one of claims 1 to 10.   48. Digital camera, characterized in that it comprises means suitable for implementing the method of decoding a signal   marking according to one of claims 11 to 14.   49. Digital camera, characterized in that it incorporates a   partition determination device according to one of the claims to 25.   50. Digital camera, characterized in that it incorporates a device for decoding a marking signal conforming to one of the claims 26 to 30.