Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T1/00—General purpose image data processing
  • G06T1/0021—Image watermarking
  • G06T1/0028—Adaptive watermarking, e.g. Human Visual System [HVS]-based watermarking
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T1/00—General purpose image data processing
  • G06T1/0021—Image watermarking
  • G06T1/005—Robust watermarking, e.g. average attack or collusion attack resistant
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2201/00—General purpose image data processing
  • G06T2201/005—Image watermarking
  • G06T2201/0051—Embedding of the watermark in the spatial domain
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2201/00—General purpose image data processing
  • G06T2201/005—Image watermarking
  • G06T2201/0061—Embedding of the watermark in each block of the image, e.g. segmented watermarking
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2201/00—General purpose image data processing
  • G06T2201/005—Image watermarking
  • G06T2201/0065—Extraction of an embedded watermark; Reliable detection
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2201/00—General purpose image data processing
  • G06T2201/005—Image watermarking
  • G06T2201/0083—Image watermarking whereby only watermarked image required at decoder, e.g. source-based, blind, oblivious

Definitions

• the present invention relates generally to the technical field of digital watermarking of an image or of a video sequence.
• It relates in particular to a method and device for inserting information (also called watermark or identification mark) into an image.
• information also called watermark or identification mark
• One aim of this technique is, for example, to identify the author of the tattooed image, thanks to the information inserted.
• magnification step of this solution generates uniform solid areas in the watermarked image, once the watermark is inserted, regardless of the magnification method used. These areas (in particular the luminosity transitions between two adjacent solid areas) are easily visible, which makes the watermark detectable.
• a generation of a sequence of symbols representative of a noise comprising as many symbols as there are elements included in said block, at least one of the symbols being non-zero, so that the result of a predetermined function applied to said symbols is equal to a chosen value
• the invention has the advantage of allowing the insertion of information that is robust to the change in scale of the image without which it was inserted, and that does not generate a solid in the watermarked image.
• the result can be a sum.
• the predetermined function may be a measure of central tendency.
• the predetermined function can be is the median.
• the predetermined function can be the average.
• the predetermined function (applied to the symbols) may be the sum (of the symbols).
• the intermediate information is, for example, obtained from the initial information by means of a magnification method (for example by means of a method of interpolation to the nearest neighbor).
• said chosen value is equal to zero; moreover, provision can be made, for example, for kx or ky to be greater than or equal to 2.
• a generation of a sequence of symbols representative of a noise comprising as many symbols as there are elements included in said block, at least one of the symbols being non-zero, so that the result of a predetermined function applied to said symbols is equal to a chosen value
• the invention also provides, according to another aspect, a recording medium readable by a computer, on which is recorded a computer program comprising program code instructions for the execution of the steps of the method as described above. .
• FIG. 1 illustrates an example of application of an embodiment of the information insertion and extraction methods according to the invention
• FIG. 2 shows an embodiment of an insertion method according to the invention
• Figure 4 shows in more detail a step of the embodiment of Figure 2;
• FIG. 8 shows an example of rendering of information to be inserted
• FIG. 1 describes an example of application of an embodiment of the methods of inserting and extracting information according to the invention.
• a first step E10 implements a method of inserting information into an image.
• This information can for example be a mark making it possible to recognize the author of the image for the purposes of protecting the image by copyright.
• the information (hereinafter called “watermark”) can be inserted into one or more images of a video sequence.
• the latter can be transmitted during a step E11 and then received by a client at step E12.
• the image can then be transformed by the customer, who can for example operate a change of scale, step E13.
• step E10 The information that was inserted in step E10 is extracted from the transformed image, step E14, then analyzed, for example to identify the author of the image, step E15.
• FIG. 2 illustrates in more detail an example of the implementation of the method of inserting information into an image, step E10.
• N1 and M1 being undamaged integers representing the dimensions D1 of the image l (D1).
• These dimensions can be expressed as an N1 x M1 pixel product.
• the dimensions D1 can be equal to 3840 ⁇ 2160 pixels.
• N0 and MO being non-harmful integers representing the dimensions D0 of the image l (D0). It comes N0 £ N1 and / or M0 £ M1, corresponding to the 16: 9 aspect ratio.
• the dimensions D0 are predetermined, for example equal to 640 ⁇ 360 pixels, corresponding to the so-called 360p definition.
• the dimensions D0 are chosen so that the ratio between the number of rows and columns corresponds to a standard image format such as the 16/9 image format.
• these dimensions D0 can imply a number of pixels corresponding to a standard definition SD, ie 480p.
• the dimensions D0 are chosen sufficiently small so that the watermark remains robust to image reductions in formats of acceptable quality from the viewpoint of a viewer, and large enough that the energy of the watermark remains low compared to the energy of the host image. It will be recalled that conventionally, in the field of signal processing, the energy of an image corresponds to the variance of the values of the pixels with respect to a constant value.
• a watermark W (D0) (or initial information) is generated, step E22.
• the watermark W (D0) is also formed of pixels distributed in rows and columns. Methods of calculating a tattoo at a given size from an image having this same size are known to those skilled in the art. Calculation methods are described for example in the Digital watermarking manual mentioned above. It is this initial information that will subsequently be extracted for analysis.
• an enlarged watermark W (D4) is obtained so as to have N4 rows and M4 columns (dimensions D4), step 23, according to an embodiment of the method of the invention.
• N43N1 and M43M1 on the one hand
• N4 and M4 being undesirable integers
• kx and ky being integers greater than or equal to 1.
• the dimensions N4 and M4 are slightly greater than N1 and M1.
• D1 3840 * 2160 pixels
• D4 4480 * 2520 pixels.
• the enlarged watermark W (D4) is inserted into the image 1 (D1), step 24, so as to obtain the watermarked image IW (D1).
• the insertion of the enlarged watermark W (D4) in the image l (D1) is done for example by addition, that is to say by adding the values of the pixels of the watermark W (D4) to the values of the corresponding pixels in image l (D1).
• the enlarged tattoo once inserted is therefore superimposed on the image.
• the image l (D1) can be a still image or one of the images of a video sequence, several images (or even all of the images) of the video sequence being able to be watermarked.
• the image can be a color component, such as luminance or a chrominance component.
• Each color component is made up of elements, one element corresponding to the color component of a pixel.
• FIG. 3 schematically illustrates the changes in dimensions implemented during the various stages. For this figure, it is considered, for the sake of simplification, that the first image 1 (D1) is formed of 4 * 4 pixels.
• FIG. 4 illustrates in more detail the step of resizing the watermark, E23.
• the resizing is done in two stages.
• W second dimension
• FIG. 5 Using a known magnification method for example the so-called Closest Neighbor interpolation method or PPV (in English: “Nearest Neighbor interpolation”), with an integer enlargement factor sx for the rows and sy the columns, we obtain a block V of sx * sy elements v, at step E230.
• PPV Nearest Neighbor interpolation
• a noise b is then generated with as many symbols b [i] as there are elements in the block V (i being an integer such as 1 ⁇ i £ (sx * sy)) . At least one of these symbols b [i] is different from zero.
• the noise and in particular the distribution of the noise is generated so that the result of a predetermined function applied to the symbols b [i] is equal to the same value chosen for the block considered, whatever the values taken by these symbols. bi].
• the predetermined function is a measure of central tendency (applied to the symbols b [i]), here the average of the symbols b [i] or, alternatively, the median of the symbols b [i].
• the values taken by the symbols can be between -50 and 50, that is to say in the interval [-50; 50]
• the noise distribution can be a Gaussian distribution.
• said resulting value is chosen so that the energy of the enlarged watermark does not exceed a predetermined threshold relating to the energy of the host image.
• a predetermined threshold relating to the energy of the host image.
• this chosen value is zero.
• the predetermined function (applied to the symbols b [i]) can be the sum of the symbols b [i].
• the noise is added to all the blocks of the watermark, so that the watermarked image has a better visual quality.
• the blocks are then processed successively according to a predetermined processing order.
• the noise can be produced using a random or pseudorandom numbers generator (P) RNG.
• Pseudo-random generators produce a sequence of mixed bits that depend on an initial seed. This must be initialized with a different value each time it is used. This initial value can be for example the time of the system used, in milliseconds. Each run of the program will use a different seed, resulting in a different random sequence. Conversely, reusing the same seed results in the same pseudo-random bit sequence. Random generators produce sequences of bits in a completely random fashion, the bits being chosen independently and uniformly at random. Each sequence of bits from a new generation is therefore different from the previous ones.
• the noise sequence comprises p symbols (p is an integer greater than 1, which is equal in this example to sx * sy), only p1 symbols with p1 ⁇ p (p1 being an integer greater than or equal to 1) are generated using the generator described above.
• the predetermined function preferably comprises an addition,; as already indicated, the predetermined function is for example a measure of central tendency (such as an average or a median), or, alternatively, a variance, this list not being limiting.
• the chosen value therefore corresponds to a sum, resulting from the addition of the symbols b [i].
• the applied function is the average
• the average (and therefore the sum) of the values taken by the noise components for the block considered is zero.
• the average of the noise components for the whole image is therefore also zero.
• the noise components are respectively added to each element of block V so as to obtain the enlarged and noisy watermark W (D4, V, b) as illustrated in figure 7.
• D4 D1
• the information can be inserted. in the tattoo image l (D1).
• D4> D1 the information must be reduced before being inserted into the image.
• the noise components can take on different values. This is possible by changing the noise generator or the initial seed as explained above.
• the noise generator or the initial seed as explained above.
• the tattoo information therefore varies over time, making them very difficult to detect because they do not follow a regular pattern.
• the energy of the watermark was not reduced because the initial values of the elements were on average not altered.
• the invention also relates to a watermarked image according to an embodiment described above.
• FIG. 8 represents an example of rendering for a watermark W (D4, V) formed of four elements after enlargement and the enlarged watermark obtained W (D4, V, b), after insertion of the noise of step E231.
• the noise generated follows a normal law. Note that the transition between the elements has disappeared after the addition of noise, thus limiting the appearance of adjacent solids in the watermarked image.
• FIG. 9 represents an embodiment of an extraction of a watermark from a watermarked image according to the invention, step E14. It is considered that a change of scale has been carried out on the watermarked image considered IW (D2, V, b), the watermarked image from which the watermark must be extracted, having new dimensions D2, i.e. N2 rows and M2 columns , N2 and M2 being positive non-harmful integers.
• step E201 an attempt is made to reduce the value of the dimensions of the watermarked image IW (D2, V, b) to the values of the dimensions D0. Indeed, it is in these dimensions that the tattoo can be extracted.
• the watermarked image IW (D2, V, b) is enlarged to the value of the second dimension D0, for example using the so-called Closest Neighbor interpolation method.
• D0 the watermarked image IW (D2, V, b) is enlarged to the value of the second dimension D0, for example using the so-called Closest Neighbor interpolation method.
• D0 the change of scale which was carried out on the watermarked image to bring its dimensions to values less than the D0 dimensions strongly degraded the visual quality of the image, D0 having been preferably chosen to correspond to a standard definition.
• This enlargement can for example be carried out using a bilinear algorithm.
• these dimensions D3 of the watermarked image can be reduced to the dimensions D0 using the predetermined function.
• This step will allow the extraction of the watermark with dimensions D0, in particular with the aim of analyzing it, step E201.
• the predetermined function which was used during the insertion of the watermark in step E 10, is applied to the values of the elements of each block V of the watermarked image with dimensions equal to D3.
• the predetermined function may include an addition of terms, such as the mean.
• the value of a pixel of the watermark after reduction is in this case an average sum of the values of the pixels of the block V corresponding to the watermark with dimensions D3.
• These pixel values of the V block include the values of the noise components that were added during the insertion. tattoo.
• the sum resulting from the average of these noise components will be close to, or even equal, to the chosen value, for example zero.
• the value of the pixel of the watermark after reduction is then close to an average sum of the values of the pixels of the block V before the insertion of the noise components. It is equal to it if the transformations undergone by the watermarked image have not had an impact on the values of the noise components added at step E10.
• step E201 can also comprise for each processed block, a subtraction of said chosen value from the value of the element obtained after the reduction of dimensions. We then obtain a final value of this element v corresponding to an averaged sum of the values of the pixels of the block V before the insertion of the noise components.
• step E202 It is then possible to extract the watermark according to a known method described for example in the Digital watermarking manual mentioned above, step E202, then to analyze step E14.
• FIG. 10 illustrates a particular way, among several possible, of producing a DIS device configured to implement an embodiment of an assistance method according to the invention.
• the device DIS comprises a random access memory, for example a RAM memory, an mR processing unit equipped for example with a processor, and controlled by a computer program stored in a read only memory, for example a ROM memory or a hard disk .
• a computer program stored in a read only memory, for example a ROM memory or a hard disk .
• the code instructions of the computer program are for example loaded into the random access memory RAM before being executed by the processor of the processing unit mR.
• FIG. 10 illustrates only one particular way, among several possible, of producing the device DIS so that it performs certain steps of the method according to the invention.
• these steps can be performed either on a reprogrammable computing machine (a PC computer, a DSP processor or a microcontroller) executing a program comprising a sequence of instructions, or on a dedicated computing machine (for example a set of logic gates such as an FPGA or ASIC, or any other hardware module).
• a reprogrammable computing machine a PC computer, a DSP processor or a microcontroller
• a program comprising a sequence of instructions
• a dedicated computing machine for example a set of logic gates such as an FPGA or ASIC, or any other hardware module.
• the processing means is produced with a reprogrammable computing machine
• the corresponding program (that is to say the sequence of instructions) may be stored in a storage medium, which may or may not be removable, this storage medium being partially or fully readable by a computer or processor.
• the invention therefore also relates to the computer program comprising instructions for implementing the method according to one of the above implementation modes, when said program is executed by a processor.

Landscapes

• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Editing Of Facsimile Originals (AREA)
• Image Processing (AREA)
• Image Analysis (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=69104619

Family Applications (1)

Country Status (4)

Family Cites Families (1)

• 2019
  • 2019-09-06 FR FR1909841A patent/FR3100648B1/fr active Active
• 2020
  • 2020-09-02 EP EP20761865.3A patent/EP4026085A1/de active Pending
  • 2020-09-02 US US17/639,720 patent/US20220292624A1/en active Pending
  • 2020-09-02 WO PCT/EP2020/074457 patent/WO2021043817A1/fr unknown

Also Published As

Similar Documents

Legal Events