FR2972829A1 - METHOD FOR COMPRESSION OF DIGITAL IMAGE BY PROJECTIONS - Google Patents

Abstract

the checkerboard (31) is projected in two projections according to two projection directions (35, 37) for determining the numerical image or images of the checkerboard to be compressed, each projection being defined by a set of terms obtained by adding the numerical values of the luminous intensity of the pixels of the checkerboard, along alignments (36, 38) oriented by these two directions, the terms of these two projections are compared to reference terms and the numerical image or images whose pixels are at the intersection of the alignments of these two projections for which the terms are different from the terms of reference, the one or each of the digital images thus determined is projected in different projections according to different projection directions radiating around a centering pixel (11 ) and defined by a set of terms obtained by adding the numerical values of luminous intensity of the pixels of the or each of the digital images along lines aligned by the directions of projection (15), and - the redundant digital images of the checkerboard whose pixels are at the intersection of the alignments of the two projections of the checkerboard for which the terms are identical to the terms of reference.

Description

The invention relates to a method of compressing a digital image in the form of a mosaic of pixels, for a black-and-white image, each pixel of the mosaic has a light intensity on a scale of numerical values of gray.For a color image, each pixel combines a light intensity on a scale of numerical values of gray and three intensities of light on a scale of numerical values of three colors respectively, for example red, yellow and blue, or red, green and blue.The numerical value or values of light intensity of each pixel are coded in one or more bytes by a computer system. define the size of the digital image as being equal to the product of the number of pixels of the digital image and the number of bytes per pixels. An example of a digital image is provided by the tomodensitometric section of a body, obtained by X-ray scanning. Such a method is known from the French patent application FR 2,888,374 and the international application WO 2007/006560 filed by the applicant. The high-resolution scanning described in these documents makes it possible to connect each pixel of the digital image to an elementary area of the tomodensitometric section as thin as (27 μm) 2. For a digital image in black and white, the pixels have a luminance whose numerical value is coded by the computer system for example on a byte. If the body is examined over a typical scanning width of 40 cm, the size of the digital image is equal to "40 cm 2 / (27 μm) 2" or 220 MB. The tomodensitometric section pre-exists. The body is projected into different projections in different angular directions to reconstruct it by the direct generation method described in the documents cited above. The body is thus traversed by an X-ray beam at different angles of rotation of the body relative to the beam or the beam relative to the body. The X-rays emitted by a source are collected by a detector and the variation of intensity between the emission and the reception gives access to the attenuation along the rectilinear trajectory of each X-ray and in particular to the attenuation engendered by the crossing the body. Attenuation is expressed by an exponential law, taking into account photoelectric absorption and Compton scattering. By linearization, the attenuation variation is linked to the sum of the attenuation coefficients per unit length of the X-ray trajectory. A detector having a detection width of 40 cm and a spatial resolution of 27 μm, at the example of that used in the documents cited above, cut the beam into an X-ray number equal to 40 cm / 27 microns, or about 14814. The set of these X-rays corresponds to a projection of the body of which each of the 14814 terms is obtained by adding the attenuation coefficients per unit length along the path of each X-ray. If the terms are encoded by the one-byte computer system and the body is examined at 36 different rotation angles, the reduced size of the digital image representing the tomodensitometric section becomes equal to 14814 x 36, or about 0.53 MB. Thus, the different projections lead to a compression of the digital image in a raster. contribution of 220 MB for 0.53 MB, a rate of about 415. However, such compression is inseparable from the X-ray scanning with which it is implemented, the terms of the projections being constituted, as indicated above, by the set of sums of the attenuation coefficients per unit of length along the path of the set of X-rays. The object of the invention is a method of compression independent of an implementation in ray scanning. X to allow storage or transfer of any digital images, representing or not a tomodensitometric section.

To this end, the subject of the invention is a method of compressing a digital image in the form of a mosaic of pixels each having a luminous intensity on a scale of numerical values of gray or four light intensities respectively over four scales of numerical values, for gray and for three colors, the numerical value or values of luminous intensity of each pixel being coded by a computer system, characterized in that the digital image is projected in different projections according to different directions of projection radiating around a centering pixel, each projection being defined by one or four sets of terms obtained by adding the numerical values of luminous intensity of the pixels on the gray scale or on the four gray scales and the three colors, along alignments oriented by the directions of projection. The digital image is projected in different projections according to different angular projection directions. The projections are distinguished from the X-ray scanner projections in that the terms are computed by the computer system by adding the numerical values of luminous intensity of the pixels along the pixel alignments. The terms of each projection are coded by the computer system for example in a byte on the gray scale for a black and white image. For a color image, each projection has four sets of terms for the four gray scales and three colors respectively. In each set of a projection, the terms are for example coded on a byte and the coding of a pixel thus requires four bytes. This leads to defining a reduced size of the digital image as being equal to the product of the number of terms of the set or one of the sets defining each projection, the number of projections and the number of bytes per term. The compression ratio is the ratio of the size of the digital image to its reduced size. The computer system is preferably a computer with a screen for displaying digital images. The screen has two series of alignments of pixels parallel to each other in each series and oriented, from one series to another, along two reference directions, for example perpendicular. These two reference directions intersect in a centering pixel around which the different projection directions radiate. With each rotation of the digital image at which the computer proceeds around the centering pixel, the terms of two new projections are obtained by adding the new numerical values of luminous intensity of the pixels along the two series of parallel alignments between them. By way of example, the size of a digital image displayed on a computer screen comprising two series of parallel alignments of 14814 pixels each in two perpendicular directions and where each pixel has a numerical value of luminous intensity in one-byte gray scale, equals (14814) 2, or about 220 MB. If the computer rotates the digital image 180 times with respect to the digital screen, 180 different projections are obtained whose terms are calculated, for each pair of perpendicular projections, by adding the numerical values of luminous intensity of the pixels along the two sets of parallel pixel alignments. The reduced size of the digital image becomes 14814 x 180, or about 2.6 MB, and the compression ratio is about 82.

Projections are used to reconstruct the digital image using different reconstruction methods. The reconstruction of a digital image in black and white is performed in one operation from the projections each defined by the set of terms obtained by adding the numerical values of light intensity of the pixels on the gray scale. The reconstruction of a digital image in color takes place in four operations from the projections each defined by the four sets of terms obtained by adding the numerical values of luminous intensity of the pixels respectively on the four scales of gray and the three colors. . The direct generation method was disclosed by the applicant in the documents cited above. It is implemented here for pairs (a, b) of projections, the two projections of each pair being oriented by the two reference directions of the two sets of pixel alignments. The luminous intensity Eij on a scale of gray, red, yellow or blue, the pixel defined by the intersection of the alignment i of the first series with the alignment j of the second series is calculated by the formula : (b) E (a'b) -ci + Ci a) 1 (ze) + zc (, b) ~ 1 n (a) m (b) 2n (a) m (b) i = 1 j = 1 1 In this formula, e) is the term of the projection (a) obtained by adding the numerical values of luminous intensity of the pixels along the alignment i of n (a) alignments the first series, e) the term of the projection (b) obtained by adding the numerical values of luminous intensity of the pixels along the alignment j of the m (b) alignments of the second series, E e) the load i = 1 of the projection (a) and E c (f) is the load of projection (b). By construction, the j = 1 charges of the two projections of a pair are equal.

The luminous intensities Eij thus calculated are adjusted. According to any alignment i or j of pixels of the first or the second series, the sum of the calculated luminous intensities is equal to the term of the corresponding projection: Ee, b) = cj) and ~ E ~ a'b) -c5a) i = 1 j = 1 The computer proceeds, with the help of a rotation operator, to a superposition according to the two reference directions of the digital screen, of the different digital images obtained with the different pairs of projections and reconstructs the digital image searched for by an average, in each pixel, of

10 numerical values of luminous intensity calculated. The method of reconstruction by iterations is known in particular from KAK A. C. and SLANEY M. "Principles of computarized tomographic imaging", SIAM - Classics in Applied Mathematics - Volume 33, 2001. It consists schematically to backproject successively different

15 projections by redistributing, for each projection alignment, the difference between the sum of the luminous intensity values of the reconstructed pixels with the retro-projected projections and the term of the new projection to be back-projected. An iteration is performed when all the projections have been back-projected successively. Several iterations are usually necessary to make

20 converge the reconstruction to a digital image for which the luminous intensity values of the pixels no longer evolve or oscillate. The method of reconstruction by filtered back projection is known in particular from document F. DUBOIS "Reconstruction of tomographic images by filtered retroprojection" Review ACOMEN, volume 4 (2), 1998. Again, it

25 is schematically to retro-project successively different projections. However, the projections are not retro-projected as such but are previously transformed in the Fourier space, filtered in frequency using a filter, for example a "ramp" filter and then retransformed in the real space on the base of the filtered spectrum. For each alignment of

30 projection, the difference between the sum of the luminous intensity values of the reconstructed pixels and the filtered retro-projected projections and the term of the new filtered projection to be retro-projected is uniformly redistributed. The quality of the reconstruction of the image number depends on the number of projections formed during the compression step, whether it is the compression method according to the invention or the inseparable compression of the X-ray scanner. However, for the latter, the number of projections is limited by the irradiation dose provided by each projection. In the compression method according to the invention, the number of projections can be increased in order to improve the quality of the reconstructed digital image at the expense of a reduction in the compression ratio. An index of the quality of the reconstruction is obtained in the following manner using the computer system: for each pixel, the difference e in absolute value and in percentage is calculated between the numerical value of luminous intensity on the reconstructed image and the numerical value of luminous intensity on the digital image before compression, 15 - the arithmetic mean of the deviations <e> is determined on all the pixels of the digital image and, - we obtain the quality index equal to Zoo - <e>. A quality estimator obtained is also used as follows: - the distance 20 e 'in absolute value and in percentage between the projection of the digital image obtained by the projection is calculated, term by term and for the same direction of projection the compression method according to the invention and the projection obtained by adding the luminous intensity values of the pixels of the reconstructed image along the same alignments, - the arithmetic mean <e '> of the deviations for all the 25 different projections and, - obtains the quality estimator equal to 1 oo - <e '>. The quality estimator is a little less accurate than the quality index but is more practical to use as it compares projections of digital images rather than the digital images themselves.

According to a first advantageous embodiment of the invention, the digital image is projected by compartments into groups of projections whose directions radiate around a centering pixel specific to each compartment and whose terms are obtained by adding the numerical values. luminous intensity of the pixels of each compartment along alignments oriented by the projection directions specific to each group. The advantage of this embodiment lies in the ability to compress the digital image faster at constant compression ratio. The compartments are preferably circles of the same radius whose inscribed squares are juxtaposed to mesh the digital image, the square inscribed in the circle being the square whose diagonal is equal to twice the radius. According to a second advantageous embodiment of the invention, a region of the digital image is projected into different projections whose directions radiate around a centering pixel and the terms are obtained by adding the numerical values of the luminous intensity of the images. pixels of the region aligned along alignments oriented by the projection directions. The advantage of this embodiment lies more particularly in the possibility of compressing a single region belonging to a series of digital images taken "fixed frame", that is to say under identical repetitive conditions. The region complementary to the compressed region is identified as being redundant and deleted from all but one of the digital images in the series. This results in a double compression of the series: - by projection of the region considered, and - by suppression of the redundancy affecting the complementary region of the projected region. The invention extends to a method of compressing a checkerboard of digital images, each digital image being in the form of a mosaic of pixels each having a luminous intensity on a scale of numerical values of gray or four intensities respectively on four scales of numerical values, for gray and for three colors, the digital intensity value or values of each pixel being coded by a computer system, characterized in that: - the checkerboard is projected in two projections according to two projection directions for determining the numerical image (s) of the checkerboard to be compressed, each projection being defined by one or four sets of terms obtained by adding the numerical values of luminous intensity of the checkerboard pixels to the scale of gray or respectively on the scales of gray and of the three colors, along alignments directed by these two directions, - the terms of these two projections are compared with reference terms 10 and the digital image or images whose pixels are at the intersection of the alignments of these two projections for which the terms are different from the terms of reference are determined and, - projects the or each of the digital images thus determined into different projections according to different projection directions radiating around a centering pixel and defined by one or four sets of terms obtained by adding the numerical values of the luminous intensity of the pixels of the of each of the digital images, on the gray scale or respectively on the four gray scales and the three colors, along alignments oriented by the directions of projection. The advantage of this embodiment lies in the possibility of compressing only the digital image (s) of the checkerboard, the pixels of which have undergone a variation of light intensity compared with the reference terms. It is illustrated more particularly when: a reference checker is projected in two projections according to the same two projection directions used to determine the numerical image or images of the checkerboard to be compressed, each projection being defined by one or four sets of terms reference numbers obtained by adding the numerical values of luminous intensity of the pixels of the reference checkerboard, on the gray scale or respectively on the four gray scales and the three colors, along alignments oriented by these two directions and the digital images of the checkerboard whose pixels are at the intersection of the alignments of the two projections of the checkerboard for which the terms are identical to the terms of reference are removed for redundancy. This results in a double compression of the checkerboard: by projection of the image or digital images whose pixels have undergone a variation of luminous intensity compared with the reference checkerboard, and by eliminating the redundancy affecting the others. digital images of the checkerboard. The invention also extends to a method of compressing a series of digital images, each digital image of the series being in the form of a mosaic of pixels each having a luminous intensity on a scale of values. digital gray or four light intensities respectively on four scales of numerical values, for gray and for three colors, the numerical value or values of luminous intensity of each pixel being coded by a computer system, characterized in that: 15 - projects each digital image of the series into different projections according to different projection directions radiating around a centering pixel, each projection being defined by one or four sets of terms obtained by adding the numerical values of luminous intensity of the pixels, on the grayscale or respectively on the four gray scales and the three colors, along aligned alignments by the projection directions, the projection terms of two digital images of the series are compared for the same projection directions and one of these two images is redundantly deleted when the compared terms are identical. This results in a double compression of the succession of the digital images: by projection of the non-redundant digital images, and by the suppression of the redundancy affecting the identical digital images.

Other advantages of the invention will emerge in the light of the description of the embodiments illustrated by the drawings. Figure 1 shows a simplified digital image projected in a first direction and a second direction of projection, perpendicular to each other. Figure 2 shows the digital image of Figure 1 projected along a third direction and a fourth direction of projection, perpendicular to each other and rotated 45 ° relative to the two preceding directions of projection. Figure 3 shows a digital image of 512 by 512 pixels representing a face. FIG. 4 shows the digital image of FIG. 3 reconstructed by filtered backprojections from 180 projections traversing the 00-180 ° arc in steps of 1 °. FIG. 5 shows the digital image of FIG. 3 reconstructed by iterative retroprojections from 180 projections traversing the 00-180 ° arc in steps of 1 ° and for 6 iterations. Figure 6 shows a digital image of 2048 by 2048 pixels representing a face. FIG. 7 shows the digital image of FIG. 6 by iterative retroprojections from 180 projections running through the 00-180 ° arc in steps of 1 ° and for 6 iterations. Figure 8 schematically shows a digital image and projection directions radiating around a centering pixel. Figure 9 schematically shows the digital image of Figure 8 divided into four compartments and four groups of projection directions radiating around a centering pixel specific to each group. Figure 10 schematically shows a checkerboard of digital images and two projection directions for determining the digital image (s) of the checkerboard to be compressed.

Figure 11 schematically shows a passive compression box architecture. Figure 12 schematically shows an active compression box architecture.

The compression method according to the invention is first described with the aid of a simplified example illustrated by FIGS. 1 and 2. A digital image 1 is displayed on a computer screen 12 in the form of a mosaic of pixels 2 each having, for a black and white image, a luminous intensity taking a numerical value 1 for black and 0 for white.

1, the digital image 1 is projected in a first direction 3 and a second direction of projection, perpendicular to each other. 2, the digital image is projected along a third direction 7 and a fourth projection direction 9 perpendicular to each other and rotated 45 ° with respect to the two preceding projection directions.

According to the invention, the projection directions radiate around a centering pixel 11 and the projection terms are obtained by adding the numerical values of luminous intensity of the pixels along the alignments 4, 6, 8 and 10 oriented parallel to each of the projection directions 3, 5, 7 and 9. As indicated in FIG. 1, e) is the term of the projection (a) obtained by adding the numerical values of luminous intensity of the pixels along the alignment i of n (a) = 11 alignments 4 parallel to the first direction 3, c (jb) the end of the projection (b) obtained by adding the numerical values of luminous intensity of the pixels along the alignment j of the m (b) = 11 alignments 6 parallel to the second direction 5, E e) = 17 the load of the projection (a) and i = 1 E cab) = 17 the load of the projection (b). Similarly, in FIG. 2, cr is the term of j = 1 the projection (c) obtained by adding the numerical values of luminous intensity of the pixels along the alignment i of the n (c) = 15 alignments 8 parallel to the third direction 7, cr the end of the projection (d) obtained by adding the numerical values of luminous intensity of the pixels along the alignment j of 30 m (d) = 15 alignments 10 parallel to the fourth direction 9, Ee ) = 17 the i = 1 load of the projection (c) and E c ~ d = 17 the load of the projection (d). These four j = 1 projections are used to reconstruct the digital image by the direct generation method described above. In practice, the number of projections is higher and depends on the size of the digital image to be compressed.

Figures 3 to 7 illustrate an implementation of the compression method according to the invention applied to a digital image representing a face. Figure 3, the face is represented by a digital image of 512 by 512 pixels. 4, the digital image is reconstructed by retroprojections filtered from 180 projections 180 radiating projection directions in steps of 1 ° between 0 ° and 180 °. In this example, the compression ratio is equal to {(512) 2} / {180 x 512} or about 3. Figure 5, the digital image of Figure 3 is reconstructed by iterative retroprojections from 180 projections following 180 projection directions radiating in steps of 1 ° between 0 ° and 180 °. The compression ratio is again equal to {(512) 2} / {180 x 512} or about 3. The quality index depends on the number of iterations. In this example, it reaches 97 for 6 iterations. If 90 projections are made instead of 180, the compression ratio is doubled to 6 while maintaining the quality index at 97, provided however that the number of iterations is increased to 10. Figure 6, the same face is represented by a digital image of 2048 x 2048 pixels. Figure 7, the digital image is reconstructed by iterative retroprojections from 180 projections 180 radiating projection directions in steps of 1 ° between 0 ° and 180 °. Compared with FIGS. 4 and 5, the compression ratio is multiplied by 4, at {(2048) 2} / {180 × 2048}, ie about 12, and the quality index is close to 98. Here again, if the 90 projections are made instead of 180, the compression ratio is doubled to 24 while maintaining the quality index at 97, provided however that the number of iterations is increased to 10. Figures 8 and 9 illustrate the advantageous embodiment of the invention for compressing the digital image faster at constant compression ratio. FIG. 8 schematically shows a digital image 13 and 180 projection directions radiating around a centering pixel 17 while FIG. 9 schematically shows the digital image of FIG. 8 divided into four compartments 19-22. and four groups of 90 projection directions 23-26 radiating around a centering pixel 27-30 specific to each group. The compartments 19-22 are circles of the same radius whose inscribed squares are juxtaposed to mesh the digital image 13, the square inscribed in the circle being the square whose diagonal is equal to twice the radius. For the sake of clarity, only 18 and 9 projection directions are shown in Figures 8 and 9.

In the preceding example, the 180 projections in steps of 1 degree made it possible to compress the digital image 82 times. By choosing to mesh the digital image with 4 circles of the same radius so that the squares inscribed in the circles juxtapose and each have a quarter of the total number of pixels, ie (7407) 2, the compression ratio of 82 maintained by forming 4 groups of 90 projections each, by virtue of equality: 14814 x 180 = 4 x (7407 x 90). Since the computer system is configured with multi-core processors to perform the 90 rotations of each group at the same time, the calculation of the 4 x 90 projections of all 4 groups is 4 times faster than the calculation of the 180 projections. digital image treated as a whole. If one chooses a mesh in 16 circles, under the same conditions, the calculation of the terms of the 16 x 45 projections of all 16 groups is 16 times faster than the calculation of the terms of the 180 projections of the digital image treated as a whole. The compression method according to the invention applies to digital images acquired in radiology or scanning. The applicant has described a high resolution mammography device in the French patent application FR 2 947 658 and the international application WO2011 / 000718. A detector with a spatial resolution of 27 μm scans a detection area of 20cm x 20cm. The intensities detected are displayed on a computer screen in a digital image having a size of {(20 cm) 2 / (27 μm) 2} or 55 MB. This digital image is projected in 180 projections to be compressed at a compression ratio equal to {(20 cm) 2 / (27 μm) 2} / {180 x (20 cm) / (27 μm)} is about 41. The breast is examined under 36 radiological views taken around the axis of rotation of the device. Each can be compressed at the rate of 41. However, if a singularity is present in the breast, it will occupy only one region of each radiological view. In the following, it is assumed that the singularity extends over 6 cm x 6 cm. The compression method according to the invention advantageously makes it possible, on the one hand, to consider the images as a single image of 6 × 6, having a size equal to 62 × {(20 cm) 2 / (27 μm) 2}, ie 1 , 97 GB, and project this single image 180 projections to compress at a compression ratio equal to {62 x (20 cm) 2 / (27 microns) 2} / {180 x (6 x 20 cm) / ( 27 μm)} or 247. On the other hand, the region corresponding to the singularity is projected, for each of the 36 radiological views, into 54 projections whose terms are obtained by adding the numerical values of luminous intensity of the pixels of the region. along the parallel alignments to each of the 54 projection directions. The region corresponding to the singularity is thus compressed with a compression ratio equal to {(2 cm) 2 / (27 μm) 2} / {54 x (2 cm) / (27 μm)} ie 41, that is, ie the same compression rate as the radiological views considered individually. The region will therefore be reconstructed with the same quality index as the radiological views, while the compression ratio of the 36 radiological views will reach {62 x (20 cm) 2 / (27 1.tm) 2 + 62 x (6 cm) 2 / (27 11m) 2} / {180 x (6 × 20 cm) / (27 μm) + 54 × 36 × (6 cm) / (27 μm)} is 175. The compression method according to the invention allows in this case store several dozen years of black and white mammograms on a 4 TB SAN according to the following estimates: - size of two mammograms of 36 radiological views each compressed at the rate of 175: 25 MB - size of 30 daily examinations: 0.74 GB 20 - size of 30 x 160 examinations performed annually: 0.12 TB - number of years of storage on the SAN: 34. This number of years of storage is greater than a legal standard of 30 year medical data retention. The compression method according to the invention also applies to digital images acquired by surveillance cameras. At 25 frames per second, a surveillance camera takes 2.16 million digital images in 24 hours, each of which has a typical size of 621 x 425 pixels. For black-and-white digital images with one-byte pixel light intensity, one-day storage reaches 0.56 to 30 TB. Because the camera is used as a "fixed frame," digital images can be repeated identically for more or less important time periods, for example when the surveillance concerns a car park or a residential space. The compression method according to the invention provides for forming a checkerboard 31 of 5 x 5 digital images 33 for each sequence of one second. The checker formed at time t is compared to a reference checker formed at a previous instant and the projected digital image or images are those whose pixels have undergone a variation in light intensity compared to the reference checkerboard. To determine this or these digital images, the checkerboard 31 is projected into two projections along two privileged directions 35, 37 for example perpendicular, each projection being defined by the set of terms obtained by adding the numerical values of the luminous intensity of the pixels of the checker on the gray scale, along alignments 36, 38 parallel to the two preferred directions 35, 37 and the terms of the two projections are compared to the homologous terms of the reference checkerboard.

Figure 10 shows the checkerboard in which an event occurs in front of the surveillance camera. This event starts about 2/5 of the sequence of one second, that is to say 400 ms from the initial moment and affects the digital images 11 to 25 of the checkerboard, rows successively in line from left to the right. If the event lasts 10 seconds, it will affect the following checkers up to the 10th. Figure 10 shows the difference profile 39, 41 between the terms of the checkerboard considered and the terms of the reference checkerboard for each of the two directions. privileged projection 35, 37. The digital images 11 to 25 to be compressed are at the intersection of the pixel alignments along which the terms of the considered checkerboard and those of the reference checkerboard are different and form the non-flat segments of the two difference profiles. 39, 41. Images 1 to 10 are declared redundant with respect to the reference checkerboard and are deleted. This results in a double compression of the checkerboard: - by projecting the image or digital images whose pixels have undergone a variation of luminous intensity compared to the reference checkerboard, and - by removing the redundancy affecting the other images digital checkerboard.

The checkers that precede the beginning of the event occurring in front of the surveillance camera and those that follow the end of this event are advantageously removed for redundancy with the reference checker, by virtue of the compression method according to the invention previously described for a series of digital images and applied here to a series of digital image checkers. This results in a triple compression of a series of checkers: - by projection of the image or digital images of a checkerboard whose pixels have undergone a variation of luminous intensity compared to the reference checkerboard, - by deletion the redundancy affecting the other digital images of the checkerboard and, - by eliminating the redundancy affecting the checkers identical to the checkerboard of reference. It should be noted that the use of the two preferred projection directions can lead to a certain overdetermination of the digital images to be compressed. A digital image that has not undergone any variation in luminous intensity can be masked by a digital image having undergone a variation in luminous intensity and aligned with it according to one of the preferred projection directions. If necessary, it is planned to proceed with different storage of the digital images in the checkerboard 5 x 5 to reduce overdetermination. The compression method according to the invention, in its various embodiments, makes it possible to transfer or store digital images in the form of their projections, which are then used to reconstruct these images. This results in a reduction in the size of the digital images decreasing their transfer time or increasing their ease of storage, in proportion to the compression ratio. The computer system by means of which the compression method according to the invention is implemented is preferably a computer. The computer screen is used to display the digital image. The pixels that make up the mosaic of the digital image may correspond to the display pixels of the screen in a ratio of 1 to 1 or in a different ratio. The computer is loaded by a program with instructions that allow this computer to compress a digital image or one or more of the digital images of a checkerboard or a series of digital images according to the compression method of the computer. 'invention. It is also planned to implement the compression method according to the invention with the aid of a computer system comprising a computer and a passive compression box, this box comprising, FIG. 11, an input memory 43 for receiving a digital image or a checkerboard or a series of digital images and transfer this digital image, checkerboard or series of digital images to the computer and a mass memory of which a zone 45 contains the program indicated previously to be loaded on the computer. computer and another area 47 is provided for storing the digital image or digital images compressed by the computer. It is also planned to implement the compression method according to the invention with the aid of an active compression box comprising, FIG. 12, an input memory 43 for receiving a digital image or a checkerboard or a series of FIG. digital images, one or more processors 49 for executing the program indicated above and a mass memory of which a zone 45 contains this program and of which another zone 47 is provided for storing the digital image or the digital images thus compressed.

Claims (7)

REVENDICATIONS1. A method of compressing a checkerboard (31) of digital images (33), each digital image being in the form of a mosaic of pixels each having a luminous intensity on a scale of numerical values of gray or four light intensities respectively on four scales of numerical values, for gray and for three colors, the numerical value or values of luminous intensity of each pixel being coded by a computer system, characterized in that: - the checkerboard (31) is projected in two projections following two projection directions (35,37) for determining the digital image (s) of the checkerboard to be compressed, each projection being defined by one or four sets of terms obtained by adding the numerical values of luminous intensity of the checkerboard pixels to the gray scale or respectively on the four scales of gray and the three colors, along alignments (36,38) oriented by these two directions, - o n compares the terms of these two projections to terms of reference and determines the digital image or images whose pixels are at the intersection of the alignments of these two projections for which the terms are different from the terms of reference and, - we project the or each of the digital images thus determined in different projections according to different projection directions radiating around a centering pixel (11) and defined by one or four sets of terms obtained by adding the numerical values of luminous intensity of the pixels of the or each of the digital images, on the gray scale or respectively on the four gray scales and the three colors, along alignments oriented by the projection directions (15). 2. A compression method according to claim 1, characterized in that: - a reference checkerboard is projected in two projections according to the same two projection directions (35, 37) for determining the digital image or images of the checkerboard to be compressed, each projection being defined by one or four sets of reference terms obtained by adding the numerical values of the luminous intensity of the pixels of the reference checkerboard, on the gray scale orally on the four scales of gray and the three colors, along of alignments oriented by these two directions (35, 37) and, - redundancy is removed digital images of the checkerboard whose pixels are at the intersection of the alignments of the two projections of the checkerboard for which the terms are identical to the terms of reference . 3. A method of compression according to claim 2, characterized in that the method is implemented from several different storage of the same digital images in the checkerboard (31). A method of compressing a series of digital images, each digital image (1) of the series being in the form of a mosaic of pixels (12) each having a light intensity on a scale of numerical values of gray or four light intensities respectively on four scales of numerical values, for gray and for three colors, the numerical value or values of luminous intensity of each pixel being coded by a computer system, characterized in that: - each digital image is projected of the series in different projections according to different projection directions (3,5,7,9) radiating around a centering pixel (11), each projection being defined by one or four sets of terms obtained by adding the numerical values d luminous intensity of the pixels, on the gray scale or respectively on the four gray scales and the three colors, along alignments (4,6,8,10) oriented by the projection directions, the projection terms of two digital images of the series are compared for the same projection directions, and one of these two images is redundantly deleted when the compared terms are identical. A computer program comprising instructions which, when loaded into a computer system, enable the computer system to code the light intensity values of each pixel of a checkerboard or series of digital images. and compressing one or more of the digital images of this checkerboard or series according to the compression method defined by one of claims 1 to 4. A computer system comprising a computer and a compression box, which housing includes an input memory (43) for receiving a checkerboard or series of digital images and transferring that checkerboard or series of digital images to the computer and a mass memory of which a zone (45) contains a computer program according to claim 5 to be loaded onto the computer and a further zone (47) of which is provided for storing the digital image or digital images compressed by the computer. 'computer. A compression box comprising an input memory (43) for receiving a checkerboard or series of digital images, one or more processors (49) for executing a computer program according to claim 5 and a memory mass of which a zone (45) contains the computer program and of which another zone (49) is provided to store the digital images thus compressed.