WO2015091518A1

WO2015091518A1 - Method, device and system for image correction processing

Info

Publication number: WO2015091518A1
Application number: PCT/EP2014/078010
Authority: WO
Inventors: Vincent Martin; Christophe Latry; Philippe Kubik
Original assignee: Centre National D'etudes Spatiales
Priority date: 2013-12-20
Filing date: 2014-12-16
Publication date: 2015-06-25
Also published as: FR3015740A1; FR3015740B1

Abstract

The present invention relates to a method of correction processing of a digital image acquired by an image acquisition device comprising at least one matrix of elementary sensors, in relative rectilinear displacement with respect to an observed zone, an acquired image being represented by at least one matrix of pixels corresponding to an image signal acquired by the elementary sensors and sampled on an initial sampling grid, having an initial horizontal sampling interval and an initial vertical sampling interval, each pixel having an associated photometric barycentre (12, 24), the image acquisition device generating, in at least one direction of irregularity from among the horizontal direction and the vertical direction, a shift of the photometric barycentres with respect to the initial sampling grid. The method comprises an extraction of signals according to sampling sub-grids (GI, GP), a re-sampling filtering of each of the signals extracted and a re-sampling on a final sampling grid (GM). It also relates to an associated device and an associated system for image correction processing.

Description

Method, device and image correction processing system

The present invention relates to an image correction processing method acquired by an image acquisition device in rectilinear displacement relative to an observed area. It also relates to an image correction processing device and an associated image correction processing system.

The present invention is in the field of improving the quality of digital images, and finds applications in particular in the field of remote sensing images, satellite imagery and industrial inspection.

In the field of remote sensing, it is known to use, for capturing images, elementary sensor arrays, which may be two-dimensional or one-dimensional, also called arrays, which are attached to a satellite and which scroll to the right of the observed area, perpendicular to the direction of movement of the satellite. Thus, each elementary sensor scrolls in view of the observed landscape, each line of the acquired digital image corresponding to the acquisition made at a given moment. The resulting digital image is composed of one or more two-dimensional matrixes of samples called pixels, each sample having an associated radiometric value. In this acquisition mode, each column of a obtained digital image matrix is physically associated with an elementary sensor, each pixel of an image line corresponding to a signal acquisition of this elementary sensor.

Each image matrix corresponds to a sampling of a photometric signal on a two-dimensional sampling grid, of sampling pitch P, where P is the distance between theoretical photometric centers of gravity of two successive pixels in each direction.

This method of acquisition by rectilinear displacement of a matrix of sensors is known as scanning acquisition, or "push-broom" in English. The image acquisition device is relatively straight relative to the observed area.

Elementary sensors integrate the electromagnetic radiation of a surface observed in a given spectral band and transform it into electric charges. Each sensor is characterized by a spread function called PSF (for "point spread function") which has a module in the Fourier domain denoted FTM (for Modulation Transfer Function ").

In case of saturation, an accumulation of excess electrical charges is observed at the level of the elementary sensors. In order to evacuate the excess charges, it has been proposed to use anti-blooming drain devices, positioned laterally with respect to the elementary sensors, alternately, both elementary sensors. of a bar. Such a side anti-glare device placement system is known as an alternate side glare system.

Thus, the excess charges are eliminated for each pixel of the acquired image, each pixel being associated with an elementary sensor.

However, in the absence of saturation, the presence of such an alternating lateral anti-dazzle system has an impact on the accumulation of electric charges in a potential well corresponding to the elementary sensor, which has the effect of shifting the photometric center of gravity of each pixel relative to the ideal sampling grid.

Indeed, if each lateral anti-dazzle device is implanted on the incident face of the photons on the elementary sensors, such an anti-glare device introduces a partial or total occultation of the photosensitive surface. If each lateral anti-glare device is implanted on the face opposite to the incident face of the photons, it disrupts the electric fields modifying the distribution of the electron fluxes created by the incident photons.

Apart from lateral anti-glare devices, in type sensors

CMOS (for "complementary metal-oxide semi-conductor"), various electronic functions are located near the elementary sensors, for example the implementation of electric bus between two sensors. In this case also, photosensitive surfaces are obscured or shifted, and therefore the photometric center of gravity of at least some pixels is shifted relative to an ideal sampling grid.

The aim of the invention is to improve the quality of the digital images acquired under these acquisition conditions resulting in an offset of the photometric center of gravity of at least part of the acquired pixels with respect to an initial sampling grid which is the grid ideal sampling.

For this purpose, the invention proposes, according to a first aspect, a digital image correction processing method acquired by an image acquisition device comprising at least one elementary sensor matrix, in relative displacement with respect to a zone observed, an acquired image being represented by at least one pixel matrix corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid, having a sampling step initial horizontal and an initial vertical sampling step, each pixel having an associated photometric barycenter, the image acquisition device generating, in at least one direction of irregularity among the horizontal direction and the vertical direction, a shift of the centers of gravity photometrically with respect to the initial sampling grid. The method according to the invention comprises the steps, applied to at least a portion of a said pixel matrix, of:

extracting said portion of a first subset of pixels corresponding to the sampled image signal on a first sampling sub-grid and a second subset of pixels corresponding to the sampled image signal on a second sub-set of pixels; sampling sub-grid, each of said first and second sampling sub-grids having a sampling interval in the direction of irregularity equal to twice the corresponding sampling interval of the initial sampling grid;

obtaining a first image signal from said first subset of pixels and a second image signal from said second subset of pixels,

applying a first resampling filter on the first image signal and a second resampling filter on the second image signal,

- re-sampling the filtered image signals to obtain a resampled image signal on a final sampling grid having a horizontal sampling step equal to the initial horizontal sampling step and a vertical sampling step equal to the step initial vertical sampling,

said first and second resampling filters being determined as a function of said offset of the photometric centers of gravity relative to the initial sampling grid and a desired positioning of the final resampling grid relative to the initial sampling grid .

Advantageously, the correction processing method according to the invention makes it possible to improve the quality of the digital images acquired by resampling from regular sub-grids, making it possible to compensate for the irregularities induced by the presence of an acquisition device. of image, in relative displacement with respect to an observed zone, inducing a photometric barycenter offset of the pixels.

The method according to the invention may have one or more of the following characteristics, taken independently or in any technically feasible combination.

The acquired image is acquired by a one-dimensional matrix of elementary sensors, and said first subset of pixels comprises all the pixels of the pixel matrix portion belonging to the even-numbered columns or lines and said second subset of pixels comprises all the pixels of the processed image portion belonging to the odd-numbered columns or lines of said processed matrix portion.

The image acquisition device induces areas of occultation of the acquisition of charges by the elementary sensors to which they are associated, arranged every two pixels in the direction of irregularity, the determination of said first and second filters being function the width of a said occultation zone.

The first and second filters are defined from a resampling filter G by the following formulas:

G _l (x) = G (x + d _y )

And

G ₂ (x) = G (d ₂ + 1 - x).

where d1 is a distance, in the direction of irregularity, between the photometric centers of gravity of the first sampling sub-grid and the final sampling grid, and d2 is a distance, in the direction of irregularity, between the centroids of the second sub-sampling grid and the final sampling grid.

The determination of the first and second filters is performed from a resampling filter defined by:

where sine is the cardinal sine function, cotan is the co-tangent function, and a is equal to the half-width of a so-called occultation zone, x and a being expressed in pixels of the initial sampling grid.

The determination of the first and second filters comprises a determination of the support length of each of said filters and the application of an apodization window.

Determining said first and second filters comprises normalizing the coefficients of each of said first and second filters.

Resampling on the final sampling grid comprises pixel-by-pixel addition of the filtered image signals to obtain a pixel value of the resampled image signal.

Obtaining a first image signal consists in inserting a column or row of zeros between two columns or consecutive lines of said first subset of pixels, and obtaining a second image signal consists in inserting a column or row of zeros between two columns or consecutive lines of said second subset of pixels. According to a second aspect, a digital image correction processing device acquired by an image acquisition device in rectilinear displacement relative to an observed area, comprising at least one elementary sensor matrix, an acquired image being represented by at least one pixel array corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid, having an initial horizontal sampling step and an initial vertical sampling step, each pixel having an associated photometric barycenter, the image acquisition device generating, in at least one direction of irregularity among the horizontal direction and the vertical direction, an offset of the photometric centers of gravity with respect to the initial sampling grid.

This device comprises means, applied to at least a portion of a said pixel matrix, of:

obtaining a first image signal from said first subset of pixels and a second image signal from said second subset of pixels;

applying a first resampling filter to the first image signal and a second resampling filter to the second image signal; resampling the filtered image signals to obtain a signal image resampled on a final sampling grid having a horizontal sampling interval equal to the initial horizontal sampling step and a vertical sampling step equal to the initial vertical sampling step.

This device further comprises means for determining said first and second resampling filter as a function of said offset of the photometric centers of gravity with respect to the initial sampling grid and a desired positioning of the final resampling grid with respect to to the initial sampling grid.

The digital image correction processing device has advantages similar to the advantages of the associated method.

This device comprises means for implementing the digital image correction processing method as briefly described above. According to a third aspect, the invention relates to an image correction processing system, this system comprising an image acquisition device in rectilinear displacement relative to an observed zone, comprising at least one matrix of elementary sensors, capable of acquiring a digital image represented by at least one pixel matrix corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid, having an initial horizontal sampling step and a vertical sampling step initial, each pixel having a photometric barycenter associated, the image acquisition device generating, in at least one direction of irregularity among the horizontal direction and the vertical direction, an offset of the photometric centers of gravity relative to the sampling grid initial, and a correction processing device as briefly defined above .

The digital image correction processing system has advantages similar to the advantages of the digital image correction processing method briefly described above.

In one embodiment, each elementary sensor is capable of converting electromagnetic radiation into electric charges, and the image acquisition device comprises anti-glare devices capable of evacuating excess electrical charges, anti-glare devices. dazzle being positioned laterally in a given direction both elementary sensors.

According to a particular characteristic, each elementary sensor has a photosensitive face oriented towards the observed zone, and each anti-glare device is positioned on the photosensitive face of an elementary sensor.

According to another particular characteristic, each elementary sensor has a photosensitive face oriented towards the observed zone, and each anti-glare device is positioned on the face opposite to the photosensitive face of an elementary sensor.

In one embodiment, the image acquisition device is a CMOS device.

Other features and advantages of the invention will emerge from the description given below, by way of indication and in no way limiting, with reference to the appended figures, among which:

FIG. 1 diagrammatically illustrates the principle of image acquisition by a matrix of elementary sensors in the case of satellite imagery;

FIG. 2 illustrates an ideal sampling grid corresponding to an image matrix, in the absence of use of an alternating lateral anti-dazzle system; FIG. 3 represents a matrix of pixels corresponding to an irregular sampling grid when using an alternating lateral anti-dazzle system,

FIG. 4 is a diagram illustrating the functional blocks of an image processing chain according to one embodiment;

FIG. 5 is a block diagram of an image correction processing method according to one embodiment of the invention;

FIGS. 6 and 7 schematically represent matrices of pixels corresponding to subgrid sampling;

- Figure 8 schematically shows a set of functional blocks of a device adapted to implement the invention.

The invention will be described hereinafter more particularly in the field of satellite image acquisition, by one-dimensional matrices or elementary sensor strips, in the mode of acquisition by the so-called "push-broom" mode, in wherein each column of the obtained digital image is physically associated with a sensor of the one-dimensional matrix of elementary sensors, and wherein the sampling grid is regular, having a horizontal sampling pitch equal to the vertical sampling step.

It should however be noted that the concept of the invention applies in other cases of application, in particular in the case of the acquisition of images by a one-dimensional or two-dimensional matrix of elementary sensors, according to other modes. scanning a landscape or if the vertical sampling step is different from the horizontal sampling step. For example, the invention also applies when the matrix of elementary sensors is fixed, and the acquired landscape scrolls to the right of the matrix of elementary sensors.

The context of satellite digital image acquisition is schematically illustrated in Figure 1.

A not shown satellite having a displacement trajectory T comprises an image acquisition device 2, which is a one-dimensional matrix of elementary sensors 4, which is able to acquire images via an optical system 6. Each elementary sensor 4 realizes the acquisition of a portion of the observed landscape.

The observed landscape passes in front of each elementary sensor 4 which integrates the received electromagnetic flux during a exposure time and transforms it into a proportional electric charge. The charges obtained are converted into electrical voltages proportional to the observed electromagnetic flux. For example, CCD ("charged-device") sensors are used as elementary sensors.

A digital image matrix, composed of image samples or pixels 8, each pixel having an associated radiometry value, also called associated intensity value, is obtained from the previously obtained electrical voltages.

The digital image matrix I is composed of lines of pixels Li and column of pixels Ci, each image line corresponding to the acquisition of the landscape at a given moment. In the example of FIG. 1, each column of pixels is associated with an elementary sensor 4.

The matrix I is formed on a theoretical sampling grid, called the initial sampling grid, illustrated in FIG.

FIG. 2 illustrates a sampling grid 10 formed of theoretical photometric centers 12 (represented by a symbol '+' in the figure) of each acquisition surface 14 of electromagnetic radiation by an elementary sensor 4.

The horizontal sampling pitch of the sampling grid 10, corresponding to the distance in the horizontal direction between two successive centroid centers 12, is noted P. The sampling pitch P is 1 pixel in each direction in this example.

More generally, for a sampling grid, a vertical sampling step and a horizontal sampling step are defined.

For a regular sampling grid, the horizontal sampling step is equal to the vertical sampling step.

In the case where a device for lateral anti-glare, called AEL device, is positioned both sensors, on the photosensitive face of the sensors, this device obscures part of the electromagnetic radiation received by the elementary sensors 4.

In other implementation modes, for example when the AEL device is positioned both sensors, on the face opposite to the photosensitive face of the sensors, an effect equivalent to a direct concealment of the collected electron flux is found.

FIG. 3 illustrates an irregular sampling grid 20 associated with a bar of elementary sensors 4 when an alternating AEL device is positioned every n = 2 elementary sensors 4. In the example of FIG. 3, a six-column matrix portion denoted by C ₀ to C ₅ is obtained by a bar of six sensors.

As illustrated in this figure, the positioning of the AEL devices causes the appearance of occultation bands Ζ to Z ₃ , each of these bands corresponding to the positioning of an AEL device between two neighboring sensors, both sensors.

The width of the occultation zones induced by the AEL devices is constant and predetermined.

In a numerical example, when the acquisition surfaces 14 have a dimension of 15 μηι by 15 μηι (micrometers), the width of an AEL band is of the order of 5 μηι corresponding to = = M3 of pixel of a regular sampling grid of not equal to 1 pixel.

Thus, each acquisition surface 22 corresponding to a given elementary sensor is reduced in width, and the sampling grid 20 has a horizontal irregularity.

The centroids 12 corresponding to a regular sampling grid 10 are denoted by the symbol '+' in FIG.

The actual photometric barycentres 24 of each acquisition surface 24 are offset relative to the barycentres 12, differently depending on the parity of the index of the column. In the example of FIG. 3, the photometric centers of gravity of columns of even index C _2j , denoted by the symbol 'ο', are shifted by a distance il A to the left of the theoretical center of gravity 12, whereas the photometric centers of gravity columns of even index C _{2j +} i, denoted by the symbol 'x', are shifted by a distance il A to the right of the theoretical center of gravity 12.

The set of barycenters of the even indexed columns denoted by 'o' form a first GP sample sub-grid, called the even grid, which is regular in each direction, of 2P horizontal sampling steps and no sampling steps. vertical P.

The set of barycenters of the odd-numbered columns denoted by 'x' form a second sampling subgreen G1, called the odd grid, which is regular in each direction, with no horizontal sample sampling and no sampling. vertical P.

However, the sampling grid 20 formed by the set of photometric centers of gravity 24 is irregular, the direction of irregularity being the horizontal direction.

The invention proposes to modify the radiometric treatment of the acquired image by an acquisition device inducing an offset of the photometric barycenters of the pixels in at least one of the directions, in order to improve the quality of the resulting digital image. FIG. 4 shows a block diagram of a processing chain 30 for acquisition of images in an acquisition mode as described above with reference to FIG. 1, forming part of a correction processing system of FIG. images according to the invention.

The image processing chain 30 comprises a unit 32 for amplifying the electrical signal S obtained at the output of the sensors 4 connected at the input of an analog-digital encoder 34.

A radiometric correction unit 36 adapted to implement a correction processing method according to the invention is connected to the output of the unit 34 and to the input of a restoration processing unit 38, such as deconvolution or denoising.

The radiometric correction processing 36 is applied before any resampling of the digital image necessary for an application of the acquired digital image.

A flow chart of the main steps of an image correction processing method according to the invention is shown in FIG.

In this embodiment, the case of an offset of the photometric centers of gravity in adjacent columns, as illustrated in FIG. 3, is considered.

The ideal continuous image signal is called f, which is independent of any representation of an image on a sampling grid for a digital representation.

For each line of a digital image portion I to be processed, the sampled values f (i, j), i = 0, ..., M; j = 0,1, 2 ...., N, where M is the maximum line index and N is the maximum column index of the image portion to be processed. The ideal sampling grid, called the initial sampling grid, has a horizontal sampling pitch P.

In a first step 40, a first subset of pixels comprising even-numbered columns is extracted from the complete sampling grid of the image portion to be processed. The first subset of pixels corresponds to a first image sub-signal, denoted f1, sampled on the first GP sample sub-grid, of horizontal sampling pitch P '= 2P,

The first image sub-signal f1 is completed by zeros in the next step 42 to obtain a first image signal F1 of horizontal sampling pitch equal to the initial sampling pitch P.

The following formula applies: F _l (i, 2j) = f (i, 2j) and F _l (i, 2j + 1) = 0

As illustrated in FIG. 6, a matrix of pixels l _pa i _r , or first set of pixels, of the same size as the portion of digital image to be processed is obtained. The first one pixel matrix corresponds to the first F1 image signal sampled on a sampling grid having the same sampling pitch as the initial grid in both horizontal and vertical directions.

In a step 44, a second image sub-signal denoted f2, corresponding to the columns of odd index, sampled on the second odd sampling sub-grid G1, of sampling pitch P '= 2P, is extracted from the complete sampling grid of the image portion to be processed.

The second image sub-signal f2 is completed by zeros in the next step 46 to obtain a second horizontal sample pitch image signal F2 equal to the initial sampling pitch P.

The following formula applies: F ₂ (i, 2j + 1) = f (i, 2j + 1) and F ₂ (i, 2j) = 0.

As illustrated in FIG. 7, an _odd pixel matrix, or second set of pixels, of the same size as the digital image portion to be processed is obtained. The second pixel array corresponds to the second image signal F2 sampled on a sampling grid having the same sampling pitch as the initial grid in both the horizontal and vertical directions.

After step 46, a step 50 consists in obtaining two resampling filters G1, G2. These filters are each defined by a set of coefficients, a filter support size, apodization weighting values.

The resampling filters G1, G2 are normalized: the sum of the filter coefficients on their finite size support is equal to 1.

In the preferred embodiment, the following mathematical formula is used to provide the coefficients of a resampling filter G:

G (x) = sinc (pc) cotanf π - ^ ^{+ a} ] ^■ - ^■ sinc ² | π · - j where sine is the sine function l 2) 2 l 2)

cardinal, cotan is the co-tangent function, and a = £ / 2 is the half-width of the occultation zones, as shown in Figure 3.

Advantageously, the function G (x) is the mathematical function that makes it possible to reconstruct, from an offset sampling grid (corresponding to the symbols 'o' and 'x'), the image that would have been acquired by a device regular acquisition without shifts of the photometric center of gravity (corresponding directly to the symbols '+') with the same level of sharpness, under certain assumptions concerning the acquisition device: image and unrestricted media filters, FTM identical between the two under - sampling grids, perfect regularity of subgrid sampling, absence of spectrum folding at the final step P, absence of instrumental acquisition noise. The advantageous properties of the function G (x) are largely preserved when these hypotheses are not perfectly respected, in the case of a real acquisition device.

In this embodiment, the filtering is performed per line of pixels of each image signal, inasmuch as the photometric centers of gravity are evenly spaced in the vertical direction, so no radiometric correction is needed at the column level to improve the quality. image.

A first resampling filter G1 (x) and a second resampling filter G2 (x) are respectively defined for filtering the first image signal F1 and the second image signal F2.

In the preferred embodiment, the first and second resampling filters are defined as follows:

This definition by the formulas (Eq 1) and (Eq 2) of the resampling filters makes it possible to perform the registration of each of the image signals F1 and F2 on a final sampling grid which is a median grid, noted GM in FIG. 3, whose photometric centers of gravity are located at equal distances from the barycentres of the respective even and odd subgrids.

Advantageously, the contribution of the first and second image signals is balanced when the resampling filters are defined by the formulas (Eq 1) and

(Eq 2).

The coefficients of each of the filters G1 and G2 are obtained on a finite support, preferably between 16 and 256, and typically equal to 128.

An apodization or windowing is applied, consisting of multiplying the filters G1 and G2, on their support, by a symmetrical function (centered in the middle of the support) equal to 1 in the center and soft decay towards the edges. Classically, the so-called Hann, Hamming, Gaussian or raised cosine functions are used.

In a first resampling step 50, the first filter G1 is applied to the first signal F1, line by line, to obtain a first filtered image signal, denoted F _l * G _l , where the symbol ^* represents the product of convolution. Then, during a second resampling step 52, the second filter G2 is applied to the signal F2, line by line, to obtain a second filtered image signal, denoted F ₂ * G ₂ .

Finally, during a summing step 54, the two filtered signals are added, pixel by pixel, to obtain a resampled final signal F, which corresponds to the initial image signal f reconstituted on a final regular GM step grid. P sampling.

F a, j) = F, a, j) * G, u) + F ₂ a, j) * G ₂ U)

It should be noted that according to one variant, the steps 50, 52 and 54 are carried out simultaneously.

According to a variant of the embodiment described above, during step 50, the first and second resampling filters are defined as follows:

G _l (x) = G (x) (Eq 3)

And

G ₂ (x) = G {a + 1 - x) (Eq 4)

With the definition given by the formulas (Eq 3) and (Eq 4) above, the resampling gate chosen is a horizontal sampling pitch grid P, positioned on the photometric centers of gravity 24 noted 'o' on the In this embodiment, the photometric centers of brightness of even-numbered columns are not modified, only the photometric centers of the columns of odd index are modified.

According to another variant, during step 50, the first and second resampling filters are defined as follows:

G _l (x) = G {a + 1-x) (Eq 5)

And

G ₂ (x) = G (x) (Eq 6)

With the above definition, the resampling gate chosen the resampling gate chosen is a horizontal sampling pitch grid P, positioned on the photometric centers 47 noted 'x' in FIG. 3. In this embodiment , the photometric centers of the columns of odd index are not modified, only the photometric centers of the columns of even index are modified.

In a more general manner, the resampling filters G1 and G2 are defined from the resampling filter formula G and the half-width has occultation zones, for a first offset d1 relative to photometric centers of the first sampling sub-grid and an offset d2 with respect to photometric centers of gravity of the second sampling sub-grid. The following condition applies d _l + d ₂ = a to compensate for the half-width a of the masking areas.

The offsets d1 and d2 define the positioning of the final sampling grid:

G _l (x) = G {x + d _l ) (Eq 7)

And

G ₂ (x) = G {d ₂ + 1 - x) (Eq 8) The correction processing methods of the invention are implemented by a programmable computer-type device, workstation, whose main functional blocks are illustrated in Figure 8.

A programmable device 60 capable of implementing the methods of the invention comprises a central processing unit 68, or CPU, able to execute computer program instructions when the device 60 is powered up. In one embodiment, a multiprocessor CPU is used to perform parallel computations. The device 60 also comprises information storage means 70, for example registers, capable of storing executable code instructions allowing the implementation of programs comprising code instructions able to implement the correction processing method. of images according to the invention.

The device 60 comprises control means 64 for updating parameters and receiving commands from an operator. When the programmable device 60 is an onboard device, the control means 64 comprise a telecommunication device for receiving remote commands and parameter values.

Alternatively and optionally, the control means 64 are means for inputting commands from an operator, for example a keyboard.

Optionally, the programmable device 60 comprises a screen 62 and additional pointing means 66, such as a mouse.

The various functional blocks of the device 60 described above are connected via a communication bus 72.

An image correction processing system according to the invention comprises an image acquisition device, such as the image acquisition device 2 of FIG. 1, generating an offset of the photometric centers of gravity in at least one direction, and a correction processing device implemented by a programmable device as described above with reference to FIG. 8.

According to a first embodiment, the image acquisition device is a matrix of elementary sensors carried by a satellite and provided with an AEL device positioned both sensors on the photosensitive face of the sensors.

According to a second embodiment, the image acquisition device is a matrix of elementary sensors carried by a satellite and provided with an AEL device positioned every two sensors, on the face opposite to the photosensitive face of the sensors, having an effect equivalent to a direct concealment of the electron flux collected is found.

According to a third embodiment, the image acquisition device is a matrix of CMOS sensors comprising an electronic function, for example an electric bus, implanted both sensors in a given direction, resulting in an occultation of the photosensitive surfaces and in a shift of the photometric centers of gravity in the given direction.

The invention applies, more generally, in any system comprising an image acquisition device making a concealment of the incident surface of the electromagnetic radiation or a disturbance of the electron flow of this electromagnetic radiation.

The principle of the invention has been described above in an example in which an irregularity of the photometric barycenters in the horizontal direction is present.

Similarly, the invention applies when the direction of irregularity of the photometric centers of gravity is the vertical direction.

In addition, the method of the invention is also separably applicable in the horizontal direction on the one hand and in the vertical direction on the other hand, to compensate for vertical and horizontal irregularities, which occur in particular when using the an acquisition device comprising a two-dimensional matrix of elementary sensors.

Claims

1 .- Digital image correction processing method acquired by an image acquisition device comprising at least one elementary sensor matrix, in relative displacement with respect to an observed zone,

an acquired image (I) being represented by at least one pixel array corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid (10), having an initial horizontal sampling step and a step vertical sampling, each pixel having a photometric barycentre (12) associated, the image acquisition device generating, in at least one direction of irregularity of the horizontal direction and the vertical direction, an offset of the photometric centers of gravity ( 24) with respect to the initial sampling grid (10), characterized in that it comprises the steps, applied to at least a portion of a said pixel matrix, of:

extraction (40, 44) of said portion of a first subset of pixels corresponding to the sampled image signal on a first sampling sub-grid (GP) and a second subset of pixels corresponding to the image signal sampled on a second sampling sub-grid (Gl), each of said first and second sampling sub-grids (GP, Gl) having a sampling interval in the direction of irregularity equal to twice the no corresponding sampling from the initial sampling grid,

obtaining (42, 46) a first image signal from said first subset of pixels and a second image signal from said second subset of pixels,

- applying (50, 52, 54) a first resampling filter (G1) on the first image signal and a second resampling filter (G2) on the second image signal,

- resampling (50, 52, 54) of the filtered image signals to obtain a resampled image signal on a final sampling grid (GM) having a horizontal sampling step equal to the sampling step initial horizontal and a vertical sampling step equal to the initial vertical sampling

said first (G1) and second (G2) resampling filter being determined (48) as a function of said offset of the photometric centers of gravity relative to the initial sampling grid and a desired positioning of the final resampling grid compared to the initial sampling grid.

2. A method according to claim 1, characterized in that said acquired image is acquired by a one-dimensional matrix of elementary sensors, and in that said first subset of pixels comprises all the pixels of the pixel matrix portion belonging to the even-numbered columns or lines and said second subset of pixels includes all the pixels of the processed image portion belonging to the odd-numbered columns or lines of said processed matrix portion.

3. A method according to any one of claims 1 or 2, characterized in that the image acquisition device induces zones of occultation (ZZ ₂ , Z ₃ ) of the acquisition of charges by the elementary sensors to which they are associated, arranged every two pixels in the direction of irregularity, and in that the determination (48) of said first and second filter is a function of the width of a said occultation zone.

4. Method according to claim 3, characterized in that said first and second filters are defined from a resampling filter G by the following formulas:

G _l (x) = G (x + d _l )

And

G ₂ (x) = G (d ₂ + 1 - x).

5. A method according to claim 4, characterized in that the determination of the first and second filter is performed from a resampling filter defined by:

G (x) = sinc (^ zx) - K · cotanf l κ · 2)] · - 2 · sine ² fl · -] where sine is the sine function

2)

cardinal, cotan is the co-tangent function, and a is equal to the half-width of a so-called occultation zone, x and a being expressed in pixels of the initial sampling grid.

6. Method according to any one of claims 1 to 5, characterized in that the determination (48) of said first and second filter comprises a determination. of support length of each of said filters and the application of an apodization window.

7. - Method according to claim 6, characterized in that the determination of said first and second filter comprises a normalization of the coefficients of each of said first and second filter.

8. - Method according to any one of the preceding claims, characterized in that the resampling on the final sampling grid (GM) comprises a pixel-by-pixel addition (54) of the filtered image signals to obtain a value pixel of the resampled image signal.

9. - Method according to any one of the preceding claims, characterized in that obtaining a first image signal consists in inserting a column or row of zeros between two columns or consecutive lines of said first subset of pixels. , and in that obtaining a second image signal consists in inserting a column or row of zeros between two columns or consecutive lines of said second subset of pixels.

10. A digital image correction processing device acquired by an image acquisition device in rectilinear displacement relative to an observed area, comprising at least one matrix of elementary sensors, an acquired image being represented by at least one a matrix of pixels corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid, having an initial horizontal sampling step and an initial vertical sampling step, each pixel having an associated photometric barycenter, the image acquisition device generating, in at least one direction of irregularity among the horizontal direction and the vertical direction, an offset of the photometric centers of gravity with respect to the initial sampling grid, characterized in that it comprises means, applied to at least a portion of a said pixel matrix, of:

extracting said portion of a first subset of pixels corresponding to the sampled image signal on a first sampling sub-grid and a second subset of pixels corresponding to the sampled image signal on a second sub-set of pixels; sampling sub-grid, each of said first and second sampling sub-grids having a sampling step according to the direction irregularity equal to twice the corresponding sampling interval of the initial sampling grid,

obtaining a first image signal from said first subset of pixels and a second image signal from said second subset of pixels; applying a first resampling filter on said first subset of pixels; the first image signal and a second resampling filter on the second image signal,

the device further comprising means for determining said first and second resampling filter as a function of said offset of the photometric centers of gravity with respect to the initial sampling grid and a desired positioning of the final resampling grid with respect to to the initial sampling grid.

1 1. - Image correction processing system characterized in that it comprises an image acquisition device in rectilinear displacement relative to an observed area, comprising at least one elementary sensor matrix, capable of acquiring a digital image represented by at least one pixel array corresponding to an image signal acquired by the elementary sensors sampled on an initial sampling grid, having an initial horizontal sampling step and an initial vertical sampling step, each pixel having a associated photometric barycenter, the image acquisition device generating, in at least one direction of irregularity among the horizontal direction and the vertical direction, an offset of the photometric centers of gravity with respect to the initial sampling grid, and a device for correction process according to claim 10.

12. - image correction processing system according to claim 1 1, characterized in that each elementary sensor is adapted to convert electromagnetic radiation into electrical charges, and in that the image acquisition device comprises devices anti-dazzle able to evacuate excess electrical charges, the anti-dazzle devices being positioned laterally in a given direction both elementary sensors.

13. - image correction processing system according to claim 12, characterized in that each elementary sensor has a photosensitive face oriented towards the observed area, and in that each anti-glare device is positioned on the photosensitive face of an elementary sensor.

14. - image correction processing system according to claim 12, characterized in that each elementary sensor has a photosensitive face oriented towards the observed area, and in that each anti-glare device is positioned on the opposite side to the photosensitive face of an elementary sensor.

15. - image correction processing system according to claim 1 1, characterized in that the image acquisition device is a CMOS type device.