Title of Invention	A METHOD OF ACQUIRING AND PROCESSING SATELLITE OBSERVATION IMAGE AND CAMERA INSTRUMENT USED THEREFOR
Abstract	The invention concerns a method of acquiring and processing an earth satellite observation image using at least an array or matrix of load transfer sensors moving opposite a zone under observation, which consists in acquiring a plurality of pixels according to a given sampling. The invention is characterized in that said sampling network includes "e" sub-systems corresponding to a sampling verifying the Shannon theory, where "e" is an integer or rational number greater than 1, and in using a process for interpolating a noise-corrected image whereof the resolution corresponds to that of a sampling sub-system from the initially acquired image pixels.

Title of Invention

A METHOD OF ACQUIRING AND PROCESSING SATELLITE OBSERVATION IMAGE AND CAMERA INSTRUMENT USED THEREFOR

Abstract

The invention concerns a method of acquiring and processing an earth satellite observation image using at least an array or matrix of load transfer sensors moving opposite a zone under observation, which consists in acquiring a plurality of pixels according to a given sampling. The invention is characterized in that said sampling network includes "e" sub-systems corresponding to a sampling verifying the Shannon theory, where "e" is an integer or rational number greater than 1, and in using a process for interpolating a noise-corrected image whereof the resolution corresponds to that of a sampling sub-system from the initially acquired image pixels.

Full Text	The present invention relates to acquiring a satellite observation image of the earth by push-broom scanning using one or more strips (or matrix(ces)) of detectors of the charge coupled device (CCD) type travelling relative to the observed zone. FIELD OF THE INVENTION The principle of push-broom type scanning is shown in Figure 1 for the case of a single strip 1 of detectors. As the satellite carrying the strip 1 travels, the strip observes successive lines L1, L2, . . . , Ln perpendicular to the travel direction (arrow D). At each instant, an optical system 2 of the instrumentation forms the image of a line of landscape on the line of detectors, with the strip 1 being placed in the focal plane of the optical system 2, perpendicularly to the speed vector of the satellite. The landscape scans over each detector which integrates light flux over an exposure time and transforms it into a proportional electric charge. Figure 2 shows a conventional system for processing images taken in this way. That processing system comprises, in outline: a unit 3 for processing and amplifying the output from the detectors of the strip 1; an analog-to-digital encoder 4 that receives the signal output by the unit 3 ,-transmitter means 5 for taking the digital images picked up in this way and transmitting them from the satellite to the ground; and a unit 6 on the ground for reconstituting the images. The unit 3 includes, in particular, a shift register into which the information as integrated and stored in charge form in each detector of the strip 1 is transferred at the end of an exposure time. Thereafter, the register transfers the charges in the form of 2 electrons, and the charges are converted into a succession of voltages that are proportional to the received and integrated light fluxes. The unit 6 on the ground reconstructs the images, in particular by implementing deconvolution processing to compensate for instrument defects, and where appropriate interpolation processing to reconstitute certain pixels within the image. STATE OF THE ART AND OBJECT OF THE INVENTION It is known that the finer the resolution of an image, the more the signal-to-noise (SNR) tends to diminish, in particular because of the lack of light flux, so that images cease to be acceptable. Unfortunately, presently known solutions for mitigating that drawback are not satisfactory. In particular, one possible solution consists in increasing the dimensions of the instrument, and in particular the diameter of the pupil of the telescope. However that solution is expensive. Another solution consists in controlling the attitude of the satellite so as to slow down its rate of scanning. That technique allows light flux to be integrated for longer, but it implies a loss of continuity in the track of the satellite, and consequently leads to a loss of data. A third solution consists in using special time delay integration (TDI) detectors. These detectors are constituted as matrices in which the rows are shifted electronically so as to compensate for the speed of the satellite. Nevertheless, such a system is complex to embody and implement. Furthermore, it does not enable sampling to be performed that is adapted to the modulation transfer function so as to ensure that spectrum folding is negligible and acquisition data rate is optimized. 3 An object of the invention is thus to provide an acquisition method of this type in which instrument noise is reduced and which does not suffer from the drawbacks of the above-mentioned solutions. THEORETICAL BACKGROUND It is known that image acquisition by push-broom scanning can be modelled linearly by the formulation: where: designates the convolution operation; is the landscape whose image is to be acquired; is the impulse response of the instrument; is the noise superposed on the filtered landscape; is the raw image; and is a two-dimensional Dirac comb which means that the continuous image has been made discrete. The image is digital and encoded on a limited number of bits. A spectral representation is often preferred. It is represented by the Fourier transform (FT) of the raw image, i.e.: where the hats indicate that these are Fourier transforms, and where the abbreviation MTF stands for modulation transfer function, which is the Fourier transform of the impulse response. In the spectral representation, the convolution operation is equivalent to a multiplicative operation. The parameters that are mainly involved in the above equations are described in greater detail below. Modulation transfer function or MTF The MTF is the attenuation factor of spatial frequencies. The higher the frequency the lower the MTF. The geometrical location where the MTF becomes zero is the boundary of the instrumental cutoff. The instrument 4 spectral medium is the low frequency domain as limited by the instrumental cutoff boundary. The MTF of an instrument is the product of the MTF of its optical system multiplied by the MTF of its detector and by the MTF of its displacement in the travel direction. Noise Instrumental noise is always present. A simplified approach consists in considering that it is characterized by the mean SNR (generally known once the mission of the satellite has been defined), and that, compared with other noise, Poisson type photon noise is preponderant at the mean luminance. With such a distribution, the standard deviation of the noise varies with the square root of the number of photons N picked up by the detector. Consequently Sampling array-Present satellites generate a square orthogonal array (speed and strip are in orthogonal directions). The sampling step size in the speed direction on the ground is ?t = v.te (where v is the speed of the satellite and te is the sampling time), and the sampling step size in the strip direction is equal to the distance between two adjacent elementary CCDs. Consequently, using the arrays implemented in present satellites, the two pitches are equal and the projections of ground pixels do not overlap (the ground is scanned once only). However, with acquisition as performed by current satellites, the sampling frequency is equal to the cutoff frequency so Shannon's condition is not satisfied (te 5 To mitigate that drawback, proposals have recently-been made, in particular in French patent FR 2 678 460 in the name of the Applicant, to use a technique whereby two offset strips make it possible to perform oversampling so as to limit spectrum folding. In its patent application WO 97/05451, the Applicant has also proposed techniques for implementing staggered oversampling adapted to the modulation transfer function of the instrument so as to enable the amount of spectrum folding to be negligible while nevertheless optimizing acquisition data rate. SUMMARY OF THE INVENTION The invention proposes an instrumental solution which makes it possible to obtain images in which firstly spectrum folding is limited and secondly noise is minimized. More particularly, the invention provides a method of acquiring and processing a satellite observation image of the earth by means of at least one strip or matrix of detectors of the charge coupled device type, said detectors travelling over the observed zone, in which a plurality of pixels are acquired in a given sampling, the method being characterized in that the array of said sampling includes "e" subarrays corresponding to sampling satisfying Shannon's condition, where "e" is an integer or rational number greater than 1, and in that processing is implemented that enables a noise-reduced image to be interpolated at a resolution corresponding to that of a sampling subarray on the basis of the initially acquired image pixels. In other words, the initial image is equivalent to "e" images corresponding to sampling that satisfies Shannon's condition, with which a new image at the same resolution is reconstituted. For each pixel of the new image, "e" times more flux is available than in the case of normal acquisition, such that the mean noise in the 6 new image is divided by ve compared with the noise normally obtained for an image of the same resolution. The invention also provides a camera instrument for implementing the method. BRIEF DESCRIPTION OF THE DRAWINGS Other characteristics and advantages of the invention appear further from the following description. The description is purely illustrative and non-limiting. It should be read with reference to the accompanying drawings, in which: Figure 1, described above, shows the principle of acquisition by push-broom scanning; • Figure 2, also described above, shows a system for processing images picked up by push-broom scanning; Figures 3 and 4 are diagrams showing one possible implementation of the invention using a single strip of detectors; Figure 5 shows another possible implementation of the invention using two strips of detectors ; - Figure 6 shows a possible embodiment of the invention using a matrix of detectors; and * Figures 7 and 8 show two ways in which strips can be organized so that they can be used for increasing field when the invention is implemented using a plurality of detector strips. DESCRIPTION OF PARTICULAR EMBODIMENTS Figures 3 and 4 show a linear strip 11 placed in the focal plane of a satellite observation instrument, and also the projection in said focal plane of the travel direction D of the satellite carrying the strip 11. The strip 11 is oriented in said focal plane so as to form an angle a relative to the direction D where a is equal to arctan(l/3). 7 The sampling pitch AT in the travel direction D is equal to pv10/6 , where p is the pitch between two elementary detectors of the strip 11. Such sampling is equivalent to acquiring three images by staggered sampling complying with Shannon's condition, i.e., for example, three images obtained by sampling of the type described in the Applicant's patent application WO 97/05451. Consequently, with sampling of the kind shown in Figures 3 and 4, it is possible to obtain a new image, after interpolation, e.g. by adding together the three above-mentioned images corresponding to staggered sampling or by implementing processing of the type described in greater detail below, which new image is of resolution that corresponds to that obtained with staggered sampling, but has its noise divided by V3 compared with the noise of an image as normally obtained with such sampling. More generally, sampling performed with a strip 11 that is oriented at an angle a relative to the direction D where a = arctan(l/n), where n is an integer, and implementing a sampling pitch equal to in said direction D, is equivalent to acquiring n images by means of staggered sampling in compliance with Shannon's condition, and this makes it possible by interpolation to obtain an image whose resolution corresponds to that of such staggered sampling, but in which noise is divided by vn. Other variant embodiments can naturally be envisaged. In particular, as shown in Figure 5, it is possible to use two linear strips 21 and 22 in the camera instrument, each having the same pitch p between elementary detectors. 8 The two strips 21 and 22 are oriented at an angle a relative to the direction D where a = arctan(l/n). They are disposed in the camera instrument in a configuration where they are offset by p/2 in the direction in which they extend When the sampling pitch is equal to then acquisition is performed which is equivalent to acquiring n staggered images in compliance with Shannon's condition. Consequently, by interpolating these subimages, it is possible to obtain an image which has the same resolution and whose noise is divided by compared with that normally obtained at said resolution. Similarly, as shown in Figure 6, the above described solutions for linear strips can be generalized to matrices having rows that are oriented at an angle a relative to the direction D, where a = arctan(l/n). When only one matrix is used, a sampling pitch in the travel direction is used that is equal to It is also possible to use two matrices that are offset from each other, with the sampling pitch in the travel direction equal to Compared with the examples described with reference to Figures 3 to 5, which apply to linear strips, images are obtained in which noise is further divided by the square root of the number of rows in the matrix or matrices used. Whether matrices or strips are used, the interpolation processing can be performed on the ground 9 or on-board. Performing it on-board has the advantage of enabling the image for transmission to be compressed by a factor of n. In particular, when the undersampling is of the square type, the interpolated image can be given by: (2) where E is the oversampled array and Ib is the noisy image obtained with the array; where (n1, m1) is a pair of dummy variables corresponding to the coordinates in the interpolated array; and where p1 and p2 are the elementary sampling pitches in the array E in the direction in which the strip extends and in the direction perpendicular thereto (which implies in particular that they satisfy the equation: pl.p2/p2 = e where e is equal to n for all of the cases described with reference to Figures 1 to 5). More generally, for subimages corresponding to arbitrary regular sampling, it is advantageous to implement the following interpolation: where I is the function which corresponds to the interpolated image, Ib is the function which corresponds to the oversampling image; where z' belongs to the oversampled array E and z belongs to the subarray E0; and where RO is the inverse Fourier transform of the function characteristic of RO which is the reciprocal cell of EO. In a variant, interpolation can consist more simply in adding together the pixels of the sampling subarrays 10 (after interpolating each image so as to make the pixels coincide). Furthermore, as will have been understood, although the technique described above with reference to Figures 3 to 6 enables more flux to be obtained for each pixel, i.e. enables an image to be obtained with less noise, it nevertheless gives rise to a corresponding loss of field. This loss of swath can be compensated by adding strips in the focal plane. To this end, as shown in Figure 7, it is possible to place a row constituted by a plurality of parallel strips 41 in the focal plane where each of the strips is inclined at the same angle a relative to the direction D. It is also possible, as shown in Figure 8, to use a plurality of superposed rows of strips 41. 11 We Claim: 1. A method of acquiring and processing a satellite observation image of the earth by means of at least one strip or matrix of detectors of a. charge coupled device type, said detectors traveling over an observed zone, in which a plurality of image pixels are initially acquired in a give sampling, characterized in that the array of said sampling has "e" subarrays corresponding to sampling satisfying Shannon's condition, where "e" is an integer or rational number greater than 1, and in that processing is implemented that enables a noise-reduced image to be interpolated at a resolution corresponding to that of a sampling subarray on the basis of the initially acquired image pixels, and in that the strip or matrix of detectors is oriented relative to the projection of the travel direction of the satellite on a focal plane containing the detectors at an angle a where a=arctan (1/n) and where n is an integer, and in that the sampling pitch in said travel direction is equal to pv1+n2/2n where p is the pitch between two elementary detectors of the strip or matrix. 2. A method as claimed in claim 1, wherein said image is acquired by means of two strips or matrices oriented at an angle a, where a=arctan (1/n), relative to the projection of the satellite displacement direction on a focal plane, in that the two strips or matrices are offset by p/2 in the direction in which they extend and b/2n+Kp/n in the 12 direction perpendicular thereto, where K is an integer and where p is the pitch between two elementary detectors in each of said strips, and in that the sampling pitch in the displacement direction is equal to pvl+n2/2n. A method as claimed in claim 1, wherein to interpolate the noise- reduced image, the pixels of the sampling subarrays are added together. A method as claimed in claim 1, wherein the interpolation processing consists in determining: where I is the function corresponding to the interpolated image, lb is the function corresponding to an initial image; where E is the array of the initial image and EO is the subarray of the interpolated image; and where RO is the inverse Fourier transform of the function characteristic or RO which is the reciprocal cell of EO. A method as claimed in claim 4, wherein when the sampling of the sub arrays is of the square type, the interpolation processing consists in determining: where 1b, is the function which corresponds to an initially acquired image and E is the array of the initial image; where (n1, m1,) is a pair of dummy variables corresponding to the 13 coordinates in the interpolated array; and, where p1 and p2 are the elementary sampling pitches in the array E in the direction in which the strip or matrix extends and in the direction perpendicular thereto. A camera instrument for implementing the method as claimed in claim 1, wherein it comprises a strip or matrix of detectors oriented at an angle a=arctan(l/n), where n is an integer, relative to the projection of the satellite travel direction in a focal plane. An instrument as claimed in claim 6, wherein it has at least one row of strips or matrices oriented at an angle a=arctan (1/n) extending in parallel. An instrument as claimed in claim 6, wherein it has two strips oriented at an angle (x. where a=arctan(1/n), relative to the projection of the satellite travel direction in a focal plane, and in that the two strips are offset by p/2 in the direction in which they extend and by p/2n + Kp/n in the direction perpendicular thereto, where K is an integer and where p is the pitch between two elementary detectors in each of said strips. 14 The invention concerns a method of acquiring and processing an earth satellite observation image using at least an array or matrix of load transfer sensors moving opposite a zone under observation, which consists in acquiring a plurality of pixels according to a given sampling. The invention is characterized in that said sampling network includes "e" sub-systems corresponding to a sampling verifying the Shannon theory, where "e" is an integer or rational number greater than 1, and in using a process for interpolating a noise-corrected image whereof the resolution corresponds to that of a sampling sub-system from the initially acquired image pixels.

Full Text

The present invention relates to acquiring a satellite observation image of the earth by push-broom scanning using one or more strips (or matrix(ces)) of detectors of the charge coupled device (CCD) type travelling relative to the observed zone.
FIELD OF THE INVENTION
The principle of push-broom type scanning is shown in Figure 1 for the case of a single strip 1 of detectors.
As the satellite carrying the strip 1 travels, the strip observes successive lines L1, L2, . . . , Ln perpendicular to the travel direction (arrow D). At each instant, an optical system 2 of the instrumentation forms the image of a line of landscape on the line of detectors, with the strip 1 being placed in the focal plane of the optical system 2, perpendicularly to the speed vector of the satellite. The landscape scans over each detector which integrates light flux over an exposure time and transforms it into a proportional electric charge.
Figure 2 shows a conventional system for processing images taken in this way.
That processing system comprises, in outline: a unit 3 for processing and amplifying the output from the detectors of the strip 1; an analog-to-digital encoder 4 that receives the signal output by the unit 3 ,-transmitter means 5 for taking the digital images picked up in this way and transmitting them from the satellite to the ground; and a unit 6 on the ground for reconstituting the images.
The unit 3 includes, in particular, a shift register into which the information as integrated and stored in charge form in each detector of the strip 1 is transferred at the end of an exposure time. Thereafter, the register transfers the charges in the form of
2

electrons, and the charges are converted into a succession of voltages that are proportional to the received and integrated light fluxes.
The unit 6 on the ground reconstructs the images, in particular by implementing deconvolution processing to compensate for instrument defects, and where appropriate interpolation processing to reconstitute certain pixels within the image.
STATE OF THE ART AND OBJECT OF THE INVENTION
It is known that the finer the resolution of an image, the more the signal-to-noise (SNR) tends to diminish, in particular because of the lack of light flux, so that images cease to be acceptable.
Unfortunately, presently known solutions for mitigating that drawback are not satisfactory.
In particular, one possible solution consists in increasing the dimensions of the instrument, and in particular the diameter of the pupil of the telescope. However that solution is expensive.
Another solution consists in controlling the attitude of the satellite so as to slow down its rate of scanning. That technique allows light flux to be integrated for longer, but it implies a loss of continuity in the track of the satellite, and consequently leads to a loss of data.
A third solution consists in using special time delay integration (TDI) detectors. These detectors are constituted as matrices in which the rows are shifted electronically so as to compensate for the speed of the satellite. Nevertheless, such a system is complex to embody and implement. Furthermore, it does not enable sampling to be performed that is adapted to the modulation transfer function so as to ensure that spectrum folding is negligible and acquisition data rate is optimized.
3

An object of the invention is thus to provide an acquisition method of this type in which instrument noise is reduced and which does not suffer from the drawbacks of the above-mentioned solutions.
THEORETICAL BACKGROUND
It is known that image acquisition by push-broom scanning can be modelled linearly by the formulation:

where:
designates the convolution operation; is the landscape whose image is to be acquired; is the impulse response of the instrument; is the noise superposed on the filtered landscape; is the raw image; and
is a two-dimensional Dirac comb which means that the continuous image has been made discrete.
The image is digital and encoded on a limited number of bits.
A spectral representation is often preferred. It is represented by the Fourier transform (FT) of the raw image, i.e.:

where the hats indicate that these are Fourier transforms, and where the abbreviation MTF stands for modulation transfer function, which is the Fourier transform of the impulse response. In the spectral representation, the convolution operation is equivalent to a multiplicative operation.
The parameters that are mainly involved in the above equations are described in greater detail below.
Modulation transfer function or MTF
The MTF is the attenuation factor of spatial frequencies. The higher the frequency the lower the MTF. The geometrical location where the MTF becomes zero is the boundary of the instrumental cutoff. The instrument
4

spectral medium is the low frequency domain as limited by the instrumental cutoff boundary.
The MTF of an instrument is the product of the MTF of its optical system multiplied by the MTF of its detector and by the MTF of its displacement in the travel direction.
Noise
Instrumental noise is always present. A simplified approach consists in considering that it is characterized by the mean SNR (generally known once the mission of the satellite has been defined), and that, compared with other noise, Poisson type photon noise is preponderant at the mean luminance. With such a distribution, the standard deviation of the noise varies with the square root of the number of photons N picked up by the detector. Consequently
Sampling array-Present satellites generate a square orthogonal array (speed and strip are in orthogonal directions). The sampling step size in the speed direction on the ground is ?t = v.te (where v is the speed of the satellite and te is the sampling time), and the sampling step size in the strip direction is equal to the distance between two adjacent elementary CCDs. Consequently, using the arrays implemented in present satellites, the two pitches are equal and the projections of ground pixels do not overlap (the ground is scanned once only).
However, with acquisition as performed by current satellites, the sampling frequency is equal to the cutoff frequency so Shannon's condition is not satisfied (te 5

To mitigate that drawback, proposals have recently-been made, in particular in French patent FR 2 678 460 in the name of the Applicant, to use a technique whereby two offset strips make it possible to perform oversampling so as to limit spectrum folding.
In its patent application WO 97/05451, the Applicant has also proposed techniques for implementing staggered oversampling adapted to the modulation transfer function of the instrument so as to enable the amount of spectrum folding to be negligible while nevertheless optimizing acquisition data rate.
SUMMARY OF THE INVENTION
The invention proposes an instrumental solution which makes it possible to obtain images in which firstly spectrum folding is limited and secondly noise is minimized.
More particularly, the invention provides a method of acquiring and processing a satellite observation image of the earth by means of at least one strip or matrix of detectors of the charge coupled device type, said detectors travelling over the observed zone, in which a plurality of pixels are acquired in a given sampling, the method being characterized in that the array of said sampling includes "e" subarrays corresponding to sampling satisfying Shannon's condition, where "e" is an integer or rational number greater than 1, and in that processing is implemented that enables a noise-reduced image to be interpolated at a resolution corresponding to that of a sampling subarray on the basis of the initially acquired image pixels.
In other words, the initial image is equivalent to "e" images corresponding to sampling that satisfies Shannon's condition, with which a new image at the same resolution is reconstituted. For each pixel of the new image, "e" times more flux is available than in the case of normal acquisition, such that the mean noise in the
6

new image is divided by ve compared with the noise normally obtained for an image of the same resolution.
The invention also provides a camera instrument for implementing the method.
BRIEF DESCRIPTION OF THE DRAWINGS
Other characteristics and advantages of the invention appear further from the following description. The description is purely illustrative and non-limiting. It should be read with reference to the accompanying drawings, in which:
Figure 1, described above, shows the principle of acquisition by push-broom scanning;
• Figure 2, also described above, shows a system for
processing images picked up by push-broom scanning;
Figures 3 and 4 are diagrams showing one possible
implementation of the invention using a single strip of
detectors;
Figure 5 shows another possible implementation of
the invention using two strips of detectors ;
- Figure 6 shows a possible embodiment of the invention using a matrix of detectors; and
* Figures 7 and 8 show two ways in which strips can
be organized so that they can be used for increasing
field when the invention is implemented using a plurality
of detector strips.
DESCRIPTION OF PARTICULAR EMBODIMENTS
Figures 3 and 4 show a linear strip 11 placed in the focal plane of a satellite observation instrument, and also the projection in said focal plane of the travel direction D of the satellite carrying the strip 11.
The strip 11 is oriented in said focal plane so as to form an angle a relative to the direction D where a is equal to arctan(l/3).
7

The sampling pitch AT in the travel direction D is equal to pv10/6 , where p is the pitch between two elementary detectors of the strip 11.
Such sampling is equivalent to acquiring three images by staggered sampling complying with Shannon's condition, i.e., for example, three images obtained by sampling of the type described in the Applicant's patent application WO 97/05451.
Consequently, with sampling of the kind shown in
Figures 3 and 4, it is possible to obtain a new image,
after interpolation, e.g. by adding together the three
above-mentioned images corresponding to staggered
sampling or by implementing processing of the type
described in greater detail below, which new image is of
resolution that corresponds to that obtained with
staggered sampling, but has its noise divided by V3
compared with the noise of an image as normally obtained
with such sampling.
More generally, sampling performed with a strip 11
that is oriented at an angle a relative to the direction
D where a = arctan(l/n), where n is an integer, and
implementing a sampling pitch equal to in said
direction D, is equivalent to acquiring n images by means of staggered sampling in compliance with Shannon's condition, and this makes it possible by interpolation to obtain an image whose resolution corresponds to that of such staggered sampling, but in which noise is divided by
vn.
Other variant embodiments can naturally be
envisaged. In particular, as shown in Figure 5, it is
possible to use two linear strips 21 and 22 in the camera
instrument, each having the same pitch p between
elementary detectors.
8

The two strips 21 and 22 are oriented at an angle a relative to the direction D where a = arctan(l/n).
They are disposed in the camera instrument in a configuration where they are offset by p/2 in the direction in which they extend When the sampling pitch is equal to then acquisition is performed which is equivalent to acquiring n staggered images in compliance with Shannon's condition. Consequently, by interpolating these subimages, it is possible to obtain an image which has
the same resolution and whose noise is divided by compared with that normally obtained at said resolution.
Similarly, as shown in Figure 6, the above described solutions for linear strips can be generalized to matrices having rows that are oriented at an angle a relative to the direction D, where a = arctan(l/n). When only one matrix is used, a sampling pitch in the travel
direction is used that is equal to It is also
possible to use two matrices that are offset from each other, with the sampling pitch in the travel direction
equal to
Compared with the examples described with reference to Figures 3 to 5, which apply to linear strips, images are obtained in which noise is further divided by the square root of the number of rows in the matrix or matrices used.
Whether matrices or strips are used, the interpolation processing can be performed on the ground
9

or on-board. Performing it on-board has the advantage of enabling the image for transmission to be compressed by a factor of n.
In particular, when the undersampling is of the square type, the interpolated image can be given by:

(2)
where E is the oversampled array and Ib is the noisy image obtained with the array;
where (n1, m1) is a pair of dummy variables corresponding to the coordinates in the interpolated array; and
where p1 and p2 are the elementary sampling pitches in the array E in the direction in which the strip extends and in the direction perpendicular thereto (which implies in particular that they satisfy the equation:
pl.p2/p2 = e where e is equal to n for all of the cases described with reference to Figures 1 to 5).
More generally, for subimages corresponding to arbitrary regular sampling, it is advantageous to implement the following interpolation:

where I is the function which corresponds to the interpolated image, Ib is the function which corresponds to the oversampling image;
where z' belongs to the oversampled array E and z belongs to the subarray E0; and
where RO is the inverse Fourier transform of the function characteristic of RO which is the reciprocal cell of EO.
In a variant, interpolation can consist more simply in adding together the pixels of the sampling subarrays
10

(after interpolating each image so as to make the pixels coincide).
Furthermore, as will have been understood, although the technique described above with reference to Figures 3 to 6 enables more flux to be obtained for each pixel, i.e. enables an image to be obtained with less noise, it nevertheless gives rise to a corresponding loss of field.
This loss of swath can be compensated by adding strips in the focal plane.
To this end, as shown in Figure 7, it is possible to place a row constituted by a plurality of parallel strips 41 in the focal plane where each of the strips is inclined at the same angle a relative to the direction D.
It is also possible, as shown in Figure 8, to use a plurality of superposed rows of strips 41.
11

We Claim:
1. A method of acquiring and processing a satellite observation image of the earth by means of at least one strip or matrix of detectors of a. charge coupled device type, said detectors traveling over an observed zone, in which a plurality of image pixels are initially acquired in a give sampling, characterized in that the array of said sampling has "e" subarrays corresponding to sampling satisfying Shannon's condition, where "e" is an integer or rational number greater than 1, and in that processing is implemented that enables a noise-reduced image to be interpolated at a resolution corresponding to that of a sampling subarray on the basis of the initially acquired image pixels, and in that the strip or matrix of detectors is oriented relative to the projection of the travel direction of the satellite on a focal plane containing the detectors at an angle a where a=arctan (1/n) and where n is an integer, and in that the sampling pitch in said travel direction is equal to pv1+n2/2n where p is the pitch between two elementary detectors of the strip or matrix.
2. A method as claimed in claim 1, wherein said image is acquired by
means of two strips or matrices oriented at an angle a, where
a=arctan (1/n), relative to the projection of the satellite displacement
direction on a focal plane, in that the two strips or matrices are offset
by p/2 in the direction in which they extend and b/2n+Kp/n in the
12

direction perpendicular thereto, where K is an integer and where p is
the pitch between two elementary detectors in each of said strips, and
in that the sampling pitch in the displacement direction is equal to
pvl+n2/2n.
A method as claimed in claim 1, wherein to interpolate the noise-
reduced image, the pixels of the sampling subarrays are added
together.
A method as claimed in claim 1, wherein the interpolation processing
consists in determining:
where I is the function corresponding to the interpolated image, lb is
the function corresponding to an initial image;
where E is the array of the initial image and EO is the subarray of the
interpolated image; and
where RO is the inverse Fourier transform of the function
characteristic or RO which is the reciprocal cell of EO.
A method as claimed in claim 4, wherein when the sampling of the
sub arrays is of the square type, the interpolation processing consists
in determining:
where 1b, is the function which corresponds to an initially acquired
image and E is the array of the initial image;
where (n1, m1,) is a pair of dummy variables corresponding to the
13

coordinates in the interpolated array; and,
where p1 and p2 are the elementary sampling pitches in the array E
in the direction in which the strip or matrix extends and in the
direction perpendicular thereto.
A camera instrument for implementing the method as claimed in
claim 1, wherein it comprises a strip or matrix of detectors oriented at
an angle a=arctan(l/n), where n is an integer, relative to the projection
of the satellite travel direction in a focal plane.
An instrument as claimed in claim 6, wherein it has at least one row
of strips or matrices oriented at an angle a=arctan (1/n) extending in
parallel.
An instrument as claimed in claim 6, wherein it has two strips
oriented at an angle (x. where a=arctan(1/n), relative to the projection
of the satellite travel direction in a focal plane, and in that the two
strips are offset by p/2 in the direction in which they extend and by
p/2n + Kp/n in the direction perpendicular thereto, where K is an
integer and where p is the pitch between two elementary detectors in
each of said strips.

14
The invention concerns a method of acquiring and processing an earth satellite observation image using at least an array or matrix of load transfer sensors moving opposite a zone under observation, which consists in acquiring a plurality of pixels according to a given sampling. The invention is characterized in that said sampling network includes "e" sub-systems corresponding to a sampling verifying the Shannon theory, where "e" is an integer or rational number greater than 1, and in using a process for interpolating a noise-corrected image whereof the resolution corresponds to that of a sampling sub-system from the initially acquired image pixels.

Documents:

« Previous Patent

Next Patent »

Patent Number

202503

Indian Patent Application Number

IN/PCT/1999/00086/KOL

PG Journal Number

08/2007

Publication Date

23-Feb-2007

Grant Date

23-Feb-2007

Date of Filing

15-Nov-1999

Name of Patentee

CENTRE NATIONAL D'ETUDES SPATIALES

Applicant Address

A FRENCH COMAPNY, OF 2 PLACE MAURICE QUENTIN, 75001 PARIS, FRANCE

Inventors:

#	Inventor's Name	Inventor's Address
1	BERNARD ROUGE	19, RUE BONNAT, 31400 TOULOUSE, FRANCE
2	JEAN-MICHEL MOREL	18 BIS RUE PIERRE DEMOURS, 75017 PARIS, FRANCE

PCT International Classification Number

H 04 N 7/18

PCT International Application Number

PCT/FR99/00622

PCT International Filing date

1999-03-18

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	9803450	1998-03-20	France