Method for measuring and / or compensating for an aberration in an image acquired by an optical instrument comprising a multispectral detector array.

Abstract

A method (100) for measuring an aberration in an image acquired by a detector array and a plurality of spectral filter arrays, comprising the following steps: acquisition (101) of a known point source; determination (102) of an initial value (VIP) of the parameters of a model of an impulse response; sampling (103); calculation (104) of a cost as a function of the distance between the calculated impulse response and the observed impulse response; calculation (105) of an overall cost (CG) for all spectral bands; estimation (102bis) of a value (VEP) of each parameter; sampling (103bis); calculation (104bis) of the cost; calculation (105bis) of the overall cost (NCG); comparison (106) between the overall cost (NCG) and the overall cost (CG); repetition (107). Figure 1

Description

Title of the invention: Method for measuring and / or compensating for an aberration in an image acquired by an optical instrument comprising a matrix of multispectral detectors.

[0001] The present invention relates to the field of calibration of an optical instrument comprising a matrix sensor, in particular a multi-spectral optical matrix sensor, by observation of a known point source, necessary in particular for the compensation of one or more optical aberrations of the optical instrument in an image acquired by the optical instrument.

[0002] It is known to compensate for one or more optical aberrations of the optical instrument in an image acquired by an optical instrument comprising a matrix of detectors, using parametric modeling of an impulse response of the instrument, taking into account the different contributors to the formation of the image by the instrument, i.e. an optical combination, a detector, a movement of the instrument during the exposure time, etc.

[0003] The known methods do not, however, deal with the case of an optical instrument comprising a matrix of multispectral detectors.

[0004] The invention therefore aims to propose a solution to all or part of these problems.

[0005] To this end, the present invention relates to a method for measuring an aberration in an image acquired by an optical instrument comprising a matrix of multispectral detectors, the matrix of multispectral detectors comprising a matrix of detectors, and a plurality of matrices of spectral filters, each spectral filter matrix being associated with a spectral band, the method comprising the following steps: - acquisition by the detector matrix of the optical instrument of an image of a known point source, the image comprising, for each spectral filter matrix, a part of an image task of the point source acquired in the spectral band of said spectral filter matrix; - for each spectral filter matrix, determining an initial value of each parameter of a set of parameters of a parametric model of a modeled impulse response of the optical instrument, the set of parameters comprising a subset of parameters common to the plurality of spectral filter matrices, and a subset of parameters specific to each spectral filter matrix; - for each spectral filter matrix, sampling of an impulse response calculated on the basis of the initial value of the parameters applied to the model parametric, the sampling being carried out according to a sampling mask associated with the spectral filter matrix, - for each spectral filter matrix, calculation of a value of a cost in the spectral band, the value of the cost being a function of a distance between the calculated impulse response and the impulse response observed for each of the points of the sampling mask, - calculation of a value of an overall cost for all spectral bands based on the value of the cost in each spectral band, - for each spectral filter matrix, estimation of an estimated value of each parameter of the set of parameters, the estimated value being a function of an estimated variation from the value of the overall cost, the estimated value being equal to the estimated variation added to the initial value of the parameters; - for each spectral filter matrix, sampling of a new impulse response calculated on the basis of the estimated value of the parameters applied to the parametric model, the sampling being carried out according to the sampling mask associated with the spectral filter matrix, - for each spectral filter matrix, calculation of a new cost value in said spectral band, the new cost value being a function of the distance between the new calculated impulse response and the impulse response observed for each of the points of the sampling mask, - calculation (105bis) of a new value of the overall cost for all spectral bands as a function of the new value of the cost in each spectral band. - comparison between the new value of the overall cost and the value of the overall cost to obtain a result of the comparison; - determining that the initial value of the parameters is equal to the estimated value of the parameters and that the value of the overall cost is equal to the new value of the overall cost, and for each spectral filter matrix, repeating the steps between the step of estimating an estimated value of the parameters, up to the step of comparing the new value of the overall cost with the value of the overall cost until a stopping criterion is satisfied by the result of the comparison, so that at each iteration of the repetition step, during the estimation step the variation is estimated from an initial value of the parameters equal to the estimated value of the parameters estimated during the previous iteration, and from a value of the overall cost equal to a new value of the overall cost calculated during the previous iteration; - determining a measure of the aberration on the basis of the estimated value of the parameters of the impulse response.

[0006] According to one embodiment, the invention comprises one or more of the following characteristics, alone or in a technically acceptable combination.

[0007] According to one embodiment, the point source is a star observed outside the Earth's atmosphere, a spectral response of the observed star being known.

[0008] According to one implementation mode, the initial value of the parameters is equal to 0 for each parameter.

[0009] According to one embodiment, each spectral filter matrix is ​​superimposed on the detector matrix, a position of the spectral filter matrix relative to the detector matrix being offset along at least one of the two dimensions of the detector matrix, such that each sampling point of the sampling mask associated with a spectral filter matrix is ​​offset relative to a corresponding point of the sampling mask associated with another spectral filter matrix.

[0010] According to these provisions, the greater the number of spectral filter matrices, the more robust the estimation of the value of the common parameters of the parametric model, and the better the quality of the aberration compensation obtained using the model.

[0011] According to one embodiment, the acquisition step comprises the acquisition of several known point sources and / or several acquisitions of the same known point source, the set of parameters of the parametric model comprising a subset of parameters specific to each acquisition, the parametric impulse response model being associated with the different impulse responses observed in the spectral band of each spectral filter matrix, and for each point source and / or for each acquisition of the same point source, and in which the following steps are implemented for each spectral filter matrix, and for each point source, and / or for each acquisition of the same point source: - step of determining the initial value of the parameters, and step of estimating the estimated value of the parameters, - steps of sampling the impulse response calculated on the basis of the initial value of the parameters, respectively of the new impulse response calculated on the basis of the estimated value of the parameters, - steps of calculating the cost value, respectively the new cost value, and in which the global cost value, respectively the new global cost value, is calculated for all spectral bands and for all point sources.

[0012] According to one embodiment, among the plurality of successive acquisitions of a point source, at least one parameter of the parametric model takes a value during one acquisition and another different value during at least one other acquisition, the value and the other value of said at least one parameter being unknown during of the acquisition step. The variation applied between the value and the other value is not necessarily known. If the variation of the parameter is known, the estimation of the value of the parameters of the parametric model will be more robust.

[0013] According to one embodiment, the plurality of spectral filter matrices comprises at least three spectral filter matrices, configured to each filter in a spectral band complementary to the other matrices.

[0014] According to one embodiment, the plurality of spectral filter matrices is of the Bayer matrix type.

[0015] According to one embodiment, in the step of estimating the estimated value of each parameter, the variation is estimated from a Jacobian matrix of the parametric model of the impulse response, so that the repetition of the steps between the step of estimating an estimated value of the parameters, and the step of comparing the new value of the overall cost and the value of the overall cost, until a stopping criterion is satisfied by the result of the comparison, corresponds to an implementation of an iterative algorithm of the Levenberg-Marquardt type.

[0016] According to one embodiment, an amplitude of the estimated variation is constrained, so that the estimated variation is between a minimum value and a maximum value.

[0017] According to one embodiment, the value of the overall cost, respectively the new value of the overall cost, is equal to a sum of the value of the costs, respectively the new value of the costs, over all the spectral bands.

[0018] According to one embodiment, the result of the comparison is equal to a difference between the value of the new overall cost and the value of the overall cost, the predetermined stopping criterion being satisfied when the result of the comparison is less than a threshold.

[0019] According to one embodiment, the set of parameters comprises at least one of the parameters of an impulse response representative of an optical part of the optical instrument, the parameters of an impulse response representative of the integration of the charges on at least one elementary detector of the detector matrix, the parameters of an impulse response representative of a motion, i.e. of a linear movement, of the detector matrix during an exposure time, the parameters of an impulse response representative of a movement, i.e. of a non-linear movement, of the detector matrix during the exposure time, the parameters representative of the coordinates of a geometric image of the point source, said geometric image being point-like.

[0020] According to one embodiment, the method comprises a step in which the measurement of the aberration is used to compensate for the aberration by processing to improve the image acquired by the optical instrument, the processing improvement being a function of the measurement of the aberration, the processing comprising for example an application of a compensation filter calculated from the measurement of the aberration, or the processing comprising for example a restoration of the image as a function of the measurement of the aberration

[0021] According to one embodiment, the processing is a command of the optical instrument to actuate an active optic of the optical instrument, the command being based on the estimated value of the parameters of the parametric model.

[0022] According to one embodiment, the control is the application of an instruction for correcting optical aberrations to correction means, for example by acting on one or more degrees of freedom of a secondary mirror of the optical combination or on a deformable mirror of the optical combination. For example, the refocusing may consist of translating the secondary mirror along an optical axis by a quantity determined from the measurement of the aberration according to the method.

[0023] According to one embodiment, the control of the optical instrument is a focusing control.

[0024] According to one aspect, the invention also relates to a computer program comprising a set of instructions configured to implement the method according to one of the implementation modes described above when it is executed by a processor of an image processing and control unit, of an optical instrument comprising a matrix of multispectral detectors, the matrix of multispectral detectors comprising a matrix of detectors, and a plurality of matrices of spectral filters, each spectral filter matrix being associated with a spectral band.

[0025] According to another aspect, the invention also relates to a computer-readable medium comprising the computer program described above.

[0026] For a better understanding, an embodiment and / or implementation of the invention is described with reference to the attached drawings representing, by way of non-limiting example, an embodiment or implementation respectively of a device and / or a method according to the invention. The same references in the drawings designate similar elements or elements whose functions are similar.

[0027] [Fig.l] is a schematic representation of the sequence of steps of the method according to an exemplary implementation of the invention.

[0028] The method according to the invention is based on a combined optimization of parameters of a model of an impulse response modeled from an acquisition set of a known point source, for example a star in the case of observations outside the Earth's atmosphere. From a fine modeling of the acquisition chain, i.e. of the optical instrument which notably includes the optics, the detector, multispectral filters, and their arrangement, the model parameters are refined iteratively. The parameters include global parameters relating to an overall state of the acquisition chain, and parameters more specific to each spectral band.

[0029] According to an exemplary embodiment of the multispectral detector array, the plurality of spectral filter arrays comprises at least three spectral filter arrays, each configured to filter in a spectral band complementary to the other arrays. Those skilled in the art will understand that each filter is not strictly limited to one wavelength, i.e. each filter has a low but non-zero transmission in neighboring wavelengths around the central wavelength of the filter. Furthermore, the method according to the invention also applies to a combination comprising a panchromatic band, and one or more monospectral filters.

[0030] According to a particular exemplary embodiment, the plurality of spectral filter matrices is of the Bayer matrix type.

[0031] A value of the model parameters is estimated by searching for the value that minimizes a cost function, a function of a difference between the modeled impulse response and the observed impulse response. The search for the value of the parameters that minimizes the cost function is carried out using an iterative algorithm, starting from an initial value of each parameter of the set of parameters, with a stopping criterion for the iterative algorithm. The iterative algorithm is for example a Levenberg Marquardt algorithm.

[0032] Thus, for each observation of a point source and for each spectral band, a value of the parameters is estimated, for each parameter of the set of parameters considered, which minimizes said cost function.

[0033] The value of the parameters thus obtained is used to calculate, from the parametric model, a value of the modeled impulse response for all the pixels of the detector matrix.

[0034] For each spectral band, the impulse response modeled on the basis of the estimated value of the parameters is then sampled according to a sampling mask associated with the spectral filter matrix.

[0035] A deviation between the modeled impulse response and the observed impulse response in said spectral band is calculated on the basis of a difference between the modeled impulse response and the observed impulse response for each of the points of the sampling mask for said spectral band.

[0036] In other words, to calculate the deviation, the modeled impulse response is sampled according to the sampling mask of each spectral filter, and the difference between the modeled impulse response and the observed impulse response is not calculated only for these points of the mask. Each point is analyzed only once.

[0037] A cost, a function of the deviation thus calculated, is determined for each spectral band, and an overall cost, for all the spectral bands considered, is calculated on the basis of all the deviations calculated for each spectral band.

[0038] The above steps are repeated until a difference between the value of the overall cost and the next value of the overall cost is less than a predetermined threshold, in other words until the value of the overall cost approaches sufficiently an undetermined limit value.

[0039] The method according to an exemplary implementation of the invention, as described above, makes it possible to improve the precision of determining the physical parameters of the optical instrument, in particular of the optical combination; it also makes it possible to calibrate the actuation and correction devices of the instrument by considerably simplifying the operational conditions: the implementation of the method requires only a few images acquired by the instrument of known source points, for example stars.Unlike current methods (refocusing, absolute calibration, MTF measurement) which require either the observation of varied targets or strong operational constraints (imperative need for voluntary defocusing of the instrument for acquisitions at different operating points), the proposed method requires only a few acquisitions on known stars: there is no need to have several focusing points, and if several focusing points are used, it is not necessary to know precisely the variation applied between these focusing points. In the case of a symmetrical optical pupil, there is sign ambiguity on the symmetrical part of the optical aberrations. For example, in the case of defocus, we do not know in which direction to move the secondary mirror.This ambiguity is removed if we have acquisitions for which we have voluntarily introduced a known defocus, or if the pupil is not symmetrical, which is the usual case due to the presence of radial "arms" due to the physical supports of said secondary mirror.

[0040] The method makes it possible, for example, to compensate for a distortion in an image acquired by an optical instrument, the optical instrument comprising in particular an optical combination and a matrix of multispectral detectors, the optical combination being arranged to form an image of a scene observed on the matrix of multispectral detectors placed in a focal plane of the optical combination.

[0041] The distortion compensation processing is, according to an exemplary implementation, a transformation of the image by applying a geometric transformation matrix, said geometric transformation matrix being a matrix for passing angular coordinates of a point of the observed scene to coordinates in a target reference frame. The transformation matrix is ​​determined from pa meters which allow the modeling of an impulse response of the instrument, in order to compensate for the different distortion factor(s) introduced by the different components of the optical instrument.

[0042] The method also makes it possible, for example, to compensate for the distortion by an actuation command of an active optic of the optical instrument, in order to arrive at the target optical aberration state after actuation. This is applicable to the refocusing of the optical instrument and thus contributes to the calibration of an active optic.

[0043] The method also allows, for example, the estimation of geometric parameters (position / rotation) of mirrors of the optical combination, and more generally of any parameter involved in the optical model;

[0044] The method also allows, for example, the measurement of the FTM to calibrate the restoration algorithms, or the calculation of the absolute calibration coefficient allowing the conversion of digital count values ​​into equivalent luminance.

[0045] The method also allows, for example, a re-estimation of a detector FTM which translates the effect of integration of the charges on an elementary detector (diffusion of the charges widening the PSF) to improve the model of the acquisition chain.

[0046] The method also allows, for example, an estimation of the tracking (linear movement) to estimate the satellite pointing drifts;

[0047] The method also allows, for example, an estimation of the movement (non-linear remainder of the movements) to estimate the vibrations.

[0048] The method 100 according to an exemplary implementation of the invention will now be described with reference to [Fig.l].

[0049] The method comprises a step 101 of acquisition, by the matrix of detectors of the optical instrument, of an image of one or more known point sources; in particular, the emission spectrum and the apparent magnitude of each of these stars is known.

[0050] According to an exemplary implementation, the acquisition step 101 may also comprise several acquisitions of the same known point source.

[0051] The point source(s) are for example stars, a spectral response of each of the observed stars being known.

[0052] The image comprises, for each point source, and for each spectral filter matrix, an image task acquired in the spectral band of said spectral filter matrix; for each point source, the image of the point source is spread over a set of adjacent pixels belonging to potentially different spectral bands. It is necessary to distinguish the monochromatic instrumental impulse response, image of a monochromatic point source of given wavelength and unit integral (spatial and spectral Dirac distribution) and the image of a polychromatic point source which will be a spectral integral of the impulse responses monochromatic spectral responses weighted by the spectral response of the stars. In other words, the image task of a point source differs from one spectral band to another spectral band.

[0053] The image task of a point source, in each spectral band, corresponds to an impulse response of the optical instrument observed in said spectral band. The impulse response of the optical instrument in each spectral band is modeled, in a manner known to those skilled in the art, as a convolution product of the impulse responses in said spectral band of the different elements of the acquisition chain constituting the optical instrument considered. The optical instrument comprises in particular the optics of the optical instrument, the elementary detectors of the detector matrix, without forgetting, where appropriate, the other contributors to the impulse response of the instrument such as possible displacements during the pause time (stretching and / or moving).

[0054] The parametric model of the impulse response is thus a function of a set of physical parameters characteristic of the different contributors of the acquisition chain, i.e. of the optical instrument. The set of parameters comprises global parameters, such as for example the coefficients of the Zernike polynomials, characteristic of a global state of the optical combination, and parameters specific to certain elements of the acquisition chain of the optical instrument, in particular specific to each spectral band and to each acquisition. Thus, for example, the set of parameters may comprise the parameters of an impulse response representative of a pupil of the optical instrument, and / or the parameters of an impulse response representative of the integration of the charges on at least one elementary detector of the matrix of detectors, and / or the parameters of an impulse response representative of a streak, i.e.of a linear movement, of the matrix of detectors during a pause time, and / or the parameters of an impulse response representative of a movement, i.e. of a non-linear movement, of the matrix of detectors during the pause time, and / or the parameters representative of the coordinates of a geometric image of the point source, said geometric image being point-like.. .

[0055] For each spectral filter matrix, optionally for each point source, the following step of the method 100 is a step 102 of determining an initial value VIP of each parameter of the set of parameters of the parametric model of the modeled impulse response of the optical instrument, the set of parameters comprising in particular a subset of parameters common to the plurality of spectral filter matrices, and a subset of parameters specific to each spectral filter matrix.

[0056] According to an example of implementation of the method 100, the determination step 102 of an initial value of the parameters of the parametric model, comprises the choice of a value equal to 0, for example, for each parameter. According to another exemplary implementation, the step 102 of determining an initial value comprises the choice, for each parameter, of a value equal to a value previously determined, for example by a previous calibration.

[0057] The following step of the method 100, after the step 102 of determining an initial value of the parameters of the parametric model, to calculate, for each spectral filter matrix, and where appropriate for each point source, an impulse response, on the basis of the initial value VIP of the parameters, then to sample 103 the calculated impulse response, according to a sampling mask associated with the spectral filter matrix: in other words, the value of a sample of the impulse response is calculated, on the basis of the value of the parameters determined previously, at a point of the sampling mask, said sampling mask comprising a set of points distributed regularly on the spectral filter matrix considered, for example at the central point of each elementary filter of the spectral filter matrix.In a particular example, each spectral filter matrix is ​​superimposed on the detector matrix, and a position of the spectral filter matrix relative to the detector matrix is ​​offset along at least one of the two dimensions of the detector matrix, such that each sampling point of the sampling mask associated with one spectral filter matrix is ​​offset relative to a corresponding point of the sampling mask associated with another spectral filter matrix.

[0058] For each spectral filter matrix, and where appropriate for each point source, a value of a cost, in the spectral band considered, which is a function of a distance between the calculated impulse response and the observed impulse response, is then calculated, during a step 104 of calculating a value of a cost. This cost is calculated, for example, on the basis of a difference between the modeled impulse response and the observed impulse response for each of the points of the sampling mask. For example, the cost is equal to a sum of the squares of the differences, for each of the points of the sampling mask, optionally weighted by a normalization coefficient.

[0059] A value of a global cost CG for all the spectral bands, where appropriate for each point source, is then calculated during a step 105 of calculating a value of the global cost CG. For example, the global cost is equal to a sum of the costs previously calculated for each spectral band.

[0060] For each spectral filter matrix, and where appropriate for each point source, an estimated value of each VEP parameter of the set of parameters is estimated during an estimation step 102bis, the estimated value of each parameter VEP being estimated from the initial value VIP of the parameters, determined in step 102, and from a variation VAR; said variation VAR being estimated as a function of the value of the overall cost CG, the estimated value VEP being equal to the estimated variation VAR added to the initial value of the parameters VIP.

[0061] For each spectral filter matrix, and where appropriate for each point source, the steps 103, 104, and 105, previously described, are implemented again with the estimated value of the VEP parameters. These steps are represented in [Fig.l], respectively with the references 103bis, 104bis, 105bis, so that step 105bis is a step of calculating a new value of the overall cost NCG for all the spectral bands, where appropriate for each point source, for the estimated value of the VEP parameters.

[0062] The new value of the overall cost NCG calculated in step 105bis for the estimated value of the VEP parameters is compared, during a comparison step 106, to the value of the overall cost CG calculated in step 105 for the initial value of the VIP parameters determined in step 102. If the result of the comparison RC satisfies a stopping criterion CRA then the estimated value of the VEP parameters is retained for the modeling of the impulse response of the optical instrument. For example, the result of the comparison RC is a difference between the new value of the overall cost NCG and the value of the overall cost CG, and the stopping criterion CRA is satisfied if said difference is less than a predetermined threshold.

[0063] If the result of the comparison RC does not satisfy the stopping criterion CRA then the steps of the method 100 are repeated 107, from the step 102bis of estimating a value of the parameters VEP, passing through the step 104bis of calculating a new value of the cost per spectral band, and through the step 105bis of calculating a new value of the overall cost NCG for all the spectral bands, up to the step 106 of comparing the value of the overall cost CG, calculated at the previous iteration of the repetition 107, and the new value of the overall cost NCG calculated at the current iteration of the repetition 107, until the stopping criterion CRA is satisfied by the result of the comparison RC. At each new iteration, the estimation step 102bis of an estimated value of the VEP parameters is carried out from a value of the parameters equal to the estimated value of the VEP parameters obtained at the estimation step 102bis of the previous iteration.For this purpose, at each new iteration, during a determination step 107bis, the initial value of the VIP parameters becomes equal to the estimated value of the VEP parameters of the previous iteration, and the value of the overall cost CG becomes equal to the new value of the overall cost NCG, for each spectral filter matrix.

[0064] According to an exemplary implementation, at each new iteration, during the estimation step 102bis of an estimated value of the VEP parameters, the VAR variation of the estimated value of the VEP parameters estimated at the previous iteration, is estimated from a Jacobian matrix of the parametric model of the impulse response, said Jacobian matrix being determined with respect to the estimated value of the VEP parameters estimated at the previous iteration, so that the repetition 107, of the steps between the estimation step 102bis of an estimated value of the VEP parameters, and the comparison step 106 between the new value of the global cost NCG and the value of the global cost CG, until a stopping criterion CRA is satisfied by the result of the comparison RC, corresponds to an implementation of an iterative algorithm of the Levenberg-Marquardt type.

[0065] According to a particular example of implementation, an amplitude of the estimated VAR variation is constrained, such that the estimated VAR variation is between a predetermined minimum value and a predetermined maximum value.

[0066] According to these provisions, the estimated value of the VEP parameters gradually converges, as successive iterations progress, towards an estimated value of the VEP parameters increasingly close to a limit value. Thus the impulse response modeled according to the value of the VEP parameters estimated at the end of the repetition step 107bis, 107 is representative of the observed impulse response, so that a measurement of the aberration of the optical instrument can be determined, during a step of determining the measurement of the aberration 108, on the basis of the estimated value of the parameters; optionally this measurement of the aberration can be used to compensate for the aberration, by a processing using the impulse response thus modeled on the basis of the estimated value of the parameters.

[0067] According to an exemplary implementation, the processing for compensating for the aberration is a processing for improving the image acquired by the optical instrument, for example by applying an image compensation filter calculated from the measurement of the aberration, i.e. the estimated value of the parameters of the parametric model, or by applying an image restoration processing, depending on the measurement of the aberration, i.e. the estimated value of the parameters of the parametric model.

[0068] According to another exemplary implementation, the processing for aberration compensation is a control of the optical instrument to actuate an active optic of the optical instrument, the control being based on the estimated value of the parameters of the parametric model.

[0069] Optionally, the processing for aberration compensation may combine a control of the optical instrument to operate an active optic of the optical instrument, and an image enhancement processing.

[0070] It should be noted that the greater the number of spectral filter matrices, the more robust the estimation of the value of the common parameters of the parametric model is, and the better the quality of the aberration compensation, obtained using the model.

[0071] As indicated, the set of parameters comprises global parameters characteristic of a global state of the optical combination, and parameters specific to certain elements of the acquisition chain of the optical instrument, in particular specific to each spectral band. According to a particular example, the set of parameters of the parametric model comprises a subset of parameters specific to each new acquisition of a known point source, the parametric impulse response model being representative of the different impulse responses observed in the spectral band of each spectral filter matrix, where appropriate for each point source and / or for each acquisition of the same point source.The following steps are implemented for each spectral filter matrix, where appropriate for each point source, and / or for each acquisition of the same point source: . - step 102 of determining the initial value of the VIP parameters, and step 102bis of estimating the estimated value of the VEP parameters, - sampling steps 103,103bis of the impulse response calculated on the basis of the initial value of the VIP parameters, respectively of the new impulse response calculated on the basis of the estimated value of the VEP parameters, - calculation steps 104, 104bis of the cost value, respectively of the new cost value, and in which the value of the overall cost, respectively the new value of the overall cost, is calculated 105, 105bis for all spectral bands and for all point sources.

[0072] According to a particular example of implementation, among the plurality of successive acquisitions of a point source, at least one parameter of the parametric model takes a value during one acquisition and another different value during at least one other acquisition, the value and the other value of said at least one parameter being unknown during the acquisition step. The variation applied between the value and the other value is not necessarily known. If the variation of the parameter is known, the estimation of the value of the parameters of the parametric model will be more robust.

[0073] According to one aspect, the invention relates to a computer program comprising a set of instructions configured to implement the method 100 according to one of the implementation modes described above, when it is executed by a processor of an image processing and control unit of an optical instrument comprising a matrix of multispectral detectors, the matrix of multispectral detectors comprising a matrix of detectors, and a plurality of matrices of spectral filters, each spectral filter matrix being associated with a spectral band.

[0074] According to another aspect, the invention relates to a computer-readable medium including the computer program described above.

Claims

Claims

1. A method (100) of measuring an aberration in an image acquired by an optical instrument comprising a multispectral detector array, the multispectral detector array comprising a detector array, and a plurality of spectral filter arrays, each spectral filter array being associated with a spectral band, the method (100) comprising the following steps: - acquisition (101) by the detector matrix of the optical instrument of an image of a known point source, the image comprising, for each spectral filter matrix, a part of an image task of the point source acquired in the spectral band of said spectral filter matrix; - for each spectral filter matrix, determining (102) an initial value (VIP) of each parameter of a set of parameters of a parametric model of a modeled impulse response of the optical instrument, the set of parameters comprising a subset of parameters common to the plurality of spectral filter matrices, and a subset of parameters specific to each spectral filter matrix; - for each spectral filter matrix, sampling (103) of an impulse response calculated on the basis of the initial value of the parameters (VIP) applied to the parametric model, the sampling being carried out according to a sampling mask associated with the spectral filter matrix, - for each spectral filter matrix, calculation (104) of a value of a cost in the spectral band, the value of the cost being a function of a distance between the calculated impulse response and the impulse response observed for each of the points of the sampling mask, - calculation (105) of a value of a global cost (CG) for all the spectral bands as a function of the value of the cost in each spectral band, - for each spectral filter matrix, estimation (102bis) of an estimated value (VEP) of each parameter of the set of parameters, the estimated value (VEP) being a function of a variation (VAR) estimated from the value of the global cost (CG), the estimated value (VEP) being equal to the estimated variation (VAR) added to the initial value of the parameters (VIP); - for each spectral filter matrix, sampling (103bis) a new impulse response calculated on the basis of the estimated value (VEP) of the parameters applied to the parametric model, the sampling being carried out according to the sampling mask associated with the spectral filter matrix, - for each spectral filter matrix, calculation (104bis) of a new value of the cost in said spectral band, the new value of the cost being a function of the distance between the new impulse response calculated and the impulse response observed for each of the points of the sampling mask, - calculation (105bis) of a new value of the overall cost (NCG) for all the spectral bands as a function of the new value of the cost in each spectral band. - comparison (106) between the new value of the overall cost (NCG) and the value of the overall cost (CG) to obtain a result of the comparison (RC);- determining (107bis) that the initial value of the parameters (VIP) is equal to the estimated value of the parameters (VEP) and that the value of the overall cost (CG) is equal to the new value of the overall cost (NCG), and for each spectral filter matrix, repeating (107) the steps between the step of estimating (102bis) an estimated value of the parameters (VEP), up to the step of comparing (106) between the new value of the overall cost (NCG) and the value of the overall cost (CG) until a stopping criterion (CRA) is satisfied by the result of the comparison (RC), so that at each iteration of the repetition step (107), during the estimation step (102bis) the variation (VAR) is estimated from an initial value of the parameters (VIP) equal to the estimated value of the parameters (VEP) estimated during the previous iteration, and from a value of the overall cost (CG) equal to a new global cost calculated during the previous iteration;- determining (108) a measure of the aberration on the basis of the estimated value of the parameters (VEP) of the impulse response. .;

2. Method according to claim 1, in which the acquisition step (101) comprises the acquisition of several known point sources and / or several acquisitions of the same known point source, the set of parameters of the parametric model comprising a subset of parameters specific to each acquisition, the parametric impulse response model being associated with the different impulse responses observed in the spectral band of each spectral filter matrix, and for each point source and / or for each acquisition of the same point source, and in which the following steps are implemented for each spectral filter matrix, and for each point source, and / or for each acquisition of the same point source: - step of determining (102) the initial value of the parameters (VIP), and step of estimating (102bis) the estimated value of the parameters (VEP), - steps of sampling (103, 103bis) the impulse response calculated on the basis of the initial value of the parameters (VIP), respectively the new impulse response calculated on the basis of the estimated value of the parameters (VEP), - steps of calculating (104, 104bis) the value of the cost, respectively the new value of the cost, and in which the value of the overall cost, respectively the new value of the overall cost, is calculated (105,105bis) for all spectral bands and for all point sources.,

3. The method of claim 1 or 2, wherein the plurality of spectral filter arrays comprises at least three spectral filter arrays, each configured to filter in a spectral band complementary to the other arrays.

4. Method according to one of claims 1 to 3, in which, in the step 102bis of estimating the estimated value (VEP) of each parameter, the variation (VAR) is estimated from a Jacobian matrix of the parametric model of the impulse response, so that the repetition (107), of the steps between the step of estimating (102bis) an estimated value of the parameters (VEP), and the step of comparing (106) the new value of the global cost (NCG) and the value of the global cost (CG), until a stopping criterion (CRA) is satisfied by the result of the comparison (RC), corresponds to an implementation of an iterative algorithm of the Levenberg-Marquardt type.

5. Method according to one of claims 1 to 4, in which the value of the overall cost, respectively the new value of the overall cost, is equal to a sum of the value of the costs, respectively the new value of the costs, over all the spectral bands.

6. Method according to one of claims 1 to 5, in which the result of the comparison is equal to a difference between the value of the new cost global and the value of the global cost, the predetermined stopping criterion being satisfied when the result of the comparison is less than a threshold.

7. Method according to one of claims 1 to 6, in which the set of parameters comprises at least one of the parameters of an impulse response representative of an optical part of the optical instrument, the parameters of an impulse response representative of the integration of the charges on at least one elementary detector of the matrix of detectors, the parameters of an impulse response representative of a streak, i.e. a linear movement, of the matrix of detectors during an exposure time, the parameters of an impulse response representative of a shake, i.e. a non-linear movement, of the matrix of detectors during the exposure time, the parameters representative of the coordinates of a geometric image of the point source, said geometric image being point-like.

8. Method according to one of claims 1 to 7, in which the measurement of the aberration is used to compensate for the aberration by an improvement processing of the image acquired by the optical instrument, the improvement processing being a function of the measurement of the aberration, the processing comprising for example an application of a compensation filter calculated from the measurement of the aberration, or the processing comprising for example a restoration of the image as a function of the measurement of the aberration.

9. A method according to one of claims 1 to 8, wherein the measurement of the aberration is used to compensate for the aberration by a control of the optical instrument to actuate an active optic of the optical instrument, the control being based on the measurement of the aberration.

10. A computer program comprising a set of instructions configured to implement the method 100 according to one of claims 1 to 9 when executed by a processor of an image processing and control unit, of an optical instrument comprising a matrix of multispectral detectors, the matrix of multispectral detectors comprising a matrix of detectors, and a plurality of matrices of spectral filters, each spectral filter matrix being associated with a spectral band.

11. A computer-readable medium comprising a computer program according to claim 10.

Images