HIGH AND LOW LEVEL IMAGING PROCESS AND SYSTEM

The field of the invention is that of imaging and more particularly low light level imaging.

In the visible domain, the human being is endowed with a capacity of day vision which decreases for decreasing levels of illumination. In a night environment, the different sky conditions, mainly modulated by the phase of the Moon and the cloud cover, allow ground illumination of between 1 lux and 0.1 mlux, which is also qualified by night levels (d 'after the Aero 790 40 standard) ranging from night 1 (starting at 1 lux) to night 5 (ending at 0.1 mlux) respectively. The 0.1 mlux level is the ultimate level of illumination to achieve and improve for vision. An operational benefit is the increased nighttime image quality level 5. However, under weak lighting conditions, very few photons are received from the observed scene. When the received flow decreases, the signal-to-noise ratio (SNR) decreases until it becomes insufficient for vision. The signal is then drowned in the noise generated by the detector and the image coming from the detector is unusable or very degraded. Thus, the sensitivity and noise of the detector limit performance for imaging at very low light levels.

Generally, Low Light Level imagers are optimized by increasing the pixel area and increasing the integration time as far as possible.

Currently, improving night-time performance can also go through the use of low read noise and dark technologies and, for example, by using an amplifying medium introducing multiplication gains on the signal. We can cite :

■ ILCMOS: Light intensifier tube (IL tube) (with micro-channel plate or MCP Micro Channel Plate) optically coupled to a CMOS matrix;

■ EBCMOS: Photocathode + CMOS matrix (bombarded by photoelectrons) integrated in a tube;

■ EMCCD: CCD matrix with an Electron Multiplier which gives an amplification by avalanche on the multiplexed signal;

■ APD array of photodiodes (Avalanche PhotoDiode) either in linear mode (multiplying gain) or in Geiger counter mode (triggering of a current pulse on the generation of a photoelectron).

For scientific applications such as in astronomy or bio-medical imaging, where we know that we will receive very few photons of what we want to observe, we use detection and counting imaging. photons; this technique was introduced many years ago first with the use of photomultipliers in the processing of the time signal.

Photon counting imaging is based on the principle that the number of photons arriving at a pixel array detector during a given exposure time is a stochastic process. It is characterized by a Poisson distribution of mean m r and standard deviation Vp p . As a function of the average number of photons per pixel m r , we can calculate the probability P that n p photons are detected during the exposure time t ex (taken into account by m r ). This law is defined by the relation:

The Law in figure 1 presents the Poisson distribution for different mean values ​​of photons per pixel. Thus, for example, by exploiting the Poisson law, for an average flux of 0.15 photon per pixel, we have 86% of the pixels which do not see any photon, 13% which see one and 1% see more than 1 photon. . Thus, if we place ourselves in this particular regime of the Poisson law with on average 0.15 photons per pixel (or less), then we can constrain a statistical exploitation of the received signal knowing a priori that either we are not receiving any photon (0 photon) or we only receive 1 in 99% of cases per period of exposure time. It is therefore possible to estimate the signal in a binary manner while only being wrong in less than 1% of the cases for this example of the mean flux less than or equal to 0.15 photons per pixel. For higher average fluxes we will have an increasingly important error rate with confusion of situations where the pixel counts one photon while it has collected 2 or more. The reasoning that we have just had on the reception of photons by a pixel also applies to the photoelectrons generated by the photoelectric detection process implemented in the detector.

The analog signal at the output of the detector is fluctuating and is dominated by read noise and dark noise for a pixel which has not generated a photoelectron, and multiplication noise for pixels which have converted 1 or more photoelectrons. However, knowing a priori the number of photons received, less than 0.15 on average, forces us to output two states: either we have received 0 photons or we have received 1, in 99% of cases. We can therefore replace the fluctuating analog signal by binarizing an elementary frame or an elementary acquisition of the pixel at 0 or 1. This can be done for example by a thresholding algorithm: below a certain threshold, we consider that we received 0 photons, the analog fluctuation of the initial signal being dominated by the various noise of the detector and not by the true amplitude linked to the incidence of 0 photons. Beyond the threshold, it is considered that the detection of 1 and of a single photon has indeed taken place and this “unit” amplitude replaces the analog amplitude linked to the fluctuation of the reading noise and the multiplication of the signal. The binarization of the frame by this method however leads to a residual noise linked to the error of the photon counting rate. This residual noise can be characterized by a probability of counting error which can be estimated and still bounded by using noise statistics and Poisson's law. This probability breaks down into the probability of counting 1 for 0 incident photon,

Finally, the last step of photon counting imaging consists of summing several binary elementary frames to reconstitute an image corresponding to the output image reproduction rate, for example for a video output rate of 25 Hz, i.e. 40 ms accumulation time of elementary frames obtained at a higher rate. The sum of the different binarized frames makes it possible to reconstitute an image with different levels of gray. Number of

discernible shades of gray is no longer dominated by the dispersion of the multiplication gains of the signal amplification process which is at the origin of the sensation of snow dominating the image and its nuances.

It is important to note that if the scene flux increases above 0.15 average photons per elementary frame, then we go out of this constrained regime of 0 or 1 photon to have more and more cases where the pixel sees 2 photons or more. In this case, however, the amplification noise factor of the detector no longer makes it possible to estimate the exact number of photons received with a good level of confidence. In other words, we are not able to know if we have received 1, 2, 3 or more photons. We lose information. This is an important limitation of photon counting imaging with current charge multiplication detectors.

It is important to note that this amplification noise figure is different for different types of detectors. In SPAD (Single Photon Avalanche Diode) or avalanche photodiode technologies in Geiger mode, the absorption of a photon causes a giant avalanche process. The current generated by the avalanche is detected by an electronic circuit which allows the detection of the photon. However, after such a process, a dead time is necessary to evacuate the charges and therefore prevents the detection of a new photon during this evacuation time. It is not this type of technology which is preferred and presented here.

For certain applications in astronomy (using for example a camera from the company Nüvü or from the company First Light Imaging) or in bio-medical imaging, photon counting imaging is well indicated because it is known that we will receive only a few photons. On the other hand, for vision applications of any night scene, it is not known in advance which lighting conditions one will be able to meet and which scene dynamics will be encountered in the video image (spatially and temporally). If the scene flux is strong while performing photon counting imaging, we are outside the zone of the illumination regime which makes it possible to binarize the signal of each pixel in the frame;p (standard deviation for an average level of n p (for example n p > 10 photons per frame). On the other hand, if the scene flux is too weak while doing conventional imaging, the noise of the detector will then be preponderant and the image unusable.

Consequently, there remains to this day a need for a method and a high dynamic range imaging system, adapted to unforeseen lighting conditions and which can vary from a strong scene flow to a weak scene flow.

The imaging system according to the invention comprises several imaging modes, with at least:

■ A classic imaging mode used when the scene flow is important. This mode and the application conditions of this mode assume a high photon flux, greater than the square of the detector reading noise and its dark current. In which case the detector will generate an image limited by the photonic noise. If the detector allows it (as in EMCCDs) the use of a signal amplification or multiplication mode is not necessary and in this case the S / N is typically close to the root of the photon flux accumulated by pixel and per frame. If the photon flux decreases and becomes close to the dark current or to the square of the read noise, then an amplification or multiplication process can be initiated on the signal to limit the degradation of the signal to noise ratio.

■ A “natural” photon counting imaging mode, enabling imaging at very low light levels. At very low flux, the detectors are limited by the reading noise of the receiver. In which case it is relevant to use a device for amplifying or multiplying photoelectrons. As a result, in certain detectors the multiplied photoelectric signal becomes dominant with respect to the read noise of the matrix or even to the dark current. On the other hand, the multiplication factor is stochastic and brings a noise called amplification noise factor which will degrade the S / N. It has been shown that if this flux is less than 0.15 photoelectrons per pixel and per frame, then a thresholding logic makes it possible to constrain the analog signal delivered by

every pixel. We are in this case in photon counting imaging mode that we define as "natural". This mode makes it possible to improve the image quality which can be defined by its S / N and a greater dynamic range of gray levels restored.

And when the scene flux is intermediate, that is to say decreases without reaching a very low light level threshold suitable for “natural” photon counting, then we can force the imaging system to switch to a mode. specific imaging with “forced” photon counting, by placing oneself in conditions allowing to be in this regime constrained to 0 or 1 photon per pixel for a frame. In this case it is relevant to examine whether on the one hand it is judicious to activate the function of amplification or multiplication of the signal to reduce the influence of the reading noise and the dark current of the sensor that we have seen in classic imaging mode. We are then in the classic multiplied or amplified imaging mode. Nevertheless in this mode of multiplication we will see a degradation of the S / N and the dynamics of the image while going down in the levels of illumination. It is then possible then, paradoxically, to reduce the photon flux per pixel and per frame to enter a mode similar to natural photon counting in order to improve the image quality by its S / N and its dynamic compared to the multiplied mode. This sets the forced photon count imaging mode.

We can consider as a first approach that:

- when the scene flux is less than 0.15 photo electron per pixel and per frame, the imaging mode is natural photon counting,

- when the scene flux is greater than a few (for example 5) photo electron per pixel and per frame, the imaging mode is conventional,

- when the scene flow is intermediate, the imaging mode is forced photon counting.

The system is forced to place itself in an imaging mode by controlling certain parameters such as:

The integration time to have, for example, on average 0.15 photon / pixel / frame or less, if the photon counting imaging mode is chosen. This value 0.15 can be adjustable and depends on the chosen binarization algorithm; we can have for example 0.2 photon / pixel / frame by tolerating for example a slightly larger counting error,

the acquisition rate,

the digital aperture of the optics,

optical attenuation,

the reception spectral band, for example to improve spectral discrimination,

the pixel pitch or its fill factor if the technology allows it, for example to improve the angular resolution or the MTF, etc ...

More precisely, the subject of the invention is a method for imaging a scene by means of an imaging system making it possible to obtain an image of the scene, the imaging system comprising a device for acquiring frames according to acquisition parameters, and a unit for processing the acquired frames, characterized in that the method comprises the following steps:

A) Choice of an imaging mode from among a conventional imaging mode, a natural photon counting imaging mode and a forced photon counting imaging mode, and depending on the imaging mode chosen, determination by the processing unit of the corresponding acquisition parameters,

B) acquisition of at least one frame by the acquisition device configured with said acquisition parameters, and transmission of the acquired frames to the processing unit to obtain an image, the image being obtained at the end of the sub -following steps if the chosen imaging mode is the natural photon counting imaging mode or the forced photon counting imaging mode:

- binarization of acquired frames and

- sum of binarized frames to obtain an image,

C) estimate of the quality of the image obtained,

D) D1) depending on the quality of the image obtained, determine one or two new imaging mode (s) chosen from among the conventional imaging mode, the imaging mode natural photons and the forced photon counting imaging mode,

D2) repeat steps A, B, C and D with the new imaging mode chosen as imaging mode.

According to one embodiment of the invention, step D1) is carried out by comparison (1 st comparison) of the quality of the image obtained with a first predetermined quality corresponding to the conventional imaging mode:

- if the comparison (1 st comparison) is favorable, repeating steps A, B, C and D with the conventional imaging mode as the chosen imaging mode,

- otherwise comparison (2nd comparison) of the quality of the image obtained with a second predetermined quality corresponding to the natural photon counting imaging mode, and

o if the comparison (2nd comparison) is favorable, reiteration of steps A, B, C and D with the natural photon counting imaging mode as the chosen imaging mode, o if not, reiteration of steps A, B, C and D with forced photon counting imaging mode as the chosen imaging mode.

As recalled in the preamble, without photon counting mode, the Low Light Level imagers are optimized, for very low fluxes, by increasing the surface area of ​​the pixel and increasing the integration time as far as possible. According to the invention, the introduction of the photon counting mode (natural and forced) makes it possible to favor the reduction of the size of the pixels and / or to reduce the integration time by increasing the frame rate in order to use this mode in a greater stage dynamics.

The parameters of the acquisition device are typically parameters of rate time and / or time of integration and / or opening of the optics and / or variation of spectral band and / or variation of optical attenuation. .

The acquisition parameters according to the conventional imaging mode and the parameters according to the natural photon counting imaging mode

typically have fixed values, and the parameters according to the forced photon counting imaging mode typically have values ​​varying from one iteration to another to optimize the image quality criterion at each iteration.

According to one characteristic of the invention, the method comprises a step of spatial division of each frame into sub-frames and the steps of the imaging modes are applied to each sub-frame.

The estimate of the quality of the image obtained can be determined by calculation or by an operator.

In addition, such a method can advantageously be used in active imaging, that is to say with laser illumination. In this case, the parameters of the acquisition device can advantageously lead to reducing the laser illumination of the scene or even to increasing its range.

Thus according to another characteristic of the invention, the method comprises a preliminary stage of illumination of the scene by means of an illumination device synchronized with the acquisition device and the illumination device and the acquisition device. have the same spectral band. In this case, illumination parameters can also be defined in step A.

The illumination parameters are for example parameters of power and / or duration of illumination of the illumination device.

The subject of the invention is also a system for imaging a scene which comprises a device for acquiring frames of the scene, optionally a device for illuminating the scene, and a processing unit connected to the acquisition device. and to the possible illumination device, configured to implement the imaging method as described.

Thus, instead of undergoing the received scene stream, the imaging system according to the invention adapts its operating parameters, such as for example its acquisition rate, to be in the imaging mode exhibiting the best performance.

Other characteristics and advantages of the invention will become apparent on reading the detailed description which follows, given by way of nonlimiting example and with reference to the appended drawings in which:

FIG. 1 diagrammatically represents examples of Poisson distributions, in this case histograms of the photon fluxes acquired on a matrix, for different mean values ​​m r , FIG. 2a diagrammatically represents a first example of an imaging system according to the invention, FIG. 2b schematically represents a second example of an imaging system according to the invention, with an active imaging system,

FIG. 3 is an example of a flowchart showing different steps of the method according to the invention.

An example of an imaging system 100 according to the invention is described with reference to FIG. 2a. He understands :

A device 1 for acquiring frames of the scene comprising: an optical device 12 for forming images on a focal plane and

o in the focal plane, a detection matrix 11 which generates the acquisition of frames, and

■ a processing unit 2 connected to the acquisition device 1.

The imaging system can operate in several imaging modes, with at least:

A classic imaging mode used when the scene flow is important,

A so-called natural photon counting imaging mode, allowing imaging at very low light levels, i.e. when the scene flux is low,

A so-called forced photon counting imaging mode, allowing imaging when the scene flow is intermediate between high and low.

Thus, when the scene flow is important, we keep a classic imagery. When the scene flux is low, we switch to natural photon counting imaging. On the other hand, when the scene flow is

intermediate, then the imaging system is forced to switch to a specific mode of forced photon counting imaging, by placing the imaging system under conditions allowing it to be in this regime constrained to 0 or 1 photon per pixel for a frame: the imaging system is constrained by playing on one or more of its various parameters (integration time, rate, aperture, illumination for active imaging, etc.) to be able to be in the particular mode where from the point of view of the restored image quality, the photon counting can be carried out advantageously.

The imaging system is forced to be placed in one or the other imaging mode by controlling the parameters of the acquisition device such as:

- the acquisition rate of the detection matrix 11, or the acquisition period,

- the integration time of the detection matrix 11,

- opening of the image forming device 12,

the optical attenuation or the variation of the spectral band by choosing a filter 13 from among filters of different attenuations or different widths of spectral bands respectively.

One can for example in a preferred way:

■ Adjust the frame acquisition rate.

■ Control the integration time to have an average of 0.15 photon / pixel / frame or less, if the photon counting imaging mode is chosen. This value 0.15 depends on the chosen binarization function and on the error rate allowed on the restored image.

We can also play on other parameters (example: aperture, variable optical attenuation, illumination of the scene in active imaging, etc.) to artificially limit the flow from the scene and thus force the system to be in d mode. photon counting imaging. These parameters are controlled by the processing unit 2.

Thus, instead of undergoing the received scene stream, the imaging system according to the invention adapts its operating parameters, such as for example its acquisition rate, to be in the imaging mode exhibiting the best performance.

Switching between the different imaging modes is done according to the scene stream received or more precisely according to the quality of the image obtained according to one or the other imaging mode, this quality being linked to the stream stage received. It can be done automatically by being provided by the processing unit 2 or else manually. It can be done on the whole of the detection matrix 1 1 (-> on each frame acquired) or only on a part (^ sub-frame or window), even if the matrix allows it, per pixel, which allows to adapt the imaging mode to the stream received as finely as possible.

If the scene flux is relatively low then the imaging system adapts its parameters to remain in a photon count imaging mode.

If the scene flux is too high and the system can no longer be forced to be in the specific regime of the photon counting mode or if the performance is better in conventional imaging then we switch back to a conventional imaging mode where the photons are accumulated for an integration time equal to that of the final image (40 ms in the case of a video at 25 Hz for example).

In low light level imaging, to increase the sensitivity or the image quality, it is conventional to increase the integration time per frame and / or increase the pixel area and / or increase the band. spectral reception. According to the invention, in order to use the photon counting mode, the parameters of the acquisition device are controlled by favoring the increase in the frame rate and thus the reduction of the integration time, the limitation or even the reduction of the frame rate. spectral band.

The method according to the invention also makes it possible to use a matrix of detectors the size of the pixels of which is reduced, even if it means using binning modes (as in CCDs) which make it possible to sum the charges on adjacent pixels (2x2, 3x3 or 4x4) for example to effectively increase the size of the pixels.

Steps of an example of an imaging method according to the invention will be described in relation to FIG. 3.

A) Determination by the processing unit 2, of the acquisition parameters of the acquisition device 1 according to a chosen imaging mode. The parameters of the conventional imaging mode and those of the natural count imaging mode are predetermined and have predetermined values. Those of the parameters of the natural counting imaging mode correspond to the ultimate conditions of use of the acquisition device, corresponding for example to 0.15 photon / pixel or less for each frame.

The parameters of the forced counting imaging mode are variable: the parameters may vary from one image to another and their values ​​are optimized from one image to another to optimize the image quality criterion from one image to another. the other. They are generally optimized by the processing unit within a range of values, proceeding step by step from one iteration to another. These ranges of values ​​lie between the values ​​of the parameters in classic mode and those of the parameters in natural counting mode.

B) Acquisition of at least one frame by the acquisition device configured according to the current imaging mode and transmission of these frames to the processing unit as they are acquired.

If the current imaging mode (chosen in step A) is the conventional mode, an image is obtained from these frames by the processing unit in a conventional manner (acquisition of at least one frame and summation of the frames. frames to get the final image).

If the current imaging mode (chosen in step A) is the natural count or forced count imaging mode, the following substeps are carried out by the processing unit:

- Binarization of acquired frames,

- Sum of the binarized frames to form an image.

For the binarization of the acquired frames, we can use a threshold binarization function (we assign 1 to a pixel when the amplitude measured on this pixel is greater than a predetermined threshold, 0 otherwise), but other binarization functions can be considered as described in the following publications:

■ - E. Lantz et al “Multi-imaging and Bayesian estimation for photon counting with EMCCDs” Mon. Not. R. Astron. Soc, 386, 2262-2270 (2008)

■ - KBW Harpsoe et al “Bayesian photon counting with EMCCDs”, A&A, 537 (2012).

At the output of the processing unit, the image obtained according to one or the other mode is sent for example to a display device.

C) Estimation by the processing unit 2 of the quality of the image obtained.

D) Comparison (1st comparison) with a first predetermined image quality (quality1) corresponding to an image quality in conventional mode, by the processing unit. If the comparison (1st comparison) is favorable to the conventional imaging mode (the quality of the image obtained is for example given by the level of illumination of the scene measured in the Visible spectral range of between 100,000 lux and 0 , 01 lux, or less depending on the optical configuration and the performance of the detector), steps A, B, C and D are reiterated by choosing the conventional imaging mode (which may be the current mode). This illumination level criterion can also be carried out at the pupil entrance or on the focal plane, it can be measured in W / m 2 or in photons / s / m 2in the spectral range of the filters 13, for example in the near IR or even in the SWIR. If the comparison (1st comparison) is unfavorable to the conventional imaging mode, the quality of the image obtained is compared (2nd comparison) with a second predetermined image quality (quality2), corresponding to an image quality in natural counting mode, by means of the processing unit. If the comparison (2nd comparison) is favorable to the natural counting mode (the quality of the image obtained is for example between illuminations of 1 mlux and 1 plux), steps A, B are repeated,

therefore with an optimization of the acquisition parameters in step A. When the quality 1 and / or possibly the quality 2 are not predetermined, it is possible to acquire new frames according to two imaging modes (conventional and with forced counting, or at forced counting and natural counting) and between the two, the imaging mode which gives the best quality criterion is chosen. This amounts to defining these quality criteria in a relative manner.

As an example of a quality criterion, a level of illumination incident on the scene or again at the pupil entrance or else in the focal plane has been given; it is of course possible to take other criteria known to those skilled in the art such as the signal-to-noise ratio (temporal or spatial), the dynamics in restored gray levels.

The initial imaging mode can be either mode. This initial mode is engaged without knowing the lighting level of the received scene stream. In the example of FIG. 3, the conventional mode was chosen as the initial mode symbolized by a dotted arrow. But if we choose an imaging mode by photon counting (natural or forced), the comparison tests for qualities 1 and 2 are of course reversed to keep the acquisition parameters as long as it is not necessary. to change them.

In other words, the method comprises the following steps.

A) Choice of an imaging mode from among a conventional imaging mode, a natural photon counting imaging mode and a forced photon counting imaging mode, and determination by the processing unit 2, acquisition parameters of the acquisition device 1 according to a chosen imaging mode. The parameters of the conventional imaging mode and those of the natural count imaging mode are predetermined and have predetermined values. Those of the parameters of the natural count imaging mode correspond to the ultimate conditions of use of the acquisition device, corresponding for example to 0.15 photons / pixel or less for each frame.

The parameters of the forced counting imaging mode are variable: the parameters may vary from one image to another and their values ​​are optimized from one image to another to optimize the image quality criterion from one image to another. the other. They are generally optimized by the processing unit within a range of values, proceeding step by step from one iteration to another. These ranges of values ​​lie between the values ​​of the parameters in classic mode and those of the parameters in natural counting mode.

B) Acquisition of at least one frame by the acquisition device configured according to the current imaging mode and transmission of these frames to the processing unit as they are acquired.

If the current imaging mode (chosen in step A) is the conventional mode, an image is obtained from these frames by the processing unit in a conventional manner (acquisition of at least one frame and summation of the frames. frames to get the final image).

If the current imaging mode (chosen in step A) is the natural count or forced count imaging mode, the following substeps are carried out by the processing unit:

- Binarization of acquired frames,

- Sum of the binarized frames to form an image.

At the output of the processing unit, the image obtained according to one or the other mode is sent for example to a display device.

C) Estimation by the processing unit 2 of the quality of the image obtained.

D) D1) depending on the quality of the image obtained, determine one (or two as indicated previously) new imaging mode chosen from among the conventional imaging mode, the imaging mode with natural photon counting and the forced photon count imaging mode.

D2) repeat steps A, B, C and D with the (or each) new imaging mode chosen as imaging mode.

According to a variant, prior to the steps for obtaining images, the acquired frames are spatially divided into sub-frames or windows, this spatial division possibly going so far as to consider in each frame a fraction of the pixels (window, rows, columns or still pixels determined). The steps of the different imaging modes are then applied to each sub-frame.

This imaging method is in particular well suited to active imaging which then comprises a preliminary stage of illumination of the scene by an illumination device (laser, LED, lamp) 200 shown in FIG. 2b generally synchronized with the device d acquisition 1. We will then restrict the spectral band of the acquisition device to that of the illumination device and thus count only the photons of the scene reflected in this band. It is also possible to reduce the power or the illumination times of the illumination device. The spectral band, the power or the illumination durations are then illumination parameters also controlled by the processing unit 2.

CLAIMS

1. Method of imaging a scene using an imaging system

(100) making it possible to obtain an image of the scene, the imaging system comprising a device (1) for acquiring frames according to acquisition parameters, and a unit (2) for processing the acquired frames, characterized in that the process comprises the following steps:

A) Choice of an imaging mode from among a conventional imaging mode, a natural photon counting imaging mode and a forced photon counting imaging mode, and depending on the imaging mode chosen, determination by the processing unit of the corresponding acquisition parameters,

B) acquisition of at least one frame by the acquisition device configured with said acquisition parameters, and transmission of the acquired frames to the processing unit to obtain an image, the image being obtained at the end of the sub -following steps if the chosen imaging mode is the natural photon counting imaging mode or the forced photon counting imaging mode:

- binarization of acquired frames and

- sum of binarized frames to obtain an image,

C) estimate of the quality of the image obtained,

D) D1) depending on the quality of the image obtained, determine one or two new imaging mode (s) chosen from among the conventional imaging mode, the imaging mode natural photons or the forced photon counting imaging mode,

D2) repeat steps A, B, C and D with the new imaging mode chosen as imaging mode.

2. Method of imaging a scene according to the preceding claim, characterized in that the new imaging mode chosen from step D1) is obtained by comparison (1 st comparison) of the quality of the image obtained at a predetermined first quality

(quality) corresponding to the classic imaging mode:

- if the comparison (1st comparison) is favorable, the new imaging mode chosen is the classic imaging mode,

- otherwise comparison (2nd comparison) of the quality of the image obtained at a second predetermined quality (quality2) corresponding to the natural photon counting imaging mode, and o if the comparison (2nd comparison) is favorable, the new mode imaging mode chosen is the natural photon counting imaging mode,

otherwise, the new imaging mode chosen is the forced photon counting imaging mode.

3. Imaging method according to one of the preceding claims, characterized in that the acquisition parameters are parameters of acquisition rate and / or integration time and / or opening and / or variation. spectral band and / or optical attenuation variation.

4. Imaging method according to one of the preceding claims, characterized in that the acquisition parameters according to the conventional imaging mode and the parameters according to the natural photon counting imaging mode have fixed values, and in that the acquisition parameters according to the forced photon counting imaging mode have values ​​varying from one iteration to another.

5. Imaging method according to one of the preceding claims, characterized in that it comprises a step of spatial division of each frame into sub-frames and in that the steps of the imaging modes are applied to each sub-frame. frame.

6. Imaging method according to one of the preceding claims, characterized in that the estimate of the quality of the image obtained is determined by calculation or by an operator.

7. Imaging method according to one of the preceding claims, characterized in that it comprises a preliminary stage of illumination of the scene by means of an illumination device (200) synchronized with the acquisition device ( 1) and in that the illumination device and the acquisition device have the same spectral band.

8. Imaging method according to the preceding claim, characterized in that the illumination device has illumination parameters defined in step A.

9. Imaging method according to the preceding claim, characterized in that the illumination parameters are parameters of power and / or duration of illumination of the illumination device.

10. System (100) for imaging a scene which comprises a device (1) for acquiring frames of the scene and a processing unit (2) connected to the acquisition device, configured to implement the method. imaging device according to one of claims 1 to 6.

11. System (100) for imaging a scene which comprises an illumination device (200) of the scene, a device (1) for acquiring frames of the scene and a processing unit (2) connected to the scene. illumination device and the acquisition device, configured to implement the imaging method according to one of claims 7 to 9.