Method of acquiring and analyzing a scene by image difference
Patent Information
- Authority / Receiving Office
- ES · ES
- Patent Type
- Patents
- Filing Date
- 2021-11-16
- Publication Date
- 2026-07-09
AI Technical Summary
Existing methods for acquiring differential images are hindered by noise from ambient light and artifacts caused by scene displacement, which degrade the quality of the extracted useful signal.
A method involving synchronized acquisition of images with and without excitation light, followed by registration and integration steps to compensate for scene displacement, and including denoising and compression to reduce noise and improve signal extraction.
The method effectively reduces noise and artifacts, enhancing the quality of the useful signal by minimizing the impact of ambient light and scene movement, allowing for clearer visualization of fluorescent or phosphorescent signals.
Smart Images
Figure 00000011_0000 
Figure 00000012_0000 
Figure 00000013_0000
Abstract
Description
[0001] The present invention relates to the field of synchronized acquisition of active images, and in particular a method for processing and analyzing differential images.
[0002] It is known to illuminate a scene with a light source and simultaneously acquire an image of that illuminated scene. The scene illumination can be ambient light and / or excitation light, chosen from a specific frequency band, such as ultraviolet, to trigger a particular light emission at the scene, for example, through fluorescence or phosphorescence. The emission of the excitation light, in the form of a flash for example, is synchronized with the triggering of an acquisition period for the specific light, of fluorescent or phosphorescent origin, so that the acquisition period can be as short as possible, proportional to the intensity of the excitation flash. This allows for a very significant reduction in the sensor's exposure to ambient light, which causes light noise superimposed on the useful fluorescent or phosphorescent signal.
[0003] To further reduce noise related to ambient light, it is also known to successively acquire a first image of the scene illuminated only by ambient light, and a second image of the scene illuminated by ambient light and by the light from the excitation flash: the first image is thus formed by the ambient light reflected by the scene, the second image being formed by the ambient light reflected by the scene to which is added useful fluorescent or phosphorescent signal produced by the excitation flash, so that by a subtraction of the first image from the second image, the useful signal can be extracted from the ambient light noise.
[0004] Several factors can nevertheless reduce the effectiveness of the known process described above. In particular, even a small relative displacement of the scene with respect to the acquisition sensor, or of the acquisition sensor with respect to the scene, will introduce artifacts at the image difference level that can impair the quality of the extraction of the useful signal.
[0005] Thus, it is possible to obtain, through specific processing of images successively acquired using the known method described above, an improvement in the quality of the useful signal obtained by comparing these images. It is also possible to optimize the effectiveness of these specific processing steps. US patent application 2019 / 166355 A1 discloses an imaging device that reduces the negative impact of ambient light on an image acquired with active illumination.
[0006] US patent application 2020 / 267299 A1 discloses an imaging device that enhances foreground regions that are overexposed by a flash, and underexposed background regions.
[0007] The invention therefore aims to provide a solution to all or part of these problems.
[0008] To this end, the present invention relates to a method of acquiring and analyzing a scene by difference of images acquired with an image sensor, as described in claim 1 below.
[0009] According to claim 1, the differential image is a difference between the first image, and the second registered image.
[0010] According to these provisions, the impact on the useful signal of the differential image of a movement between the acquisition of the first image and the acquisition of the second image will be limited.
[0011] According to one embodiment, the invention comprises one or more of the following features, alone or in a technically acceptable combination.
[0012] According to one implementation method, the excitation source is a complementary light source, or a sound excitation source, or an electrical excitation source.
[0013] Depending on the implementation method, the supplementary light source has one or more filtering or polarizing elements.
[0014] According to one implementation mode, the complementary light source is configured to emit at one or more different wavelengths, for example at a wavelength of ultraviolet, infrared, or visible light.
[0015] According to one implementation method, at least one iteration of the process steps is carried out so as to obtain at least one other differential image and in which at least one other differential image is registered with respect to the differential image before calculating an integrated differential image as a function of the differential image and at least one other registered differential image, and then replacing the differential image with the integrated differential image.
[0016] According to the invention, the step of acquiring a first image is repeated at least twice, and the step of acquiring a second image is repeated at least twice, the method further comprising: a first registration step of at least two first images, the first registration step being followed by an integration step of at least two first registered images, so as to produce a first integrated image; a second registration step of at least two second images, the second registration step being followed by an integration step of at least two second registered images, so as to produce a second integrated image in which the calculation step of a second registered image and the calculation step of a differential image are applied to the second integrated image and to the first integrated image to respectively calculate a second integrated image registered with respect to the first integrated image, then the differential image equal to a difference between the first integrated image, and the second integrated image registered.
[0017] According to one implementation method, the integration of at least two differential images, or at least two first images, or at least two second images, respectively, includes the calculation of a sum or an average of the at least two differential images, or the at least two first images, or the at least two second images.
[0018] According to these provisions, the noise is reduced.
[0019] According to one embodiment, the first image is a first pixel matrix, the second image is a second pixel matrix, and the calculation of a second registered image includes determining an offset between the first matrix and the second matrix, said offset corresponding to a pixel of the first matrix, a corresponding pixel of the second matrix, the determination of the offset comprising the following substeps: calculation of a sum, for a plurality of pixels of the first matrix, of an absolute value of a difference between the value of the pixel of the first matrix and the value of the corresponding pixel of the second image, for at least two offsets; determination that the offset is, among the at least two offsets, the one for which the calculated sum meets a predetermined criterion.
[0020] According to one implementation method, the predetermined criterion is a criterion for minimizing the calculated sum; in other words, the determined offset is the one for which the calculated sum is a minimum among the at least two offsets.
[0021] According to one implementation mode, the plurality of pixels of the first matrix is a row of pixels of the first matrix and a column of pixels of the first matrix.
[0022] According to these provisions, the registration step is particularly suited to the specificity of differential imaging, to improve the performance and speed of registration.
[0023] According to one embodiment, the sensor comprises a detector array, and a color filter array superimposed on the detector array, the sensor being configured so that the first image and the second image are raw, i.e., not demosaiced, and in which the first image and the second image are not demosaiced before the differential image calculation step.
[0024] According to these provisions, the amount of data to be processed in each image is reduced by about two-thirds, compared to the volume of data that would have to be processed if the processing was carried out on demosaiced images.
[0025] According to one implementation method, the process further includes a denoising step applied to the differential image to obtain a denoised differential image.
[0026] According to one embodiment, the denoising step includes a substep, applied to the pixels of a matrix of pixels of the differential image, of determining whether a pixel is good or bad, a pixel being determined bad on the basis of a comparison of the value of said pixel to the value of neighboring pixels of said pixel in the matrix, and a pixel being determined good if it is not determined bad, and a substep of replacing the value of said pixel with a zero value or with a value based on the value of neighboring pixels, determined good, of said pixel, when said pixel is determined bad.
[0027] In one implementation, comparing the pixel's value to the values of neighboring pixels calculates a similarity score for the pixel to the neighboring pixel values, and the pixel is determined to be bad based on a criterion applied to the calculated similarity score. For example, a pixel with a low similarity value to its neighbors is determined to be bad; similarly, a pixel with a high similarity to too few neighbors is determined to be bad.
[0028] According to these provisions, since the useful signal is weak and the noise is high compared to said useful signal, the noisy values will not be averaged, therefore not spread out by the denoising step.
[0029] According to one implementation method, a dematrixing step is not carried out before the denoising step, in order to reduce the volume of data to be processed.
[0030] According to one embodiment, the method further includes a compression step applied to the denoised differential image, to obtain a compressed denoised differential image in which each pixel of the pixel matrix of the compressed denoised differential image is coded on a number of bits less than or equal to a determined number of bits, preferably less than or equal to 8 bits.
[0031] According to one implementation method, a dematrixing step is not performed before the compression step, in order to reduce the volume of data to be processed.
[0032] According to one implementation method, the process further includes a step of displaying the compressed denoised differential image.
[0033] According to one implementation mode, the display of the compressed denoised differential image includes the use of a color code to improve the visibility of areas containing a useful signal, the useful signal being determined after sorting the pixels according to their intensity and color characteristics, for example so as to select fluorescent or phosphorescent pixels.
[0034] According to one implementation method, the compressed denoised differential image is displayed as a superposition on a contextual image, the contextual image resulting from a processing of the first or second image, to reveal information about the surrounding areas containing a useful signal, for example fluorescence or phosphorescence.
[0035] According to one embodiment, the method further includes a step of illuminating the scene or an environment of the scene by a light source in addition to the ambient light, the lighting step being synchronized so that said lighting is turned off while the steps of acquiring by the sensor a first image of the scene, of exciting the scene by the excitation source, and of acquiring by the sensor a second image of the scene are carried out.
[0036] According to these provisions, the lighting of the scene allows the process operator to work in satisfactory lighting conditions without adding ambient noise during the first and second image acquisition stages.
[0037] According to one aspect of the invention, it also relates to a computer program, comprising a set of instructions configured to implement the process according to one of the embodiments described above, when said instructions are executed by a computer.
[0038] According to another aspect of the invention, it also relates to a computer-readable medium comprising a set of instructions configured to implement the process according to one of the embodiments described above, when said instructions are executed by a computer.
[0039] For its proper understanding, an embodiment and / or implementation of the invention is described with reference to the accompanying drawings, which represent, by way of non-limiting example, one embodiment or implementation of a device and / or method according to the invention. The same reference numerals in the drawings designate similar elements or elements with similar functions. [ Fig. 1 ] is a schematic view of a device for acquiring one or more pairs of successive images; [ Fig. 2 ] is a schematic presentation of the sequence of steps in a method of implementing the process according to the invention; [ Fig. 3a ] illustrates a method for implementing a step in calculating a second, re-registered image; [ Fig. 3b ] illustrates a method for implementing a step in calculating a second, re-registered image; [ Fig. 3c ] illustrates a method for implementing a step in calculating a second, re-registered image; [ Fig. 3d ] illustrates a method for implementing a step in calculating a second, re-registered image; [ Fig. 3e ] illustrates a method of implementing a step in calculating a second, recalibrated image;
[0040] The invention relates to a method 100 for acquiring and analyzing a scene 4 by comparing images acquired with an image sensor 2, for example, a detector array. The scene 4 and the sensor 2 are schematically represented in the figure 1 , as well as a complementary light source 1. Initially, the scene 4 is illuminated by ambient light which is reflected and / or scattered by the scene before contributing to the formation of a first image of the scene, acquired by the sensor 2. In a second step, the scene is excited by an excitation source 1, for example a complementary light source 1 chosen to optically excite certain components of the scene 4, or in an area of interest of the scene 4, so as to trigger, for a period of time, a light emission, for example by a fluorescence effect, or phosphorescence, or by a Raman effect, which will contribute, in addition to the ambient light reflected or scattered by the scene 4, to the formation of a second image acquired by the sensor 2.
[0041] The acquisition 103 of the second image is synchronized with the excitation 102 of the scene so that the acquisition 103 of the second image is carried out during a part of the emission time period.
[0042] The complementary light source 1 is configured to project the excitation light onto the scene with a field at least as large as that of sensor 2. In particular, the excitation light includes light in one of the frequency bands among the frequency bands corresponding to ultraviolet, visible and infrared.
[0043] The complementary light source 1 is equipped, more particularly, with one or more filtering or polarizing elements, configured to filter or polarize the excitation light.
[0044] A central unit 3, for example a computer or a processor equipped with memory containing the appropriate instructions, is configured to control the implementation of the various steps of the process 100 according to the invention.
[0045] The method 100 according to the invention aims to exploit the difference between a first image acquired without excitation of the scene, and a second image acquired with excitation of the scene, to extract from the background light generated by the ambient lighting, before visualizing in its context, the useful signal corresponding to the weak signal emitted by the excited components of the scene 4.
[0046] The method 100 according to the invention thus makes it possible to work under ambient lighting and to extract the weak signal.
[0047] To obtain the desired result by image difference, it is essential to be able to compensate for any relative displacement of the scene 4 with respect to the sensor 2, between the acquisition of the first image and the acquisition of the second image, by calculating 104 a second registered image, before calculating 105 a differential image equal to a difference between the first image and the second registered image.
[0048] According to these provisions, the impact on the useful signal of the differential image of a movement between the acquisition of the first image and the acquisition of the second image will be limited.
[0049] Thus, with reference to the figure 2 The process 100 according to a first embodiment of the invention comprises the following steps: acquisition 101 by sensor 2 of a first image of the scene 4, the first image being formed by ambient light reflected or scattered by the scene; excitation 102 of the scene by an excitation source 1, the excitation 102 triggering light emission from at least a part of the scene, for a period of emission time; acquisition 103 by sensor 2 of a second image of the scene, the acquisition 103 of the second image being synchronized with the excitation 102 of the scene so that the acquisition 103 of the second image is carried out during a part of the emission time period, the second image being formed by the light emission from at least a part of the scene added to the ambient light reflected or scattered by the scene; calculation 104 of a second image registered with respect to the first image to compensate for a relative displacement of the scene with respect to sensor 2 between the acquisition of the first image and the acquisition of the second image, calculation of a differential image 105, the differential image being a difference between the first image, and the second registered image.
[0050] Depending on the specific implementation, the excitation source may be a complementary light source, a sound excitation source, or an electrical excitation source. Thus, those skilled in the art will understand that sound excitation can be used to modify the fluorescence of at least part of the scene, for example, to allow the localization of microstructures or cracks. Similarly, those skilled in the art will understand that electrical excitation can be used to generate a magnetic or electric field in such a way as to modify the fluorescence of at least part of the scene, for example, to allow the localization of an electrical current or potential circuit.
[0051] Several first images are acquired successively, and several second images are acquired successively during several excitation periods, using the same type of excitation, for example, around an optical wavelength of 270 nm, or around 365 nm, or around a wavelength in one of the infrared bands, etc. More specifically, during these one or more excitation periods, different types of excitation are used to produce as many successively acquired images. Each of the images from the plurality of successively acquired first images is then registered with each other by calculating as many registered first images as there are acquired first images. An integrated first image, which is an average of the registered first images, is calculated.Similarly, each of the images in the plurality of successively acquired second images is then registered with respect to the others by calculating as many registered second images as there were acquired second images. A second integrated image, which is an average of the registered second images, is calculated. The calculation step 104 of a second image registered with respect to the first image is then applied to the second integrated image and the first integrated image, respectively, to calculate a second integrated image registered with respect to the first integrated image. Finally, the calculation step 105 of the differential image is applied to the first integrated image and the registered second integrated image, to calculate the differential image equal to the difference between the first integrated image and the registered second integrated image.The number of first images acquired may, for example, be different from or equal to the number of second images acquired; the first images of the plurality of first images may be acquired successively first, before the acquisition of the second images of the plurality of second images; but optionally, it is also possible that each or some of the second image acquisitions be interspersed between two first image acquisitions.
[0052] Thus, the order in which the first and second images are acquired is not particularly important for the proper functioning of the invention. However, integration after registration of the successively acquired images, whether they are first images, second images, or differential images as described below, helps to reduce noise.
[0053] The time between the first image and the second image, or between successive acquisitions of images synchronized with successive excitations of the scene, is for example less than 30 ms, so as to limit the movement between two successive acquisitions as much as possible, while ensuring a sufficient pause time for image acquisition given the sensitivity of sensor 2.
[0054] According to one embodiment, which we will call the video mode, steps 101, 102, 103, 104, and 105 of process 100 are repeated several times in succession to obtain a plurality of differential images; an integrated differential image, for example equal to an average of the differential images of the plurality of differential images, is then calculated. The time period between two successive iterations of the acquisition steps is less than 100 ms, for example.
[0055] More specifically, the integration of at least two differential images, or at least two first images, or at least two second images, includes calculating a sum or an average of at least two differential images, or at least two first images, or at least two second images, respectively. According to these provisions, noise is reduced, especially as the number of integrated images increases.
[0056] According to an implementation method, illustrated on the figure 3 The sensor 2 comprises a detector array, and each acquired image is a pixel array. The dimensions of the pixel array for each acquired image are always the same and equal to the dimensions of the detector array. Each pixel of a first image I1, i.e., of a first pixel array, corresponds to the image of a portion of the scene around a point in scene 4, which we will subsequently refer to as the image of the point in scene 4. The value of the image pixel of a point in scene 4 is a sum of a value representing the optical flux received by the corresponding detector from said point in scene 4 and a value corresponding to detection noise specific to the detector. Thus, as illustrated in the figure 3a , the first pixel matrix I1 is the sum of a matrix S of representative values of the optical flux received by each detector from the points of scene 4, and a matrix B1 of the noise values of the detectors at the time of the acquisition of the first image.
[0057] If there is no movement, i.e., no displacement of scene 4 relative to sensor 2 between the acquisition of the first image I1 and the acquisition of the second image I2, as illustrated in the figure 3b , the second image I2 is the sum of the first matrix S, considering that the ambient lighting is identical to that used for the acquisition of the first image, and a second matrix SE of the representative values of the optical flux received by each detector from the points of the scene 4 after excitation of the latter by the excitation source, and a third matrix B2 of the noise values of the detectors at the time of the acquisition of the second image.
[0058] As illustrated on the figure 3d , a displacement of the sensor 2 relative to the scene 4 between the acquisition of the first image I1 and the acquisition of the second image I2 creates a shift, for example of one pixel to the right along a line, between the position in the first pixel matrix of the image of the point of scene 4 and the position of said point in the second pixel matrix of the second image.Thus, if the value of the pixel corresponding to said point is strongly contrasted with the value of the neighboring points of said point in the first pixel matrix, a difference D2=I2-I1 between the first pixel matrix I1 and the second pixel matrix I2 will show at the point considered a significant value corresponding to the difference between the value of the pixel at this point in the first pixel matrix and the value of the neighboring pixel in the second pixel matrix, said neighboring pixel being shifted from said pixel by a shift corresponding to the displacement of the sensor 2 with respect to the scene 4 which occurred between the acquisition of the first image I1 and the acquisition of the second image I2.
[0059] When there is no movement, as illustrated on the figure 3c , the difference D1=I2-I1 only shows the noise or the useful signal, i.e. that resulting from the excitation of the scene by the excitation source 1.
[0060] On the other hand, when there is a displacement, the difference D2=I2-I1 shows negative values whose absolute value is much greater than the noise, and / or the useful signal, at the level of the contrasted areas of the first image I1, typically corresponding to the marked contours of certain objects present in scene 4.
[0061] Thus, for example, we will note that an SLC2 sum of negative values along a row and a column of the difference matrix D1, D2 will have a much larger absolute value, for example 200, for the difference matrix D2=I2-I1 obtained following a shift created by a relative displacement between the acquisition of the first image I1 and the acquisition of the second image I2, whereas the absolute value of the SLC1 sum will be minimum, for example equal to 7, for the difference matrix D1=I2-I1 obtained when the relative displacement is zero between the acquisition of the first image I1 and the acquisition of the second image I2.
[0062] The goal of the registration step is to find a registration offset that, when applied to the second pixel matrix following the first offset, will restore a zero offset between the two matrices. In one implementation, all possible registration offsets are applied to the second pixel matrix I2, and for each applied offset, an SLC2 sum of the negative values along a row and column of the difference matrix D2 is calculated; the desired registration offset is then determined by selecting the one for which the absolute value of the SLC2 sum is minimum.
[0063] Choosing the SLC2 sum of negative values along a row and column of the difference matrix D2 is just one example. Those skilled in the art will understand that the SLC2 sum of negative values can be calculated on any other subset of the difference matrix D2.
[0064] The calculation step 104 of a second image registered with respect to the first image to compensate for a relative displacement of the scene with respect to sensor 2 between the acquisition of the first image and the acquisition of the second image, then consists of applying the registration offset to the second image I2, after having determined said registration offset as indicated above.
[0065] In other words, according to an implementation method illustrated on the figures 2 And 3 , the determination of the offset for calculation step 104 of a second registered image comprising the following sub-steps: calculation 10411, for a plurality of pixels of the first matrix, of an absolute value of an SLC2 sum of a difference between the value of the pixel of the first matrix and the value of the corresponding pixel of the second image, for at least two offsets; determination 10412 that the offset is, among the at least two offsets, the one for which the calculated sum meets a predetermined criterion, for example a criterion for minimizing the calculated sum, i.e. the determined offset is the one for which the calculated sum is minimum among the at least two offsets.
[0066] According to these provisions, the registration step is particularly suited to the specificity of differential imaging, to improve the performance and speed of registration.
[0067] Another particular way of optimizing the overall performance of the process 100 according to the invention, where the sensor 2 comprises a detector array, and a color filter array superimposed on the detector array, for example in the case of a color image sensor 2, is to configure the sensor 2 so that the first image and the second image are raw, in other words undemosaiced, and in which the first image and the second image are not demosaiced before the differential image calculation step.
[0068] Thus, the amount of data to be processed in each image is reduced by about two-thirds, compared to the volume of data that would have to be processed if the processing was carried out on demosaiced images.
[0069] According to a more particular embodiment of method 100 according to the invention, method 100 further includes a denoising step 106 applied to the differential image to obtain a denoised differential image.
[0070] According to one embodiment, the denoising step includes a substep, applied to the pixels of the differential image, of determining whether a pixel is good or bad; a pixel is determined bad on the basis of a comparison of the value of said pixel to the value of neighboring pixels of said pixel in the matrix; conversely, a pixel is determined good if it is not determined bad; when said pixel is determined bad, the value of said pixel is replaced by a zero value or by a value depending on the value of neighboring pixels, determined good, of said pixel.
[0071] In particular, comparing a pixel's value to the values of neighboring pixels involves, for example, calculating a similarity score based on the pixel's value relative to its neighbors. The pixel is then classified as poor based on a criterion applied to this similarity score. For instance, a pixel with a low similarity in value to its neighbors is classified as poor; similarly, a pixel with a high similarity to too few neighbors is classified as poor.
[0072] According to these provisions, since the useful signal is weak and the noise is high compared to said useful signal, the noisy values will not be averaged, therefore not spread out by the denoising step.
[0073] According to one implementation method, the dematrixing step is not carried out before the denoising step, in order to reduce the volume of data to be processed by the denoising step.
[0074] According to a more particular embodiment of method 100 of the invention, method 100 further comprises a compression step 107 applied to the denoised differential image, to obtain a compressed denoised differential image in which each pixel of the pixel matrix of the compressed denoised differential image is encoded on a number of bits less than or equal to a predetermined number of bits, preferably less than or equal to 8 bits. In one embodiment, a dematrixing step is not performed before the compression step, so as to reduce the volume of data to be processed.
[0075] According to a more particular embodiment of method 100 of the invention, method 100 further comprises a display step 108 of the compressed denoised differential image. In particular, the display of the compressed denoised differential image includes the use of a color code to improve the visibility of areas containing a useful signal, the useful signal being determined after sorting the pixels according to their intensity and color characteristics, for example, so as to select fluorescent or phosphorescent pixels. More particularly, the compressed denoised differential image is displayed superimposed on a contextual image, the contextual image resulting from processing the first or second image, to reveal information about the surrounding areas containing a useful signal, for example, fluorescence or phosphorescence.
[0076] According to one embodiment of the method 100 of the invention, the method further comprises a step of illuminating the scene 4 or an environment of the scene 4 with a light source in addition to the ambient light. The lighting step is synchronized so that said lighting is off during the acquisition 101 by the sensor 2 of a first image, and during the excitation 102 of the scene by the excitation source 1, and during the acquisition 103 of a second image. According to these provisions, the scene illumination allows the operator of the method to work under satisfactory lighting conditions without adding ambient noise during the acquisition steps of the first and second images.
[0077] According to one aspect of the invention, it also relates to a computer program, comprising a set of instructions configured to implement the process according to one of the embodiments described above, when said instructions are executed by the central unit 3.
[0078] According to another aspect of the invention, it also relates to a computer-readable medium comprising a set of instructions configured to implement the process according to one of the embodiments described above, when said instructions are executed by the central unit 3.
Claims
1. A method (100) for acquiring and analyzing a scene (4) by difference of images acquired with an image sensor (2), the method (100) comprising the following steps: - acquisition by the sensor (2) of a plurality of first images (101) of the scene (4), each of the first images being formed by ambient light reflected or scattered by the scene; - excitation (102) of the scene by an excitation source (1), the excitation (102) triggering light emission from at least one portion of the scene, during an emission time period; - acquisition (103) by the sensor (2) of a plurality of second images of the scene, the acquisition of each second image (103) being synchronized with the excitation (102) of the scene so that the acquisition of the second image (103) is carried out during a portion of the emission time period, said second image being formed by the light emission from the at least one portion of the scene added to the ambient light reflected or scattered by the scene; - first registration of the first images with each other, the first registration step being followed by a step of integrating the first images registered with each other, so as to produce a first integrated image, the first integrated image being an average of the first images registered with each other, - second registration of the second images with each other, the second registration step being followed by a step of integrating the second images registered with each other, so as to produce a second integrated image, the second integrated image being an average of the second images registered with each other, - calculation (104) of a second integrated image registered with respect to the first integrated image, then of a differential image equal to a difference between the first integrated image and the second registered integrated image.
2. The method (100) according to claim 1, wherein at least one iteration of the steps of the method (100) is carried out so as to obtain at least one other differential image, and wherein the at least one other differential image is registered with respect to the differential image before calculating an integrated differential image based on the differential image and the at least one other registered differential image, and then replacing the differential image with the integrated differential image.
3. The method (100) according to any of the preceding claims, wherein the first image is a first pixel array, the second image is a second pixel array, and the calculation of a second registered image (104) comprising the determination of an offset (1041) between the first array and the second array, said offset causing a pixel of the first array to correspond with a corresponding pixel of the second array, the determination of the offset comprising the following sub-steps: - calculation (10411) of a sum, for a plurality of pixels of the first array, of an absolute value of a difference between the value of the pixel of the first array and the value of the corresponding pixel of the second image, for at least two offsets; - determination (10412) that the offset is, among the at least two offsets, the one for which the calculated sum meets a predetermined criterion.
4. The method (100) according to any of the preceding claims, wherein the sensor (2) comprises a detector array and a color filter array superimposed on the detector array, the sensor (2) being configured so that the first image and the second image are raw, in other words non-demosaiced, and wherein the first image and the second image are not demosaiced before the step of calculating the differential image.
5. The method (100) according to any of the preceding claims, further comprising a denoising step (106) applied to the differential image to obtain a denoised differential image.
6. The method (100) according to claim 5, further comprising a compression step (107) applied to the denoised differential image, to obtain a compressed denoised differential image in which each pixel of the pixel array of the compressed denoised differential image is encoded on a number of bits less than or equal to a determined number of bits, preferably less than or equal to 8 bits.
7. The method (100) according to claim 6, further comprising a step of displaying (108) the compressed denoised differential image.
8. The method (100) according to any of the preceding claims, further comprising a step of illuminating the scene (4) or an environment of the scene by a light source in addition to the ambient light, the illumination step being synchronized so that said illumination is switched off while the steps of acquiring by the sensor (2) a first image (101) of the scene (4), of exciting (102) the scene by the additional light source (1), and of acquiring by the sensor (2) a second image (103) of the scene (4) are performed.
9. A computer program, comprising a set of instructions configured to implement the method according to any of claims 1 to 8, when said instructions are executed by a computer.
10. A computer-readable medium comprising a set of instructions configured to implement the method according to any of claims 1 to 8, when said instructions are executed by a computer