PROCESSING METHOD OF AN IMAGE

Description

TECHNICAL FIELD

[0001] The present invention relates to a method for processing a raw image collected by a detector provided with a matrix of bolometers. In particular, the processing method according to the present invention is intended to correct non-uniformities due to dispersions of the characteristics of the bolometers of the detector.

[0002] The present invention also relates to a computer program capable of implementing all the steps of the method according to the present invention.

[0003] The invention finally relates to a detector comprising bolometers and a computer implementing the computer program. STATE OF THE PRIOR ART

[0004] An infrared detector known from the state of the art generally comprises bolometers organized in a matrix manner according to n rows and m columns.

[0005] When exposed to a scene for image acquisition, these bolometers, which are sensitive to the scene's temperature, see their electrical resistance vary. In other words, the current flowing in each of the bolometers depends on the scene temperature, but also on the ambient temperature.

[0006] In particular, the measurement S_(i,j) of the bolometers Pix_(i,j) of a matrix of bolometers evolves according to the following law: S P i j = Resp T amb T s c é ne − T amb + S 0 , T amb

[0007] Or : T amb is the ambient temperature, and more specifically the temperature of the imager; T scene is the temperature of the scene seen by the bolometer, Resp(T amb ) is the responsivity of the bolometer, the latter depends on the ambient temperature; S 0,Tamb is the output value of the bolometer for a scene temperature equal to the ambient temperature.

[0008] The term Resp(T amb ) depends on the materials used for the design of the bolometer as well as on its architecture.

[0009] Deriving the scene temperature requires knowing the ambient temperature T amb , so the detector is usually also equipped with a temperature sensor.

[0010] The detector can also be equipped with additional bolometers, called blind bolometers, which are not exposed to the scene. The current flowing in these bolometers then depends only on the ambient temperature.

[0011] Thus, according to such a configuration, the determination of the resistance variation of an exposed bolometer is based on a differential measurement between the currents flowing in said exposed bolometer and a blind bolometer.

[0012] In general, each column of the bolometer matrix is ​​associated with a blind bolometer which is implemented for each of the bolometers of said column during the differential measurement. However, other configurations can be envisaged, and in particular the pooling of a single blind bolometer for several columns of bolometers.

[0013] The raw image ( figure 1 ) of a scene that can be obtained with such a device is generally not usable, and requires additional processing.

[0014] In particular, the image illustrated in the figure 1 reveals the positioning of the detector bolometers, and more specifically a non-uniformity (“pixelation effect”). This effect originates from the significant dispersion of electrical resistances from one bolometer to another.

[0015] The image also has a columnar appearance which is due to the dispersion of electrical resistances between the blind bolometers.

[0016] In order to overcome these problems, different solutions have been considered, some of which are described in documents US20070120058 and WO0184118.

[0017] In particular, it has been proposed to implement a mechanical shutter on the detector. In particular, the mechanical shutter is placed in front of the detector so as to collect a reference image relative to the ambient temperature, which is subsequently subtracted from the image of the scene.

[0018] This arrangement, relatively simple in principle, is however not satisfactory.

[0019] Indeed, the implementation of a shutter, and the motorization associated with it, poses both cost and space problems.

[0020] Furthermore, the reference image must be refreshed whenever the ambient temperature varies.

[0021] Alternatively, it was proposed to characterize the temperature response of the detector, and in particular of each of its bolometers.

[0022] This characterization includes reference measurements at different temperatures with all of the detector's bolometers sealed with a shutter.

[0023] The reference measurements then make it possible to determine the temperature evolution of each of the bolometers and thus construct calibration tables stored in a memory space of the detector.

[0024] Thus, in operation, the detector corrects the raw image by subtracting, for each bolometer, the values ​​obtained by interpolation from the calibration tables.

[0025] This solution, which makes it possible to reduce the effect of non-uniformity of the image from one bolometer to another, is however not satisfactory.

[0026] Indeed, the procedure for acquiring reference measurements is long, and generates additional manufacturing costs for the detector.

[0027] Furthermore, the memory space dedicated to saving calibration tables, due to the cost associated with it, is not desirable.

[0028] Finally, a third method based on algorithms to correct image non-uniformity has been proposed in documents [1] and [2] cited at the end of the description.

[0029] These methods known from the state of the art are also not satisfactory.

[0030] Indeed, these methods are generally tedious to implement, and their robustness is questionable.

[0031] Furthermore, these methods require the implementation of heavy calculation means which penalize the cost of the detectors in which they are implemented.

[0032] An aim of the present invention is to propose a method of processing an image collected by a detector provided with bolometers which is simpler than the techniques known from the state of the art, and which does not require the use of mechanical parts or the robotization associated with them.

[0033] Another aim of the invention is to propose a method for correcting the columnar effect observed on a raw image. STATEMENT OF THE INVENTION

[0034] The aims of the present invention are, at least in part, achieved by a method of treatment as defined in claim 1.

[0035] According to one implementation mode, the detector includes a memory space dedicated to saving the calibrated components S_mask_cal_(i,j).

[0036] According to one implementation mode, the calibrated components S_cal_(i,j) are saved in a memory space of the detector.

[0037] According to one mode of implementation, the method further comprises the following steps: d) a calculation, for each of the columns, of a column term Col_mask_(j) corresponding to the average value of the measurements S_(i,j) of the masked bolometers of the column considered; e) a correlation between the column terms Col_mask_(j) and the calibrated coordinates C_mask_cal_(j) of the masked bolometers determined according to the relation: C _ mask _ cal _ j = 1 n ∑ k = 1 n S _ mask _ ref _ k j ; f) an additional correction of the corrected image to form a final image, the additional correction comprising the calculation of final measurements S_fin_(i,j) of the final image from the corrected measurements S_Cor_(i,j) and on the basis of the result of correlation step e).

[0038] According to one embodiment, the correlation step e) comprises a linear regression so that the column terms Col_mask_(j) and the calibrated coordinates C_mask_cal_(j) verify the following relationship: Col_mask_ j = α . C_mask_cal_ j + Col offset where α and Col offset are the terms determined during the linear regression.

[0039] According to one implementation mode, the additional correction step f) comprises the calculation of the final measurements S_fin_(i,j) for each bolometer according to the following relation: S_fin_ i j = S_Cor_ i j − α . Col_cal_ j − Col offset where, for a bolometer (Pix_(i,j)) of a given column, Col_cal_(i,j) is a calibrated coordinate determined according to the relation Col _ cal _ j = 1 n ∑ k = 1 n S _ ref _ k j .

[0040] According to one embodiment, the detector is provided with a lens mounted on a diaphragm which closes off bolometers masked at the corners of the detector.

[0041] According to one embodiment, the detector also comprises blind bolometers, each blind bolometer being used for the differential measurement of the bolometers of at least one column of bolometers specific to it, advantageously each blind bolometer is associated with a single column of bolometers.

[0042] The invention also relates to a computer program, which when executed by a computer, leads to implementing the method according to the present invention.

[0043] The invention also relates to a device comprising: a detector provided with a plurality of bolometers arranged in a matrix manner in n rows and m columns, the plurality of bolometers comprising bolometers exposed to the scene, and masked, closed bolometers, a computer provided with the computer program according to the present invention.

[0044] The invention also relates to the implementation of the device according to the present invention for detection, in particular the detection of people, in a room. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Other characteristics and advantages will appear in the following description of a method for processing an image, given as non-limiting examples, with reference to the appended drawings in which: there figure 1 is a raw image of a scene obtained by a detector equipped with bolometers arranged in a matrix of 80 rows by 80 columns; the figure 2is a schematic representation of a detector provided with a lens mounted on a diaphragm capable of being implemented according to the present invention; the figure 3 is a reference image collected by the detector shuttered by a shutter maintained with the detector at a reference temperature; the figure 4 is a representation of the components of a vector, called the column reference vector, calculated from the reference image of the figure 3 ; there Figure 5 is a representation of the components of a matrix, called the column reference matrix, determined from the reference image of the figure 3 and the vector of the figure 4 ; there figure 6 is a representation of the components of a vector, called a masked reference vector calculated from the reference image of the figure 3 ; there figure 7is a representation of the components of a matrix, called a masked reference matrix, determined from the reference image of the figure 3 and the vector of the figure 6 ; there figure 8 is a graphical representation of the values ​​of the masked terms S_mask_norm_(i,j) (on the vertical axis) as a function of the masked calibrated components S_masque_cal_(i,j) (horizontal axis); the Figures 9a and 9b are images, respectively, raw and uniformity corrected according to the present invention; the figure 10 is a graphical representation of the values ​​of the masked term C_mask_norm_(i,j) (on the vertical axis) as a function of the masked calibrated component C_mask_cal_(i,j); the figure 11 represents a final image; the figure 12 illustrates a correction of the columnar aspect only of the raw image of the Figure 9a . DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS

[0046] The present invention will now be described in relation to the figures 1 to 11 .

[0047] There figure 2 represents a detector provided with a plurality of bolometers, denoted Pix_(i,j), arranged in a matrix manner along n rows (denoted “L i”) and m columns (denoted “C j”).

[0048] A bolometer indexed i, j corresponds to a bolometer placed at the intersection of row i with column j.

[0049] Among the bolometers Pix_(i,j), we can distinguish the bolometers, called exposed Pix_Exp_(i,j), exposed to the scene intended to be imaged and the bolometers, called masked bolometers Pix_mask_(i,j), which are closed.

[0050] The masked bolometers Pix_mask_(i,j) are for example closed by the diaphragm on which is mounted a lens 2 arranged between the scene and the detector 1. The masked bolometers Pix_mask_(i,j) are, according to this configuration, arranged in the corners of the detector. The invention is however not limited to this arrangement, and the masked bolometers Pix_mask_(i,j) can for example form complete columns of bolometers, in particular the columns bordering the matrix of bolometers.

[0051] In other words, masked bolometers are blocked when an image is acquired.

[0052] The detector of the figure 2 may also include a plurality of blind bolometers Bol_Blind_(j).

[0053] The detector 1 further comprises a computer 4 provided with a calculation processor intended to execute the different steps of the method according to the present invention. The computer may also comprise a memory space for saving raw, reference, or intermediate measurements.

[0054] Finally, the detector 1 may comprise a temperature probe 5 intended to evaluate the temperature of the environment in which said detector is located. The temperature probe may, for example, comprise a PN junction.

[0055] The method according to the present invention proposes to correct the defects of a raw image collected by the detector 1.

[0056] In this regard the figure 1 , discussed in the “PRESENTATION OF THE PRIOR ART” section, exhibits uniformity defects and a columnar appearance.

[0057] The non-uniformity is mainly due to a dispersion of the electrical resistances of the bolometers forming the detector.

[0058] Darker areas can also be observed at the corners of the image. These dark areas are due to the obstruction of the masked bolometers Pix_mask_(i,j) by the diaphragm on which the lens is mounted, and will be used to correct the non-uniformities observed on the raw image. The raw measurement S_(i,j) of a masked bolometer Pix_mask_(i,j) is noted in the rest of the statement as masked raw measurement S_M_(i,j).

[0059] A brighter central area, corresponding to the raw measurements S_(i,j) (denoted raw exposed measurement S_E_(i,j)) of exposed bolometers Pix_Exp_(i,j), can be observed.

[0060] Finally, the figure 1also reveals a columnar aspect which is due to the dispersion of the electrical resistances of the blind bolometers Bol_Blind_(j). Each blind bolometer Bol_Blind_(j) is, in this respect, implemented for the differential measurement of the bolometers of at least one column of bolometers which is specific to it, advantageously each blind bolometer Bol_Blind_(j) is associated with a single column C j of bolometers Pix_(i,j).

[0061] Note that if all columns were associated with the same blind bolometer, the columnar effect would not be observed. However, this configuration is not desirable since a failure of this single blind bolometer would render the entire detector non-functional.

[0062] The method according to the present invention comprises determining the response S_(i,j), denoted S_ref_(i,j), of the bolometers Pix_(i,j) when they are closed by a shutter maintained with the detector at a temperature, called reference temperature Tr, for example at 20°C.

[0063] The responses S_Ref_(i,j) can be obtained by a simple measurement by the detector with a shutter placed in front of the bolometers Pix_(i,j). In this respect, the figure 3 represents a reference image collected by the detector.

[0064] Based on these reference measurements, a vector, called the column reference vector V_ref col (represented in figure 4 ) of components C_ref col (j) (for j ranging from 1 to m) is determined.

[0065] The components C_ref col_(j) are determined by the following relation: C _ ref _ col _ j = 1 n ∑ i = 1 n S _ Ref _ i j

[0066] In other words, each component C_ref col_(j) is the average of the reference measurements S_Ref_(i,j) of column j.

[0067] A matrix, called column reference matrix M_Pix_col_cal (represented in Figure 5 ), of calibrated components S_col_cal_(i,j) each associated with a bolometer Pix_(i,j), can also be determined.

[0068] The components S_col_cal_(i,j), for each of the bolometers Pix_(i,j) of column j, are then calculated as follows: S _ col _ cal _ i j = S _ ref _ i j − C _ ref _ col _ j

[0069] In other words, the calibrated component S_col_cal_(i,j) of a bolometer Pix_(i,j) of a column j of the detector, corresponds to the reference measurement of said bolometer from which we subtract the average of the reference measurements S_ref_(i,j) of all the bolometers of column j.

[0070] The columnar calibration matrix M_Pix_col_cal, as described in the rest of the statement, is advantageously implemented to correct the columnar aspect of the raw images collected by the detector.

[0071] In a complementary or alternative manner, always on the basis of the reference measurements S_ref_(i,j), a vector, called the masked reference vector V_ref_pix (represented in figure 6 ), of components C_mask_cal_(j), can be determined.

[0072] The components C_mask_cal_(j) are determined by the following relation:

[0073] Where the data S_mask_ref_(i,j) correspond to the responses S_Ref_(i,j) of the masked bolometers Pix_mask_(i,j) only.

[0074] In other words, each component C_mask_cal_(j) is the average of the reference measurements S_Ref_(i,j) associated with the masked bolometers of column j.

[0075] A matrix, called the mask reference matrix M_Pix_mask_cal (represented in figure 7 ), of components S_mask_cal_(i,j) each associated with a masked bolometer Pix_mask_(i,j) can also be determined.

[0076] The calibrated component S_mask_cal_(i,j), for each of the masked bolometers Pix_mask_(i,j) of column j, is determined according to: S _ mask _ cal _ i j = S _ mask _ ref _ i j − C _ mask _ cal _ j

[0077] In other words, the calibrated component S_mask_cal_(i,j) of a bolometer Pix_mask_(i,j) of a column j of the detector, corresponds to the reference measurement of said bolometer from which we subtract the average of the reference measurements of all the masked bolometers of column j.

[0078] The establishment of the masked calibration matrices M_Pix_mask_cal and columnar M_Pix_col_cal is implemented during the manufacture of the detector, and their components are advantageously saved in a dedicated memory space of the detector, for example in the memory space of the computer 4.

[0079] The method according to the present invention comprises a step a) of calculating terms, called masked terms S_mask_norm_(i,j) of a matrix, called masked matrix.

[0080] In particular, each masked term S_mask_norm_(i,j) is associated with a masked bolometer Pix_mask_(i,j).

[0081] The determination of the masked term S_mask_norm_(i,j) relating to a masked bolometer Pix_mask_(i,j) of column Cj, includes the subtraction of the average of the masked raw measurements of column Cj (noted S_M_(j)) from the masked raw measurement S_M_(i,j).

[0082] In particular, each masked term S_mask_norm_(i,j) of column j is determined according to the following relations: S _ M _ j = 1 n ∑ k = 1 n S _ M _ k j S _ mask _ norm _ i j = S _ M _ i j − S _ M _ j

[0083] The step of determining the masked terms S_mask_norm_(i,j) is followed by a step b) of correlation between said masked terms S_mask_norm_(i,j) and the masked calibrated components S_mask_cal_(i,j) of the masked bolometers Pix-mask_(i,j).

[0084] There figure 8 represents in particular for each masked bolometer Pix-mask_(i,j), the value of the masked terms S_mask_norm_(i,j) (on the vertical axis) as a function of the masked calibrated components S_mask_cal_(i,j). In this figure, the masked term S_mask_norm_(i,j) varies linearly as a function of the masked calibrated component S_mask_cal_(i,j).

[0085] The correlation between these two terms then includes a linear regression according to the relationship: S_mask_norm_ i j = β . S_mask_cal_ i j + D offset where β and D offset are the terms determined during the linear regression.

[0086] These then make it possible to correct, during a correction step c), the raw measurements S_(i,j) of each bolometer according to the relation: S_Cor_ i j = S _ i j − β . S_col_cal_ i j − D offset

[0087] THE Figures 9a and 9b represent the effect of the correction described above. In particular, the Figure 9a is a raw image, while the Figure 9b illustrates the effect of image non-uniformity correction. If the columnar appearance still appears on the Figure 9b , the latter reveals a character previously undetectable in the raw image.

[0088] The method according to the present invention can also implement a correction of the columnar aspect. The latter can be carried out independently of the uniformity correction presented above.

[0089] The method according to the present invention may further comprise, during a step d), a calculation, for each of the columns Cj, of a column term Col_mask_(j) corresponding to the average value of the measurements S_(i,j) of the masked bolometers Pix_mask_(i,j) of the column considered.

[0090] These column terms Col_mask_(j) can then be correlated, during a step e), to the calibrated coordinates C_mask_cal_(j) of the masked bolometers Pix_mask_(i,j).

[0091] There figure 10 represents in particular for each masked bolometer Pix_mask_(i,j), the value of the column terms Col_mask_(j) (on the vertical axis) as a function of the calibrated coordinates C_mask_cal_(j) of the masked bolometers Pix_mask_(i,j) of column j. In this figure, the term Col_mask_(j) varies linearly as a function of the calibrated coordinate C_mask_cal_(j).

[0092] The correlation can include a linear regression so that the column terms Col_mask_(j) and the calibrated coordinates C_mask_cal_(j) verify the following relationship: Col_mask_ j = α . C_mask_cal_ j + Col offset where α and Col offset are the terms determined during the linear regression.

[0093] The method according to the present invention may then comprise an additional correction (step f) of the corrected image to form a final image.

[0094] The additional correction includes in particular the calculation of final measurements S_fin_(i,j) of the final image from the corrected measurements S_Cor_(i,j) and on the basis of the result of the correlation step.

[0095] For example, the additional correction includes the calculation of the final measurements S_fin_(i,j) for each bolometer Pix_(i,j) according to the following relation: S_fin_ i j = S_Cor_ i j − α . Col_cal_ j − Col offset

[0096] The image thus corrected presented at the figure 11 no longer has a columnar appearance.

[0097] It is understood that the columnar and non-uniform aspects of the raw image have distinct origins and can therefore be corrected independently.

[0098] In particular, the order of implementation of these corrections may be reversed.

[0099] In this regard, the figure 12 illustrates a correction of the columnar aspect only of the raw image of the Figure 9a .

[0100] The invention also relates to a computer program, which when executed by a computer or a calculator, leads to implementing the method according to the present invention.

[0101] The present invention makes it possible to differentiatedly correct the columnar and non-uniform aspects of the raw images of detectors provided with a matrix of bolometers.

[0102] In particular, the present invention only requires a reference measurement at a reference temperature in order to correct the raw images.

[0103] The present invention can advantageously be implemented for detection, and in particular the detection of people, in rooms. REFERENCES

[0104] [1] EP2940991B1 [2] US2010237245A1

Claims

1. A method, executed by a calculator, for processing a raw image, characterized by raw measurements S_(i,j) collected by bolometers Pix_(i,j) of a detector (1) arranged in a matrix-like fashion according to n rows (Li) and m columns (Cj), the bolometers Pix_(i,j) comprising exposed bolometers (Pix_exp_(i,j)), and masked bolometers (Pix_mask_(i,j)), veiled during the acquisition of the raw measurements, the method comprising: a) calculating masked terms S_mask_norm_(i,j), associated with the masked bolometers (Pix_mask_(i,j)), the calculating consisting in subtracting a given column (Cj) from the raw measurement S_(i,j) of each masked bolometer (Pix_mask_(i,j)), the mean value of the raw measurements S_(i,j) of the masked bolometers of said column; b) correlating between the masked terms S_mask_norm_(i,j) and calibrated components S_mask_cal_(i,j) of the masked bolometers (Pix_mask_(i,j)), determined according to the relationship: S _ mask _ cal _ i j = S _ mask _ ref _ i j − 1 n ∑ k = 1 n S _ mask _ ref _ k j the terms S_mask_ref_(i,j) being reference measurements of the masked bolometers (Pix_mask_(i,j)) obtained at a reference temperature (Tr) and by masking the detector with a mask also maintained at the reference temperature (Tr), the correlation step comprising a linear regression so that each masked term S_mask_norm_(i,j) and the calibrated components S_mask cal_(i,j) comply with the following relationship: S _ mask _ norm _ i j = β . S _ mask _ cal _ i j + D offset where β and Doffset are the terms determined during the linear regression; c) a step of correcting the raw image which comprises the calculation of corrected measurements S_Cor_(i,j) of a corrected image for each bolometer (Pix_(i,j) on the basis of the result of the correlation step b), this correction step comprising the calculation of the corrected measurements S_Cor_(i,j) for each bolometer (Pix_(i,j)) according to the following relationship: S _ Cor _ i j = S _ i j − β . S _ cal _ i j − D offset where, for a given bolometer (Pix_(i,j)), S_cal_(i,j) is a calibrated component determined according to the relationship: S _ cal _ i j = S _ ref _ i j − 1 n ∑ k = 1 n S _ ref _ k j the terms S_ref (i,j) being reference measurements of the bolometers (Pix_(i,j)) at a reference temperature (Tr) and by masking the detector with a mask also maintained at the reference temperature (Tr).

2. The method according to claim 1, wherein the detector comprises a memory space dedicated to saving the calibrated components S_mask cal_(i,j).

3. The method according to one of claims 1 or 2, wherein the calibrated components S_cal_(i,j) are saved in a memory space of the detector.

4. The method according to one of claims 1 to 3, wherein the method furthermore comprises the following steps: d) calculating, for each of the columns (Cj), a column term Col_mask_(j) corresponding to the mean value of the measurements S_(i,j) of the masked bolometers Pix_mask_(i,j) of the considered column; e) correlating between the column terms Col_mask_(j) and calibrated coordinates C_mask_cal_(j) of the masked bolometers (Pix_mask_(i,j)), determined according to the relationship: C _ mask _ cal _ j = 1 n ∑ k = 1 n S _ mask _ ref _ k j ; f) additionally correcting the corrected image to form a final image, the additional correction comprising the calculation of final measurements S_fin_(i,j) of the final image from the corrected measurements S_Cor_(i,j) and on the basis of the result of the correlation step e).

5. The method according to claim 4, wherein the correlation step e) comprises a linear regression so that the column terms Col_mask_(j) and the calibrated coordinates C_mask_cal_(j) comply with the following relationship: Col _ mask _ j = α . C _ mask _ cal _ j + Col offset where α and Coloffset are the terms determined during the linear regression.

6. The method according to claim 5, wherein the additional correction step f) comprises the calculation of the final measurements S_fin_(i,j) for each bolometer (Pix_(i,j)) according to the following relationship: S _ fin _ i j = S _ Cor _ i j − α . Col _ cal _ j − Col offset where, for a bolometer (Pix_(i,j)) of a given column (Cj), Col_cal_(i,j) is a calibrated coordinate determined according to the relationship Col _ cal _ j = 1 n ∑ k = 1 n S _ ref _ k j .

7. The method according to one of claims 1 to 6, wherein the detector is provided with a lens mounted on a diaphragm which veils masked bolometers (Pix_mask_(i,j)) at the corners of the detector (1).

8. The method according to one of claims 1 to 7, wherein the detector also comprises blind bolometers (Bol_Blind_(j)), each blind bolometer (Bol_Blind_(j)) being implemented for a differential measurement of the bolometers of at least one column of bolometers which is specific thereto, advantageously each blind bolometer (Bol_Blind_(j)) is associated with one single column (Cj) of bolometers (Pix_(i,j)).

9. A computer program which, when it is executed by a computer, causes the implementation of the method according to one of claims 1 to 7.

10. A device comprising: - a detector provided with a plurality of bolometers arranged in a matrix-like fashion according to n rows (Li) and m columns (Cj), the plurality of bolometers comprising bolometers (Pix_exp_(i,j)) exposed to the scene, and veiled masked bolometers (Pix_mask_(i,j)), - a calculator fitted with the computer program according to claim 9.

11. An implementation of the device according to claim 10 for detection, in particular the detection of a person, in a room.