Image capture using radiation sensitive elements with memory effect

Abstract

A method for capturing an image which makes it possible to at least partially correct the memory effects of a sensitive element (1) of a matrix (10a) used to capture the image. The corrected image is formed by subtracting from the captured new raw image a portion of a previous raw image captured before the new raw image. The method is particularly suitable for sensitive elements having a first-order transfer function with respect to time, such as bolometers or microbolometers. Correcting the memory effects makes it possible to improve the transfer function and / or reduce the tail effects present in the image when scene elements move.

Description

Technical Field

[0001] The present description relates to a method for capturing images using a sensitive element exhibiting a memory effect, and an image sensor implementing the method. Background Art

[0002] There are a wide variety of image sensors that vary according to the type of radiation to which they are sensitive and their mode of operation used to capture images.

[0003] In this specification, radiation is understood to mean all types of external signals that can originate from a certain scene and reach an image sensor so that images of said scene can be captured by the sensor, each image being in the form of a collection of intensity values ​​respectively assigned to points of the image arranged in a matrix. Specifically, radiation can be electromagnetic radiation in any wavelength range, in particular X-rays, ultraviolet radiation, visible light, near infrared and the so-called thermal infrared range. It can also be acoustic radiation, in particular in the ultrasound field.

[0004] The image sensor comprises a matrix of sensitive elements which are individually capable of generating a detection signal which varies as a function of the intensity of the radiation received by the sensitive elements.There can thus be a one-to-one correspondence between points of the image and sensitive elements of the image sensor.

[0005] For an image sensor of the first type, the acquisition of the detection signal is achieved by exposing each sensitive element to radiation originating from the scene for a determined duration, usually called an accumulation period. Next, the detection signal that has been generated by the radiation in the sensitive element during the accumulation period is read by a dedicated circuit. The sensitive element is then reset between the reading of the detection signal carried out to capture an image and the beginning of a new accumulation period dedicated to capturing a subsequent image. This reset ensures that the consecutively captured images correspond to separate accumulation periods. This description does not take into account this first type of sensitive element or image sensor.

[0006] The present specification considers an image sensor for which each sensitive element can be in a variable instantaneous state, the variable instantaneous state depending on the radiation received by this sensitive element. In the case of this second type of sensor, an image is captured by reading the values ​​of the characteristics of the corresponding instantaneous states of the sensitive elements as a sensor matrix. Many times, the reset of each sensitive element is not possible and is not implemented between two consecutive captured images. In this case, the detection signal read from each sensitive element in order to capture the image can depend not only on the radiation received by this sensitive element at the readout time, but also on the radiation that has been previously received by the same sensitive element. In other words, the detection signal read is derived from the combination of the radiation intensities that have been continuously received by the sensitive element until the readout time. This effect is usually called "memory effect" by those skilled in the art, and due to this memory effect, each captured image is derived from a combination of scene states that have occurred continuously. When the scene changes over time, the memory effect causes a degradation of the image quality, which may take different forms depending on the nature of the patterns in the scene and the movement of some of these patterns. Specifically, the memory effect of the sensitive elements of the second type of image sensor can be in the form of reduced image contrast, tail effects of moving elements that affect the scene, etc.

[0007] Methods have been developed to reduce this memory effect, which consist in applying an erasing sequence to at least one of the sensitive elements of the matrix sensor between the moments of capturing two consecutive images. Such erasing sequences are applied by dedicated electronic circuits sometimes called thermalizing circuits. Such methods are hardware-based in the sense that they act on the sensitive elements of the image sensor.

[0008] Other methods have also been developed to reduce memory effects. They are software-based and consist in combining together several raw images directly derived from successive readings of the sensitive element. Temporal filtering functions are generated via convolution and / or linear combinations using matrix or scalar coefficients. However, the determination of these coefficients is a difficult task, which may result in a reduction of the information contained in the image and / or may amplify image noise. Moreover, such filtering operations are not particularly suitable for correcting the memory effects of the sensitive element of the image sensor. The resulting correction is therefore not optimal for the memory effect.

[0009] Technical issues

[0010] Therefore, it is an object of the present invention to reduce image degradation caused by memory effects for image sensors of the second type presented above. Specifically, it is an object of the present invention to improve the transfer function of the image sensor and / or to reduce the trailing effect present in the image when elements of the scene are moving. Summary of the invention

[0011] In order to achieve this or another object, a first aspect of the present invention proposes a method for capturing images, wherein several images are captured consecutively using the same matrix of sensitive elements, each sensitive element exhibiting a memory effect, which makes the original detection signal of the sensitive element depend on the amount of radiation received by this sensitive element at the read-out time of reading the original detection signal, and also on the amount of radiation received by the same sensitive element before the read-out time.

[0012] According to the invention, in order to form an image at least partially corrected for memory effects (referred to as the corrected image), intensity values ​​of points of the image are individually assigned to each of the sensitive elements of the matrix, this value being proportional to the difference between a raw detection signal of the sensitive element read for a new captured image (referred to as the new raw image) and a portion of the raw detection signal of the same sensitive element read for another image captured before the new raw image (referred to as the previous raw image).

[0013] By means of this correction, for each raw image consisting of raw detection signals read from a sensitive element, the image degradation caused by the memory effect of the sensitive element is reduced in a corrected image consisting of image point intensity values. The previous raw image is used to evaluate at least a part of the memory effect content, and this part of the memory effect content is subtracted from the content of the new raw image individually for each sensitive element. The resulting corrected image thus corresponds mainly to the radiation received by each sensitive element between two readout times of the raw image.

[0014] This correction of memory effects (which consists in using an image consisting of image point intensity values ​​calculated according to the invention instead of a raw image consisting of, for example, raw detection signals read out of a sensitive element) is more efficient when a "new" and a "previous" raw image are captured with a short duration between the readout times of the sensitive elements. For this purpose, when several previous raw images have been captured before a new raw image, the previous raw image used for calculating the image point intensity values ​​of the corrected image may preferably be the last of the previous raw images captured before the new raw image in the chronological order of the capture of the raw images.

[0015] In a preferred embodiment of the invention, the image point intensity value assigned to each of the sensitive elements in order to form the corrected image may be proportional to the difference between the raw detection signal of the sensitive element read for the new raw image and the result of multiplying the raw detection signal of the same sensitive element read for the previous raw image by exp(-Δ / τ). exp(.) specifies an exponential function, τ is the characteristic response time of the sensitive element, and Δ is the non-zero duration between the readout times of the sensitive element for the new raw image and the previous raw image, respectively. The new raw image and the previous raw image are so combined to produce the corrected image more appropriately when the operation of each sensitive element corresponds to a first-order transfer function with respect to time or can be roughly described by such a first-order transfer function with respect to time. In the context of the present invention, the transfer function of a radiation sensitive element, denoted f(s), is the quotient of the Laplace transform of the raw detection signal produced by the sensitive element in response to the incident radiation by the Laplace transform of the intensity of the incident radiation. In other words: f(s)=A d (s) / A R (s), where s specifies the Laplace variable, f is the transfer function of the sensitive element, and A R specifies the Laplace transform of the intensity of the radiation incident on the sensitive element, and A d Specifies the Laplace transform of the raw detection signal produced by the sensing element. This transfer function is called first order with respect to time when it has the following form: Where G is the gain factor and τ is the characteristic response time fixed for each sensitive element. The memory effect correction provided by the preferred embodiment of the present invention is particularly suitable for sensitive elements with a first-order transfer function with respect to time. However, it is also effective for sensitive elements that do not have a first-order transfer function with respect to time, but only provides partial correction of the memory effect.

[0016] For these preferred embodiments of the invention, the image point intensity value assigned to each sensitive element in order to form the corrected image may be proportional to the result of dividing the difference between the raw detection signals of the sensitive element read for the new raw image by [1-exp(-Δ / τ)] and the result of multiplying the raw detection signal of the same sensitive element read for the previous raw image by exp(-Δ / τ). The division by [1-exp(-Δ / τ)] may in particular avoid an intensity decay of the corrected image, which may be greater when the duration Δ between the respective readout times of the new and the previous raw images is shorter. Optionally, an additional scaling factor may be applied in addition to 1 / [1-exp(-Δ / τ)] in order to adjust the scaling of the image point intensity values ​​of the image corrected for memory effects. This additional scaling factor may be constant in the sense that it does not depend on the duration Δ between the respective readout times of the new and the previous raw images, nor on the characteristic response time τ.

[0017] In various embodiments of the present invention (including the aforementioned preferred embodiments), at least one of the following additional features may optionally be reproduced alone or by combining several of them together:

[0018] - the duration Δ between the corresponding readout times of the same sensitive element for a new raw image and a previous raw image, respectively, may be less than the characteristic response time τ of this sensitive element;

[0019] the duration Δ between the readout times of the same sensitive element for a new raw image and for a previous raw image, respectively, may be between 0.02 and 0.2 times the characteristic response time τ of this sensitive element;

[0020] - Each sensitive element may be a bolometer or microbolometer, a thermopile, a pyroelectric sensor, a ferroelectric sensor or a thermal deformation microsensor;

[0021] - the sensitive elements of the matrix may be sensitive to electromagnetic radiation including infrared radiation (particularly thermal infrared), X-rays, sound radiation (particularly ultrasonic radiation), etc.;

[0022] - the same value of the characteristic response time τ may be common to all sensitive elements of the matrix;

[0023] - the method may comprise a preliminary step during which a value of a characteristic response time τ is determined individually for each of the sensitive elements of the matrix. The value of the characteristic response time τ thus determined for each of the sensitive elements may then be used to calculate the image point intensity value assigned thereto for forming the corrected image;

[0024] - Each corrected image may be formed from a pair of original images, the pair of original images comprising a new original image and a previous original image that has been captured before the new original image. Thus, the pairs of original images used to form consecutive corrected images may be disjoint, non-interlaced, and temporally consecutive; and

[0025] - Several corrected images can be formed by combining the same new raw image with several previous raw images captured successively before the new raw image. In this case, each corrected image is obtained by using, for a duration Δ associated with each sensitive element, the difference between the readout times of this sensitive element, respectively for the new raw image and the previous raw images, which are used to form this corrected image. The gradual reduction of the tail effect can then be observed by comparing the corrected images obtained in this way. Possibly, this reduction of the tail effect can make it possible to assess the speed of travel of a specific element of the imaged scene in motion.

[0026] A second aspect of the present invention provides an image sensor, comprising:

[0027] a matrix of sensitive elements, each sensitive element exhibiting a memory effect causing the raw detection signal read from this sensitive element to depend on the amount of radiation received by the sensitive element at a readout time at which the raw detection signal is read, and also on the amount of radiation received by the same sensitive element before said readout time, and

[0028] - an image processing module adapted to output images at least partially corrected for memory effects, each corrected image being formed by image point intensity values ​​respectively assigned to sensitive elements of the matrix, the image processing module being adapted to calculate the image point intensity value of any one of the sensitive elements as proportional to the difference between a raw detection signal of the sensitive element read for a new raw image and a portion of a raw detection signal of the same sensitive element read for a previous raw image captured before the new raw image.

[0029] This image sensor may be suitable for implementing the method according to the first aspect of the invention, optionally comprising its preferred modes of implementation and its optional additional features. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The characteristics and advantages of the invention will become more apparent in the following detailed description of some examples of non-limiting embodiments, with reference to the accompanying drawings, in which:

[0031] [ Figure 1 ] is a block diagram of an image sensor according to the present invention;

[0032] [ Figure 2 ]arrive[ Figure 4 ] shows different sequences for producing corrected images according to the present invention;

[0033] [ Figure 5 ] shows a first advantage of the present invention taking into account image contrast; and

[0034] [ Figure 6 ] shows the second advantage of the present invention of taking into account the tail effect that may exist in the image. DETAILED DESCRIPTION

[0035] according to[ Figure 1], the image sensor 10 comprises a matrix 10a of sensitive elements 1 which are arranged at the intersections of the rows and columns of the matrix and which can be regarded as independent of one another. For example, the matrix 10a can consist of 320×240 sensitive elements 1. All sensitive elements 1 can be identical, in particular of a type known to a person skilled in the art. For example, each sensitive element 1 can be a microbolometer. It thus comprises a portion of an electrically conductive material whose resistance value varies depending on its temperature. This portion of variable resistance material is at least partially thermally insulated from its environment, so that the radiation R received and absorbed by this portion causes its temperature to increase and therefore causes its resistance value to vary. This resistance value constitutes the raw detection signal read for each captured image. This measurement principle is well known and therefore does not need to be described again here.

[0036] It is also known that this type of microbolometer has a first-order transfer function with respect to time, characterized by the value of the gain coefficient denoted G and by the value of the characteristic response time denoted τ. The transfer function denoted by f(s) and dependent on the Laplace variable s is thus Under illumination conditions with an intensity of radiation R varying sinusoidally with time according to frequency ν, the detection signal generated by the sensitive element also varies with time at frequency ν, with the amplitude being given by the formula A d =f(j2π·ν)·A R Given, where A R is the complex amplitude of the intensity of radiation R, A d is the complex amplitude of the original detection signal, and j is the imaginary unit of the complex number. Typically, the characteristic response time τ can be between 7ms (milliseconds) and 15ms, and the value of the gain G depends in particular on the geometric and thermal characteristics of the sensitive element. The detection signal generated by the sensitive element depends on the intensity of the radiation R that has been received at each moment until the moment of reading the detection signal. This behavior is called the memory effect, and reduces the sensitivity of the sensitive element to rapid changes in the intensity of the radiation R. This reduction in sensitivity is due to the weighted combination effect of the instantaneous values ​​of the intensity of the radiation previously received by the sensitive element. This achieves the time smoothing of these values ​​in the detection signal generated by the sensitive element. The characteristic response time τ defines a time scale according to which the contribution to the detection signal read from the radiation received at the time before the time of reading the detection signal occurs. This contribution to the value of the detection signal read is affected by a multiplicative attenuation coefficient of the exp(-t / τ) type, which is applicable to the intensity of the radiation received before the readout time, where t is the duration between the time when the radiation is received by the sensitive element and the readout time.

[0037] To form the image sensor 10, a matrix 10a of sensitive elements 1 is associated with a controller 10b, denoted CTRL. In a known manner, the controller 10b has a power supply and reading function for each of the sensitive elements 1, an addressing function for each sensitive element 1 within the matrix 10a, and possible additional functions for the sensitive elements 1, such as a test function, an optional thermalization function for at least some of the sensitive elements, and a digitization of the detection signals. In each readout cycle of all the sensitive elements 1 of the matrix 10a, the controller 10b outputs the value of the detection signal that has been read from all the sensitive elements 1, wherein for each captured image, one detection signal value is read per sensitive element. The readout of these detection signals, referred to in the overall part of this specification as raw detection signals, constitutes an image capture operation, and the image thus directly obtained is referred to as a raw image. The readout of the matrix of sensitive elements can be performed according to one of the following two modes: rolling shutter or snapshot mode. In the first case, the rows of the matrix are read sequentially row by row, and in the second case, all rows are read simultaneously. This difference in the reading mode of the image sensor does not affect the principle of the invention in any way, nor the results. The following detailed description is provided as a non-limiting example of a snapshot read mode. To transpose this to the case of reading an image using a rolling shutter mode, the readout times mentioned subsequently may be considered, for example, as the readout times of the first row of the matrix.

[0038] although[ Figure 1 ], optics can be used in front of the matrix 10a in order to optically conjugate the scene to be imaged with the matrix 10a of sensitive elements 1. The radiation R then passes through this conjugated lens and is then incident on the sensitive element 1.

[0039] For the present invention, the image sensor 10 further comprises an image processing module 11, which is connected so as to receive at input the raw image output by the controller 10b. The module 11 is designed to generate a processed image from the raw image so as to compensate or at least partially correct the memory effect of the sensitive element 1 already described above. For this reason, the processed image generated according to the present invention by the image sensor comprising this module 11 is called a corrected image. The image processing module 11 can be a dedicated electronic circuit, or it can be a software module hosted in a processor denoted as a CPU.

[0040] According to the invention, the image processing module 11 generates a corrected image from the original image by subtracting from it another original image captured previously, multiplied by the determined coefficient. Multiplying the original image by the coefficient should be understood to mean the operation of multiplying this coefficient by all the original detection signals constituting the original image. Furthermore, subtracting the first image from the second image should be understood to mean the operation of calculating, independently for each sensitive element 1, the difference between the detection signal read for the second image and the detection signal read for the first image. Therefore, according to the invention, denoted as S corr The corrected image of (t) can be obtained by calculating the value And group the values ​​to obtain, where t is represented by S raw (t) and the original detection signal S i,j_raw (t) is the readout time of the new original image, t-Δ is represented by S raw (t-Δ) and the original detection signal S i,j_raw (t-Δ) is the readout time of the previous original image, Δ is the readout time of the two original images S raw (t) and S raw The duration between the readout times (t-Δ), and S i,j_raw (t) is the raw detection signal read at time t for sensitive elements i, j in the matrix 10a, i and j are the row and column numbers of the sensitive element under consideration, respectively. α is the raw detection signal applied to the previous raw image S raw (t-Δ), and β is a global multiplication coefficient. α and β are positive and non-zero. For an image point associated with a sensitive element located at the intersection of the i-th row and the j-th column of the matrix 10a, S i,j_corr (t) is thus the corrected image S corr (t) is the image point intensity value.

[0041] Advantageously, the coefficients α and possibly β can be selected as a function of the readout raw signal S alone. i,j_raw (t-Δ) and S i,j_raw (t) consists of two original images S raw (t-Δ) and S raw (t) The duration between readout times varies.

[0042] Possibly, but optionally, at least one of the coefficients α and β may have different values ​​for different sensitive elements 1 in the matrix 10a. In this case, the values ​​of the coefficients α and β may be determined individually for each sensitive element 1 during a calibration or benchmark test step that may be performed before each image capture sequence, or in the laboratory.

[0043] In a preferred embodiment of the present invention, the coefficient α can be determined according to the following equation: where Δ again specifies the raw (t-Δ) and S raw (t) and τ again specifies the characteristic response time of the sensitive element. Optionally, the value for the characteristic response time τ may vary according to the sensitive element 1 within the matrix 10a. In this case, the value of the characteristic response time τ of all sensitive elements 1 may have been determined individually for each sensitive element during a calibration or benchmark test step carried out before each image capture sequence or in the laboratory. It is then stored in a memory within the image processing unit 11 or accessible to the unit 11.

[0044] In a more preferred embodiment of the present invention, the coefficient β can be determined according to the following equation: where a is a non-zero constant that can set the ratio of the image point intensity values ​​of the corrected image. For these embodiments, the corrected image S of the sensitive element i, j corr The image point intensity value of (t) is thus The use of this value of the coefficient β makes it possible to reduce the effect of attenuation of the intensity of the corrected image when the duration Δ between the respective readout times of the two original images is short.

[0045] The above listed expressions for the coefficients α and possibly β make it possible to correct memory effects in a particularly effective manner when the sensitive element is of the type with a first-order transfer function with respect to time (as described above). In this case, in fact, the original image S corresponding to the previous original image referred to in the overall part of this description is raw (t-Δ) quantifies the memory effect contribution associated with all radiation received by each sensitive element prior to the capture of this previous raw image when it is multiplied by α=exp(-Δ / τ). This memory effect contribution (which may be referred to as the long-term memory effect and which contributes to the new raw image S raw (t)) Therefore, the present invention is used in the corrected image S corr (t). However, another memory effect contribution associated with the radiation received by each sensitive element during the duration Δ between the respective readout times of the two original images remains. This may be referred to as the short-term memory effect.

[0046] When the sensitive element is not of the type with a first-order transfer function with respect to time, the expression of the coefficient α as a function of the characteristic response time τ and the duration Δ can still be used. For this purpose, an empirical value suitable for the sensitive element under consideration can be adopted for the characteristic response time τ, even if this value does not have a theoretical meaning in relation to the transfer function f(s) of the sensitive element.

[0047] [ Figure 2] shows a video sequence capture, during which raw images are periodically captured at readout times separated by a duration Δ. Thus, the final raw image S raw (t) is the original detection signal S read at time t i j_raw (t) is composed of the previous original image S raw (t-Δ) is the original detection signal S read at time t-Δ i,j_raw (t-Δ), the original image S before this raw (t-2Δ) is the original detection signal S read at time t-2Δ i,j_raw (t-2Δ), the other original image S before this raw (t-3Δ) is the original detection signal S read at time t-3Δ i,j_raw (t-3Δ), etc. Thus, a corrected video sequence consisting of images corrected according to the invention can be constructed at the same frame rate equal to 1 / Δ by combining each original image with the last original image read just before it. raw (t) and S raw (t-Δ) is combined to obtain the image point intensity value S according to the present invention. i,j_corr (t) The corrected image S composed corr (t); original image S raw (t-Δ) and S raw (t-2Δ) is combined to obtain the image point intensity value S according to the present invention. i,j_corr The corrected image S composed of (t-Δ) corr (t-Δ); original image S raw (t-2Δ) and S raw (t-3Δ) is combined to obtain the image point intensity value S according to the present invention. i,j_corr The corrected image S composed of (t-2Δ) corr (t-2Δ), etc. Figure 2 ] corresponds to the time coordinate, denoted t. The image sensor of the invention may thus comprise a display system which is controlled so as to display the corrected image and, possibly, also the original image according to the time correspondence. The improvement in image contrast and tail end attenuation resulting from the correction of the long-term memory effect can thus be evaluated. This pattern of images in the correction video sequence is particularly suitable for obtaining a corrected image sequence at the maximum image capture rate possible for the matrix 10a and the controller 10b.

[0048] [ Figure 3 ] corresponds to [ Figure 2], each corrected image is obtained by combining two consecutive original images with original image pairs dedicated to different corrected images that are non-intersecting, non-interlaced and consecutive. i,j_raw (t) and S i,j_raw (t-Δ1)) composed of the original image S raw (t) and S raw (t-Δ1) is combined to obtain the corrected image S according to the present invention corr (t). Similarly, the original detection signal S i,j_raw (t-(Δ1+Δ2)) and S i,j_raw The original image S composed of (t-(2Δ1+Δ2)) raw (t-(Δ1+Δ2)) and S raw (t-(2Δ1+Δ2)) is combined to obtain the image point intensity value S according to the present invention. i,j_corr The corrected image S composed of (t-(Δ1+Δ2)) corr (t-(Δ1+Δ2)), and so on. Δ1 is the duration between the readout times of two original images of the same pair used to obtain the corresponding corrected image, and Δ2 is the duration between the readout times of the previous original image of the last pair and the new original image of the previous pair. The video rate of the corrected images is thus 1 / (Δ1+Δ2). This further mode of correcting images in a video sequence is suitable for obtaining video of the corrected images at a low frame rate while being able to use the short duration between the two original images that are combined to obtain each corrected image.

[0049] [ Figure 4 ] also corresponds to [ Figure 2 ], again consider a video sequence of raw images captured at frequency 1 / Δ. But this time, a series of corrected images is generated by combining the same last captured raw image with a different raw image each time in chronological order and progressively using previous raw images upstream in the video sequence. Thus, the image point intensity value S i,j_corr_1 The first corrected image S composed of corr_1 By using the coefficient and The combination is respectively from the original detection signal S i,j_raw (t) and S i,j_raw (t-Δ) The original image S constructed raw (t) and S raw (t-Δ) is obtained. i,j_corr_2 The second corrected image S composed of corr_2 By reusing the original image S raw (t) and the use factor and It is combined with the original detection signal Si,j_raw (t-2Δ) composed of image S raw (t-2Δ) combination is obtained. i,j_corr_3 The third corrected image S composed of corr_3 By using the coefficient and Similarly, the image point intensity value S i,j_corr_4 The fourth corrected image S composed of corr_4 By using the coefficient and The original image S raw (t) and the original detection signal S i,j_raw (t-4Δ) constructed image S raw (t-4Δ), etc. The set of corrected images thus obtained makes it possible to evaluate the influence of the duration between the original images combined according to the invention on the quality of the corrected images. Some of these advantageous effects are presented here: better rendering of high spatial frequencies in the image in the presence of lateral movement, and reduction of the tail effect.

[0050] Rendering of image spatial frequencies in the presence of lateral motion .

[0051] As is known, one way to highlight the decay of the temporal transfer function of a matrix of sensitive elements is to capture the image of a spatial pattern that is periodic and moves at a constant speed parallel to its periodic direction. Each sensitive element thus receives radiation whose instantaneous intensity varies periodically according to a temporal frequency value that is equal to the product of the apparent travel speed and the period of the pattern. A scene consisting of bands parallel to the direction of the columns of the matrix 10a and whose brightness varies sinusoidally parallel to the direction of the rows of the matrix 10a is imaged onto the matrix 10a of sensitive elements 1. The spatial frequency of these bands in the scene image is denoted by v s , which can be pixel -1 This scene moves at a constant speed parallel to the rows of the matrix 10a, and V specifies the speed of movement of the image of the scene on the matrix 10a, which can be expressed in pixels / s (pixels / second). It is assumed that all sensitive elements 1 have the same value of the characteristic response time τ and their common transfer function is [ Figure 5 ] is a graph showing the apparent transfer function of the image sensor as a function of the spatial image frequency ν s Or a diagram of the variation of the moving speed V. The horizontal axis of the diagram indicates the spatial frequency ν s The value of is multiplied by the characteristic response time τ and by the velocity V in order to obtain a dimensionless variable. The product V·ν s corresponds to the temporal frequency of the variation of the intensity of the radiation received by each sensitive element. The vertical axis indicates the value F of the amplitude of the apparent transfer function of the image sensor. Figure 5 The lower curve appearing in FIG. 1 and designated by the words “no correction” corresponds to the original detection signal S delivered by the controller 10 b before the memory effect correction. i,j_raw , as the velocity V or spatial frequency ν s . It is therefore a readout signal constituting each raw image captured during the scene movement. The curve denoted Δ=τ corresponds to the image point intensity values ​​of the corrected image obtained by combining two raw images captured during the scene movement at readout times separated by a duration Δ equal to τ. Similarly, the curve denoted Δ=τ / 4 corresponds to the image point intensity values ​​of the corrected image obtained by combining two raw images captured during the scene movement at readout times separated by a duration Δ equal to τ / 4. These corrected images use the values ​​of the coefficient β calculated for each of the above indicated in this specification. The general increase in the amplitude F of the transfer function of the image sensor when going from the raw image to the corrected image with Δ=τ and then to the corrected image with Δ=τ / 4 corresponds to an increased correction of the memory effect of the sensitive element 1. In other words, the uncompensated short-term memory effect in the corrected image is less and less. When the dimensionless spatial frequency is greater than 1.0, the gain of the amplitude F of the transfer function of the image sensor compared to the original image is about 2.1 for the image corrected with Δ=τ, corresponding to a specific spatial frequency (v s ·τ·V). For these specific spatial frequencies, called blind frequencies, the two original images combined to obtain the corrected image are identical, so that the memory effect correction has no influence, and for the image corrected with Δ=τ / 4, the gain in the amplitude F of the transfer function of the image sensor is about 8 again compared to the original image but when the dimensionless spatial frequency is greater than 2.0, and the blind frequency corresponds to v s ·A value of τ·V that is a multiple of four.

[0052] Tail Effect .

[0053] The matrix 10a used comprises 320 columns and 240 rows of sensitive elements 1, and the characteristic response time τ of all sensitive elements is approximately 14 ms. The image sensor formed by this matrix of sensitive elements captures a video sequence with a uniform background consisting of a black body at 325k (Kelvin) via the upper part of an opaque rotating disk with three radial slit openings. The matrix 10a is optically conjugated to the rotating disk, and the rotation axis of the disk is parallel to the optical axis of the conjugated optics used. The rotation speed of the disk is 1.5 revolutions / second, and the raw image acquisition rate is 60 images / second, corresponding to the duration between the corresponding readout times of any two consecutive raw images (which is equal to 16.7 ms). [ Figure 6 ] are represented by S raw (t-Δ) and Sraw (t) is the value of the original detection signal read from the rows of the matrix 10a of two original images captured successively in the video of (t). It also reproduces the value of the original detection signal read from the rows of the matrix 10a of the two original images S according to the present invention. raw (t-Δ) and S raw (t) The corrected image S is obtained corr (t) Image point intensity values ​​of the same row of matrix 10a. The movement of one of the radial slits of the rotating disk is visible between the two original images: from the right side to the left side of the figure. [ Figure 6 ] the horizontal axis of the diagram is based on its column number n c The sensitive element 1 of the row under consideration in the matrix 10a is identified, and the vertical axis identifies the three images S raw (t), S raw (t-Δ) and S corr (t) The value of the original detection signal I or the image point intensity value. Corresponding to the original image S raw (t) and S raw The progressively sloping rising and falling edges of the intensity curves of the two edges of the gap in (t-Δ) constitute the tail effect due to the combination of the movement of the gap and the memory effect of the sensitive element 1. corr In (t), the two edges of the gap appear sharper, demonstrating the efficiency of memory effect compensation. For this example, the formula provided above in this specification for coefficients α and β implements S i,j_corr (t) = 1.437 x S i,j_raw (t)-0.437x S i,j_raw (t-Δ). In general, the adjustment of the ratio between the duration Δ separating consecutive original images of a video sequence and the characteristic response time τ of the sensitive element may depend on the trade-off between the reduction of tail effects, the improvement of the geometric rendering of scene elements in the image and the amplification of image noise in the corrected image. For example, a value of the ratio Δ / τ between 0.02 and 0.2 may be suitable when the signal-to-noise ratio of the original image is greater than 150 (respectively 15).

[0054] It will be appreciated that the invention can be reproduced by modifying the minor aspects of the mode of embodiment that has been described in detail above while retaining at least some of the advantages enumerated. Specifically, the choice of the original images that are combined in pairs to obtain the corrected image can be modified with respect to the example illustrated. Furthermore, it will be remembered that the coefficient β that depends on the duration Δ between the original images that are combined to form the corrected image is optional, even if the coefficient α used is calculated according to this duration Δ.

Claims

1. A method for capturing images, wherein several images are captured successively using the same matrix (10a) of sensitive elements (1), each sensitive element exhibiting a memory effect which causes the raw detection signal of the sensitive element to depend on the amount of radiation received by the sensitive element at a readout time at which the raw detection signal is read, and also on the amount of radiation received by the sensitive element before the readout time, It is characterized in that In order to form an image at least partially corrected for said memory effects, called corrected image, an image point intensity value is individually assigned to each of said sensitive elements (1) of said matrix (10a), said image point intensity value being proportional to a difference between a raw detection signal of a sensitive element read for a newly captured image, called new raw image, and a portion of a raw detection signal of the same sensitive element read for another image, called previous raw image, captured before said new raw image, The image point intensity value assigned to each of the sensitive elements (1) to form the corrected image is proportional to the result of dividing the difference between the raw detection signals of the sensitive elements read for the new raw image by [1-exp(-Δ / τ)] and the result of multiplying the raw detection signal of the same sensitive element read for the previous raw image by exp(-Δ / τ), where exp(.) specifies an exponential function, τ is the characteristic response time of the sensitive element, and Δ is the non-zero duration between the readout times of the sensitive elements for the new raw image and the previous raw image, respectively.

2. The method according to claim 1, characterized in that The duration Δ between the corresponding readout times of the same sensitive element (1) for the new raw image and the previous raw image, respectively, is smaller than the characteristic response time τ of the sensitive element.

3. The method according to any one of the preceding claims, characterized in that The previous original image is the last one of several images that have been captured before the new original image according to a time sequence of capturing of the original images.

4. The method according to any one of the preceding claims, characterized in that Each sensitive element (1) is a bolometer or microbolometer, a thermopile, a pyroelectric sensor, a ferroelectric sensor, or a thermal deformation micro-sensor.

5. The method according to any one of claims 1 to 4, characterized in that Each corrected image is formed from a pair of original images, including a new original image and a previous original image that has been captured before the new original image, and wherein the pairs of original images used to form consecutive corrected images are disjoint, non-interleaved, and temporally consecutive.

6. The method according to any one of claims 1 to 4, characterized in that Several corrected images are formed by combining the same new original image with several previous original images captured successively before the new original image, and each corrected image is obtained by using the difference between the readout times of the sensitive elements for the new original image and the previous original images combined to form the corrected image, respectively, within the duration Δ associated with each sensitive element (1).

7. An image sensor (10), comprising: a matrix (10a) of sensitive elements (1), each sensitive element exhibiting a memory effect which causes a raw detection signal read from said sensitive element to depend on the amount of radiation received by said sensitive element at a readout time at which said raw detection signal is read, and also on the amount of radiation received by said sensitive element before said readout time, and - an image processing module (11) adapted to output images at least partially corrected for said memory effect, called corrected images, each corrected image being formed by the image point intensity values ​​of said sensitive elements (1) respectively assigned to said matrix (10a), said image processing module being adapted to calculate said image point intensity value of any one of said sensitive elements as proportional to a part of the difference between a raw detection signal of said sensitive element read for a newly captured image, called new raw image, and a part of said raw detection signal of the same sensitive element read for another image captured before said new raw image, called previous raw image, The image point intensity value assigned to each of the sensitive elements (1) to form the corrected image is proportional to the result of dividing the difference between the raw detection signals of the sensitive elements read for the new raw image by [1-exp(-Δ / τ)] and the result of multiplying the raw detection signal of the same sensitive element read for the previous raw image by exp(-Δ / τ), where exp(.) represents an exponential function, τ is the characteristic response time of the sensitive element, and Δ is the non-zero duration between the readout times of the sensitive elements for the new raw image and the previous raw image, respectively.

8. The image sensor (10) according to claim 7, which is suitable for implementing the method according to any one of claims 1 to 6.

Images