FR3008230A1 - MATRIX COLOR IMAGE SENSOR, WITHOUT COLOR FILTERS - Google Patents

Abstract

The invention proposes a matrix color image sensor, without color filters, in which: the matrix is arranged in identical groups of X neighboring photosensitive pixels P1, P2, P3, P4, not all obtained by a rigorously identical manufacturing process , so as to obtain in a controlled manner different spectral responses in the sensitivity spectrum of the sensor; the sensitivity spectrum of the sensor is decomposed into M spectral bands juxtaposed; the response of each of the pixels of a group for each of the M spectral bands is measured by illuminating the active surface of the sensor with a known luminance level in this spectral band. For each pixel, a spectral band coefficient is obtained, which represents the response of the pixel subjected to illumination in this band. The M coefficients of each of the X pixels of the group thus obtained make it possible to solve a system of X equations with unknown M from the signal values provided by each of the X pixels as a function of the illumination received by the group of pixels, where the M unknowns are M calculated luminance values Lj representing the decomposition of this illumination in each of the M spectral bands, to obtain by calculation, a color information from the X pixels of each group.

Description

TECHNICAL FIELD The invention relates to electronic matrix image sensors for taking color images. It applies more specifically to matrix sensors with active pixels.

STATE OF THE ART Recall that a basic diagram of a matrix sensor active pixel comprises, for example, a photodiode, a charge storage capacity and a few transistors providing various pixel control functions: charge transfer, reset, charge conversion / voltage and read selection. According to the state of the art, the color image is generally obtained by means of a mosaic of color filters usually arranged on the photosensitive surface of the sensor: a filter is placed in front of each pixel; the pixel receives only the color that is not absorbed by the filter. A mosaic of known color filters is the so-called Bayer mosaic, arranged in a pattern of checkered tiles (a tile = a filter), comprising 50% of green filters, 25% of red filters, and 25% of blue filters. This drawing is specially adapted to the perception characteristics of the human eye. It presents a periodic alternation of colored filters of two colors on each line, green and blue on one line, red and green on the following line; and the same periodic alternation on the columns. The design of such a mosaic of color filters MB is schematically shown in Figure 1, where the tiles respectively marked R, V, B respectively represent the red, green and blue color filters. Bayer mosaics with colored filters corresponding to secondary colors, yellow, magenta and cyan, can also be used, with a similar checkerboard design, replacing green with yellow, red with magenta and blue with cyan .

In the context of the invention, it should be noted that the color qualifying the color filter namely "red filter", "green filter" or "blue filter" means that the color filter in question is optimized for a range of lengths. wave corresponding to this color, to let pass only this range of wavelengths and absorb as much as possible all the others. In these color filter sensors, each pixel provides an information signal from the observed scene in a color that the associated color filter passes through. The complete color information is obtained by considering the neighboring pixels that give the information in the complementary colors. According to the state of the art, the mosaic of color filters is formed by depositing colored organic resin layers on the surface of the sensor, the photosensitive elements of which are formed from silicon. DESCRIPTION OF THE TECHNICAL PROBLEM The use of organic colored filters poses different technical problems. When the sensors are intended to be used in harsh environments, especially in space applications, under conditions of high vacuum and exposure to irradiation, the organic resin reacts poorly. This can result in deposits of impurities on the internal face of the sensor glass, due to degassing of the resin, or pollution of the sensor. The optical qualities of the sensor can be strongly degraded. In theory, it would be possible to produce non-organic filters corresponding to the desired colors, but this technological alternative is very expensive. Another disadvantage of colored filters, organic or inorganic, is that they contribute about 30% to the light transmission losses in the sensor, which is not negligible. For color filters made by organic resins, these transmission losses are also variable with the wavelength. Finally, some applications use sensors designed to allow the capture of color and infrared images, night or day. This is typically the case of surveillance cameras or industrial control cameras. However, the organic resin colored filters have a quasi-absorbent power for the infrared. The electrical signal produced by each of the pixels of the sensor thus contains not only a color component corresponding to the bandwidth of the associated organic color filter, but also a near-infrared component, which must be eliminated when it is desired to take a color image. The sensors that must allow the capture of color and infrared images must then be equipped with a near infrared filtering optical element mounted removably to be removed when you want to take infrared images. SUMMARY OF THE INVENTION In the invention, it is sought to solve these various technical problems in an inexpensive manner, forming part of a standard semiconductor manufacturing process that guarantees a low production cost. The technical solution must allow good optical performance, whether in terms of colorimetric quality, spatial resolution or color separation. An idea underlying the invention is to provide a matrix sensor without absorbing optical layers producing color filters, more particularly without colored organic filters. The technical solution proposes an arrangement of the pixel matrix in identical groups of X neighboring pixels (for example X = 4), the X pixels being all differentiated in a controlled manner, in the process of manufacturing the pixel, so that they not all have the same spectral response in the sensitivity spectrum of the sensor; more precisely, the sensitivity spectrum of the sensor is subdivided into M adjacent spectral bands, with a characterization of the X pixel types of the sensor by measuring the signal value supplied by each of the X pixels of a group in each of the M spectral bands. for a given illumination; this characterization provides M response coefficients for each of the X pixels. In use, for any illumination, it is then known to calculate the luminance received by each pixel in each of the M spectral bands from a system of X equations with unknown M, using X signals provided by the X pixels of a group and M response coefficients of each of the X pixels.

As characterized, the invention thus relates to a matrix image sensor comprising np photosensitive pixels arranged according to n lines and p columns, n and p integers greater than or equal to two, with a pixel at the intersection of a line and d a respective column. According to the invention, the pixels are organized into identical groups of X adjacent pixels, X greater than or equal to two, the X pixels not having the same spectral response. The spectral response of a pixel of rank i in the group is defined by M known coefficients where j represents an index varying from 1 to M of a respective spectral band among M juxtaposed spectral bands forming the whole of the spectral band to which the sensor is sensitive, with M being less than or equal to X. The sensor comprises a calculation means for solving for each group of pixels a system of X equations with M unknown of the form: S. = j = 1 LI - x C with integer i ranging from 1 to X, j integer varying from 1 to M where the input data are signal values Si supplied by each of the X adjacent pixels receiving a common illumination, and where the unknown M are M values. Calculated luminance Li representing the luminance received by the group of pixels in each of the M spectral bands. In a first embodiment, M is greater than or equal to three and the M spectral bands comprise at least one spectral band generally corresponding to a blue color, a spectral band generally corresponding to a green color and a spectral band generally corresponding to one color. red. In a second embodiment, M is greater than or equal to four and the M spectral bands comprise at least one spectral band generally corresponding to a blue color, a spectral band generally corresponding to a green color and a spectral band generally corresponding to one color. red, and a spectral band corresponding to the near infrared. Since no color filters are used above the pixels, for each of the X pixels, there will generally be several nonzero responses in the different spectral bands because the silicon is naturally sensitive to illumination in the range. wavelengths ranging from about 400 nanometers to 1100 nanometers, including blue, green, red and near infrared. Thus, in practice, for a decomposition into four spectral bands, there will be among the M coefficients Cu at least three non-zero values for almost all the X pixels. The invention is particularly applicable with X = 4 pixels in each group, but one could also consider X = 9 pixels. X is preferably equal to M but it is not mandatory; it is generally desirable to have a number of pixels corresponding to a generally square area. For example, if M = 3, we could use only X = 3 pixels to solve a system of three equations with three unknowns but we prefer to take four pixels constituting a square-shaped group, two diagonal pixels can be identical and providing two redundant information.

To make pixels with different spectral responses, organic layers deposited above the pixels will not be used after fabrication of the photosensitive sensor on a silicon wafer, but only processing steps of the mineral layers of the silicon wafer will be used. In other words, if one refers to the prior art, in the prior art manufacturing processes, the steps of treatment of the silicon wafer and the so-called "postprocess" stages consisting in deposition and etching were distinguished. colored filters or possibly other organic layers. According to the present invention, the realization of different spectral response pixels in groups of X pixels will result only from silicon processing steps excluding any organic layer processing step. Thus, one or more of the following technologies will be used: formation of a thicker P + type surface layer on a photodiode of at least one pixel of the group of X pixels and less thick on the other pixels of the group; depositing a polycrystalline silicon layer above the photodiode of at least one pixel of the group but not above the other pixels of the group, depositing an anti-reflection mineral layer over at least one a pixel of the group, different from an anti-reflection mineral layer above the other pixels of the group; formation of a P + type buried layer below the entire N-layer of a photodiode of at least one pixel of the group, but not below the entire N-layer of the photodiodes other pixels of the group, - formation of a P + buried layer below only a portion of the N-layer of a photodiode of at least one pixel of the group, but not below the photodiodes of the other pixels in the group. Each of these actions contributes to the modification of the spectral response of the pixels, and a group of four pixels will use three or four of these possibilities (or combinations of these possibilities) to obtain three or four different spectral responses. Other features and advantages of the invention are presented in the following description, with reference to the accompanying drawings in which: FIG. 1 already described represents a mosaic of Bayer colored filters; FIG. 2 represents a matrix of pixels that can be arranged in identical groups of X adjacent pixels which are differentiated by process, in spectral response, according to the invention; FIGS. 3 and 4 respectively illustrate an example of a standard process used to make the photodiode of a pixel, and the spectral response of the pixel in the spectral band to which the sensor is sensitive; FIGS. 5 and 6 respectively illustrate a first example of a process modified with respect to the standard process for producing the photodiode of a pixel, and the spectral response of this pixel in the spectral band to which the sensor is sensitive; - Figures 7 and 8 respectively illustrate a second example of modified process, and the spectral response obtained; - Figures 9 and 10 show a modification using a polycrystalline silicon layer; FIGS. 11 and 12 respectively illustrate a modification using a buried layer, and the spectral response obtained; - Figures 13 and 14 illustrate a variant of the embodiment of Figures 11 and 12; FIGS. 15 and 16 illustrate an embodiment with a locally modified antireflection layer, with the corresponding spectral response; FIG. 17 represents a structure of four pixels according to the invention, of spectral responses that are all different from each other.

DETAILED DESCRIPTION The invention proceeds from the realistic assumption that adjacent photosensitive pixels, that is to say adjacent pixels on the same line or the same column, are subjected to the same illumination. It proposes: an arrangement of the matrix in identical groups of X neighboring photosensitive pixels; and in this arrangement the X pixels are not all obtained by a rigorously identical manufacturing method, so as to obtain in a controlled manner different spectral responses in the sensitivity spectrum of the sensor. Each group includes X "types" of pixel. Pixel fabrication technology is a silicon wafer technology and manufacturing differences are in the manufacture of the silicon wafer itself and not on the organic layers that would be reported on this wafer. Indeed, in this invention, there are no organic color filters to ensure the transmission of a single color to a given pixel. Each group of X pixels, however, makes it possible to obtain color information from the X pixels of this group, by calculation, according to the following principle: the sensitivity spectrum of the sensor is decomposed into M spectral bands juxtaposed; and for each of the M spectral bands, the response of each of the X pixels P1 of a group, i index varying from 1 to X, is measured by illuminating the active surface of the sensor, with an illumination in this spectral band and with a known luminance level in this band. For each pixel Pi, we obtain M noted characteristic coefficients where j varies from 1 to M, the coefficient representing the response of the pixel Pi subjected to an illumination of the pixel in the jth spectral band, that is to say the ratio between the signal level from the pixel and the luminance received by the pixel. If the sensor provides an output signal that varies proportionally to the integration time (which is often the case), it is assumed of course that the determination is made for a given integration time, and the time of the integration is taken into account later. integration if it is not the same as when determining the coefficient. At the scale of the sensor matrix, it is sufficient to perform these measurements on a single group of differentiated X pixels to characterize the sensor. Characteristic coefficients X are obtained which are stored in storage means provided in the sensor. It is then possible to implement a determination of the response of each of the X pixels of the sensor subjected to any illumination coming from a scene whose image is to be captured, by solving a system with X equations and M unknowns, from the signal Si measured on each of the X pixels, and the characteristic coefficients of the X pixels. The system of equations is written: S. j = 1 I L- x C - with i integer varying from 1 to X, j integer varying from 1 to M ed where the input data are the signal values If provided by each of the X pixels as a function of the illumination received by the group of pixels, and where the unknown M are M calculated luminance values Li representing the decomposition of this illumination in each of the M spectral bands. The signal Si supplied by a pixel is indeed the sum of the responses x Ci, i of this pixel in each of the spectral bands: Si = Sum for j varying from 1 to M of [Li x Ci, i]. The system of equations makes it possible to determine these M values Li, provided that there are indeed M independent equations, that is to say that there are real differences in the spectral response of the different pixels. The number X of pixels is greater than or equal to M. Indeed, to determine M unknowns it takes at least M equations.

The sensor comprises the calculation means for solving X unknown linear equations X. The calculation means can be put on the integrated circuit comprising the pixel matrix or on a separate circuit (in particular an FPGA logic gate array) receiving the analog values or preferably digital values provided by the pixel matrix. In the example described below it will be considered that X = 4, that is to say that each group has four pixels arranged in a square, and it will be considered that M = 4, corresponding to a conventional spectrum decomposition in the sensors. colors, the spectral bands corresponding globally to respective color ranges blue, green, red, and near infrared (NIR for "near infra red"). One could also take M = 3 (blue, green, red) while keeping four pixels to keep a square shape of the group of pixels; we would then use two identical pixels, for example diagonally in the square, and third and fourth pixels different from each other and different from the first two. To explain the practical implementation of the invention, we will successively detail the arrangement in the matrix of different types of pixels and groups, and practical means for differentiating pixels by the silicon processing steps to obtain data. different spectral responses; the method of calculation will be detailed in an example with four types of pixels of differentiated spectral responses.

FIG. 2 illustrates an array of pixels, comprising groups of X neighboring pixels, and differentiated in spectral response, according to the invention. In another way, the matrix comprises X types of pixels of the same number distributed in identical groups, where each group comprises the X types. In the example, X is equal to four and the four differentiated pixels are denoted P1, P2, P3 and P4. These pixels are preferably arranged on two lines and two columns, so as to form a square. The figure shows, for example in hatched areas, a first group G1 formed of 4 differentiated pixels P1, P2, P3 and P4, in the center of the pattern of the matrix, and a second group G2, formed identically with 4 differentiated pixels P1, P2, P3 and P4, located in the upper left corner in the figure. These two groups G1 and G2 have no pixel in common. The matrix can thus be broken down into groups of differentiated pixels which share no pixels. In this case, the color information is provided by the 4 pixels: the resolution of the sensor corresponds to the size of the group. But it is also possible to define groups of pixels that share pixels. For example, with X = 4, each pixel can belong to 10 different groups each having four different types. If we take the example of the pixel P1 of the central group G1 represented by an oblique striped pattern in FIG. 2, this pixel can belong to 3 other groups G1 ', G1 ", G1"' represented by dotted circles in the figure . It is thus possible to form 4 times more groups of differentiated pixels. As will be seen later, the color information on each group of X pixels is obtained by calculation, the resolution is improved by doing this calculation for all groups of X pixels that can be formed. In another example where X would be 9, with an arrangement of differentiated pixels in the 3-row and 3-column square array, each pixel can belong to 9 different groups. In practice, the number X of differentiated pixel types in a sensor is limited by various considerations: the loss of resolution, which can be compensated by considering overlapping groups of pixels, but at the expense of an increase in volume Calculation ; - the illumination must be considered uniform at the level of each group; the higher the number of pixels per group, the less this is true: the output information for each group is more inaccurate. the technological possibilities of differentiating the spectral responses of the different pixels of the group, as will be detailed later. Finally, for reasons of simplicity of implementation, the square arrangement is preferred, for example in square of 4 or 9 pixels.

But the invention is not limited to such an arrangement. In particular, it is possible to provide (but less advantageously) rectangular arrangements: on one line and 3 or 4 columns, on two lines and three columns; arrangements may also be provided with hexagonal shaped pixels, etc. Figures 3, 5, 7, 9, 11 and 13 schematically show examples of controlled differentiation of pixels for a CMOS technology sensor made on a silicon substrate and using a photodiode as a photosensitive element in each pixel; and Figures 4, 6, 8, 10, 12 and 14 illustrate exemplary spectral responses obtainable for the sensor sensitivity band, namely the visible and near infra-red spectrum. In the figures, we have taken the example of pixels obtained from a P + doped silicon substrate, which corresponds to the most common technology; the invention is symmetrically transposed to pixels obtained from an N-doped silicon substrate. FIG. 3 shows the layer structure of a pixel obtained by standard steps, at the level of the photodiode element (junction PN) 20 when the photodiode is a photodiode of "pinned photodiode" type, that is to say a photodiode with surface potential fixed by a P + type surface layer. This structure comprises, in a well-known manner, a P + doped silicon substrate 10 of ebulk thickness, covered with a P-type doped epitaxial layer 12 and with a thickness eepi. For the convenience of the representation, the thicknesses are not represented in a realistic way in the figures: the ebulk thickness is several hundred micrometers, the thickness eepi is a few micrometers. An N-type layer 14 is diffused from the surface of the epitaxial layer 12. Its thickness eN is of the order of a few hundred nanometers. The layers 12 and 14 form the PN junction constituting the photodiode. Finally, a P + doped surface layer 16 conventionally covers the N layer. This surface layer has a standard ep thickness which may be a few tens of nanometers. This layer is used to maintain the surface potential of the photodiode at zero. FIG. 4 represents a typical spectral response diagram of such a photodiode. In the abscissa, the wavelengths in the visible and near infrared spectrum have been increased from about 400 nanometers to 1000 nanometers. On the ordinate the quantum efficiency, that is to say the conversion efficiency of photons into electrons, has been increased for the different wavelengths. These curves, as well as those of the following figures, are given for illustrative purposes; they are not experimental curves; they are given to show that one can act by different technological means on the spectral response of the photodiode. The useful spectrum is decomposed in this example into four spectral bands B, V, R, NIR, corresponding approximately to blue, green, red, and near-infrared color ranges from shorter to longer wavelengths. . The pixel of FIG. 3 can then be characterized by four respective response coefficients C1,1, C12, C1,3 and C1,4 corresponding to these four bands. A pixel P1 groups of four pixels can be formed with this conventional structure. These coefficients are obtained by preliminary measurement (or possibly by simulation and / or calculation) of the response of each pixel type to an illumination each limited to one of the four spectral bands. Fig. 5 shows a photodiode structure modified to reduce the response in blue and in green, resulting in coefficients of a second pixel P2, which would be C2,1, C2,2, C2,3 and C2,4 , in which C2,2 and especially C2,1 are significantly lower than C1,2 and C1,1 respectively. This modification is obtained by increasing the thickness of the P + type surface layer 16 on the photodiode of the second pixel P2. Indeed, the short wavelengths have a low penetration depth in the silicon and then hardly reach the N layer. The electrons created in the layer 16 can hardly reach the N layer below and be stored there. They do not participate in the signal produced by the photodiode. The thickness of the layer 16 may be 100 to 300 nanometers for example.

Figure 6 shows a corresponding spectral response curve where the quantum efficiency is reduced in the spectral bands corresponding to blue and green. The group of coefficients C2,1, C2,2, C2,3 and C2,4 is different from the group C1,1, C1,2, C1,3 and C1,4.

Fig. 7 shows a photodiode structure similar to that of Fig. 5 but in which the thickness of the surface layer 16 has been further increased to virtually eliminate any response in the blue. Figure 8 shows the corresponding spectral response curve. The thickness of the layer 16 may be 300 to 600 nanometers. The pixel P2 could for example use this structure rather than that of FIG. 5. Response curves similar to those of FIGS. 6 and 8 could be obtained by another structural modification of the photodiode. This modification consists in depositing above the pixel P2 a layer of polycrystalline silicon with a thickness of a few hundred nanometers, playing the same role as the layer 16 for absorbing the photons of short wavelengths. The use of polycrystalline silicon etched locally above certain pixels is technologically easy because it is in any case necessary to use polycrystalline silicon for other circuit elements and in particular active pixel transistors. The thickness of polycrystalline silicon used above the photodiodes is then preferably limited to the thickness used for the gates of the transistors, typically about 200 nanometers in CMOS technology, up to 350 nanometers in other technologies. If polycrystalline polysilicon is used, it will in principle not put a P + type surface layer such as the layer 16, this layer is then not essential and then being more difficult to implement. FIGS. 9 and 10 represent such a possibility, without or with the P + type surface layer. In the latter case, the same standard thickness as for FIG. 3 has been kept for the surface layer 16 but that above this layer 16 a localized layer of polycrystalline silicon 18. This layer 18 may cover only a portion of the surface of the photodiode. Its thickness is all the more important that one wants to attenuate the answer in blue and green.

FIG. 11 represents yet another structural modification of the photodiode, making it possible to produce a third pixel P3 whose spectral response is now attenuated in the red, and also attenuated or even reduced to zero in the infrared. This modification 5 consists in locally providing, under the diffusion N of the photodiode of the pixel P4, a buried layer D_P of type P, which is more heavily doped than the epitaxial layer. This layer serves as a potential barrier between the lower part of the epitaxial layer and the PN junction of the photodiode. Thus, the electrons generated in this lower part under the effect of high penetration photons (red and especially infrared) remain localized below the barrier and do not join the N layer. They end up recombining and therefore do not contribute not to the useful signal provided by the photodiode. Figure 12 shows the strongly attenuated spectral response in the red and the infrared, giving a pixel with a group of response coefficients C3,1, C3,2, C3,3, and C3,4 different from the previous groups. Figure 13 shows a variant that could be used (for example for the same pixel P3). In this variant, the P-type buried layer D_P occupies only a part and not the entire extent of the photodiode. The action of this layer is the same as that of FIG. 11, but to a lesser degree, which is seen in the corresponding figure 14, where the quantum efficiency in the R and NIR spectral bands is attenuated relative to FIG. 15 shows a way of acting on the spectral response curve due to the presence of one or more antireflection layers of specific characteristics in FIGS. 4 and 6, but less attenuated than that of FIG. above the photodiode of the pixel concerned. In general, no antireflection layers have been shown above the photodiode of FIGS. 3, 5, 7, 9, 11, 13, but one or more layers are present and the spectral response which has been 30 represented takes account of it. The anti-reflective action of these layers depends on the number of layers, their thickness, the optical index of the materials used. In the simplest case, the anti-reflection layer may for example be a silicon nitride SiN or Si3N4 layer above a silicon oxide layer. In the prior art, the anti-reflection layers are uniformly deposited over the entire pixel array. In the present invention, certain pixels may be constituted with an anti-reflection layer constituted differently above certain pixels to give different reflection properties and thus locally modify the spectral response.

For example, it is possible to form a fourth type of pixel P4 with an anti-reflection layer which acts on certain spectral bands differently from the others. In one example, the constitution, for example the thickness, of an antireflection layer of silicon nitride is chosen so as to increase the reflection in the green (and thus to reduce the spectral response in this band). This modification is represented symbolically in FIG. 15 by an excess thickness of an antireflection layer of silicon nitride 20, this excess thickness being located at the pixel P4 and absent above the neighboring pixels. Fig. 16 shows a corresponding spectral response capability, resulting in a group of coefficients C4,1, C4,2, C4,3, and C4,4 different from the previous groups. Finally, FIG. 17 illustrates as an example a set of four pixels P1, P2, P3, P4 forming a group having different spectral responses. The pixels are shown juxtaposed in FIG. 17 and constitute in reality a square. The photodiodes of the various pixels are conventionally separated by isolation regions creating isolation barriers preventing the mixing of charges generated in the different photodiodes.

Numerous technological possibilities therefore exist for modifying the spectral response of a pixel by an action taken during the manufacture of the photosensitive sensor, that is to say at the level of the treatment of the mineral layers formed on a silicon wafer (including passivation layers and anti-reflection layers). The possibilities which have been described above by way of example have the advantage of being part of the standard CMOS technology sensor technology even if they require additional masking steps. Thus: the surface layer 16 exists in the conventional technology of pinned photodiodes. A mask will be added to form a deeper implantation only for the P2 type pixels in the example given above; polycrystalline silicon layers are available since they are already required to form transistor gates in the CMOS pixel matrix and transistor gates in the peripheral control circuits; deep buried layers such as D_P (FIGS. 11 and 13) are available in the peripheral circuits, to form N-type wells in which certain NMOS transistors are made; Anti-reflective layers are available and it will require a mask to deposit overthicknesses or in practice locally reduce the thickness of a thicker layer. The invention is applicable mainly for spectrum decomposition into four spectral bands but can be applied to more than four or less than four at the limit for simple decomposition into a visible (colorless) image and an infrared image. . The invention is also applicable to thinned sensors illuminated by the back side, but the technological modifications made to modify the spectral response will of course have to take this into account. For example, the buried layer D_P becomes a layer attenuating the response in the short wavelengths if the light comes from behind.

Claims (6)

REVENDICATIONS1. A matrix image sensor comprising np photosensitive pixels arranged in n rows and p columns, n and p integers greater than or equal to two, with a pixel at the intersection of a line and a respective column, characterized in that the pixels are organized into identical groups of X adjacent pixels, X greater than or equal to two, the X pixels not having the same spectral response, the spectral response of a pixel of rank i in the group being defined by M known coefficients where j represents an index varying from 1 to M of a respective spectral band among M juxtaposed spectral bands forming the whole of the spectral band to which the sensor is sensitive, with M integer less than or equal to X, the sensor comprising a means calculation method for solving for each group of pixels a system of X equations with M unknowns of the form: S. j = 1 I L- x C. - with i integer varying from 1 to X, j integer varying from 1 to M ed where the input data is d The signal values Si provided by each of the X adjacent pixels receiving a common illumination, and wherein the unknown M's are M values of calculated luminance Li representing the luminance received by the group of pixels in each of the M spectral bands. 2. Sensor according to claim 1, characterized in that M is greater than or equal to three and the M spectral bands comprise at least one spectral band generally corresponding to a blue color, a spectral band generally corresponding to a green color and a spectral band. generally corresponding to a red color. 3. Sensor according to claim 1, characterized in that M is greater than or equal to four and the M spectral bands comprise at least one spectral band generally corresponding to a blue color, a spectral band generally corresponding to a green color and a spectral band. generally corresponding to a red color, and a spectral band corresponding to the near infrared. 4. Sensor according to one of claims 1 to 3, characterized in that the pixel group forms a square. 5. Sensor according to one of the preceding claims, characterized in that the pixels of a group have different spectral responses due to local operations resulting from processing steps of a silicon wafer and in that the sensor does not have organic color filters. 6. Sensor according to one of the preceding claims, characterized in that the pixels of a group are distinguished from each other by: - the thickness of a P + type surface layer above a pixel photodiode and / or the presence or thickness of a polycrystalline silicon layer above the photodiode, and / or the presence of a buried P-type layer below a photodiode junction formed by an N-diffusion in a P-type epitaxial layer; and / or the presence of a change in thickness or anti-reflective layer formation over a pixel with respect to the thickness or constitution of this layer. layer above the other pixels.