FR3066344A1 - METHOD FOR CORRECTING NON-UNIFORMITIES IN IMAGE SENSORS - Google Patents

Abstract

A method of calibrating the pixels (41) of a sensor (40) with at least two pixels, in particular a matrix sensor, comprising the steps of: - placing the sensor in the dark, - injecting charges in a controlled manner into the minus one pixel to simulate an illumination, - measure the response of the pixel (s) to this injection of charges, - calculate a nonuniformity correction to be made to the sensor as a function of the measured response.

Description

METHOD FOR CORRECTING NON-UNIFORMITES IN IMAGE SENSORS

The present invention relates to the correction of non-uniformities in image sensors containing at least two pixels.

Background

Despite advances in the manufacturing of image sensors, manufacturing non-uniformities persist due to non-uniformities present in both the semiconductor materials and the active components involved in the reading circuit. This problem of non-uniformity is particularly pronounced in an infrared sensor whose semiconductor materials, often composite, are difficult to produce with great uniformity.

These manufacturing non-uniformities are manifested by differences in photoelectric response between the different pixels that make up the sensor, as shown schematically in FIG. IA for three pixels. They must be corrected in order to obtain a quality image. This correction operation, known as NUC for Non Uniformity Correction in English, can be carried out by digital circuits integrated in the sensor or in the device which includes it, for example a camera. It can also be performed by a computer via image processing software.

NUC processing applies a correction to the photoelectric response curve of the pixels so that this response curve becomes the reference curve after the correction. This requires a phase of taking images with different levels of uniform illumination on the pixel matrix of the sensor. These images are compared to reference images of uniform response, and the difference between the real images and the reference images makes it possible to determine the correction factors to be applied.

Depending on the nature of the non-uniformities, one, two, or even more images are required to precisely determine the correction factors to apply. For a linear photoelectric response sensor, as illustrated in Figure IA, the non-uniformities can be represented in offset and gain. To correct the offset and gain, two images are required, including a D image taken without optical illumination and the other B with uniform illumination on the matrix. Image D provides the OS offset for each pixel, and the difference between images B and D provides the gain G for each pixel. The correction circuit or the correction software applies the transformation: (I-OS) / G to each pixel I. Figure IB illustrates a two-point NUC operation.

If the pixel response is not entirely linear, multiple images with different levels of illumination are required to provide effective correction. An example of a complex correction is the article NUC Algorithm for Correcting Gain and Offset Non-uniformities Parul Goyal IJCSET / March 2011 / Vol 1, Issue 2,70-76.

Today with the progress made in microelectronics, the NUC operation no longer represents any computing difficulty. The main difficulties are found in the modeling of non-uniformities and in the taking of images (called NUC images) with different levels of illumination.

The difficulty in taking NUC images is the uniformity of illumination. It is difficult in practice to create illumination with high uniformity, especially when this NUC image capture cannot be performed in an optoelectronics laboratory. This NUC image capture can in particular be tainted by the presence of dust on the sensor surface or on the optical window of the sensor. The presence of a dust obscures the illumination locally and consequently creates an over-gain of correction for the pixels located below. When this dust is gone, this correction over-gain creates an over-bright pixel.

In addition, this need for uniform illumination makes NUC imaging almost impossible once the sensor is deployed. The photoelectric properties of a sensor can change in its operating environment. For example, temperature variation or aging can alter the response of the pixels, requiring a new NUC calibration. A recalibration of the NUC coefficients in situ would be a real plus for a good quality and lasting NUC correction. summary

The present invention aims to meet this need and it achieves this by means of a method for calibrating the pixels of a sensor with at least two pixels, in particular a matrix sensor, comprising the steps consisting in: - Placing the sensor in the dark , inject charges in a controlled manner into at least one pixel to simulate an illumination, measure the response of the pixel (s) to this charge injection, calculate a non-uniformity correction to be made to the sensor as a function of the measured response.

By placing the sensor in the dark, we must understand to prevent light from reaching its active surface.

The present invention makes it possible to replace the uniform illumination during the taking of NUC images by an injection of charges into the pixels. Charges can be injected into all pixels with the same intensity to simulate a given uniform level of illumination. This charge injection can be controlled by a programmable electrical signal, to simulate the different levels of illumination required to calculate the NUC coefficients.

It is thus possible to successively inject the charges with different intensities in each pixel, to simulate different respective levels of illumination, by measuring the response of the pixel after each injection. The injection of electrical charges makes it possible to take NUC images in situ. The charge injection can be done with one or more injection diodes separate from the photodiodes. This injection can be carried out with an injection diode by photodiode or alternatively with an injection diode for several photodiodes. We can have one photodiode per pixel or more.

A shutter can be released to place the sensor in the dark before performing the calibration.

The calibration process can be implemented whenever the environment of the sensor changes; for example, the environment temperature of the sensor is measured, and a new calibration is performed automatically when the measured temperature differs from a predefined deviation from the temperature at which the previous calibration was performed.

The sensor can be an infrared sensor and include an InGaAs substrate. The injection of the charges into the substrate of a pixel can be carried out using a diode junction polarized in direct during this injection, the injection diode being then used with a reverse polarization when the sensor is used normally .

The injection diode can be used to reduce the glare of the pixel. The subject of the invention is also a device comprising a sensor with at least two pixels, in particular a matrix sensor, comprising: for at least one pixel, a substrate and a diode junction making it possible to inject charges when directly polarized in the substrate, - a sensor calibration control circuit, for controlling the injection of charges into the pixel (s) via said junction (s), the pixels being at the time of this injection placed in a dark condition, measuring the response of the or pixels, calculating from the measured response (s) a correction for non-uniformity to be made to the sensor.

The control circuit can vary the charge injection current so that it takes successively several values, to simulate several uniform levels of illumination.

The characteristics mentioned above in connection with the method apply to the device also. The substrate can thus be made of InGaAs.

The junction of the injection diode can be that of an anti-glare diode, reverse biased during normal operation of the sensor.

The sensor can include as many injection diodes as there are photodiodes. As a variant, the sensor comprises several photodiodes per injection diode.

Each pixel can have a single photodiode.

The device may include a shutter to place the pixels in the dark. This shutter can be controlled by the control circuit.

Description of the Figures The invention will be better understood on reading the description which follows, of non-limiting examples of implementation thereof, and on examining the appended drawing, in which: - Figure IA , previously described, gives an example of non-uniformity of pixel response, FIG. 1B, also previously described, illustrates an example of NUC correction in the example of FIG. 1, FIG. 2A illustrates the generation of photoelectric charges during the illumination of a sensor, FIG. 2B illustrates the simulation of the generation of photoelectric charges by injection into the substrate of a current by a direct polarized diode, FIG. 3 schematically represents a top view of an example of matrix sensor according to the invention, FIG. 4 represents an example of pixel structure according to the invention, with the associated reading circuit, FIG. 5 is a view similar to FIG. 4 of a variant of a pixel, FIG. 6 schematically represents a device according to the invention, and FIG. 7 is a view similar to FIG. 3 of a variant of sensor according to the invention. The invention is applicable to any type of sensor based on the use of PN junction photodiodes or MOS capacitances (CCD).

FIG. 2A illustrates a photodiode 10 produced by doping at 11 of a semiconductor substrate 30. The doped area 11 of the photodiode 10 collects mobile charges in its vicinity and thus generates an electrical signal. The latter can be in the form of charges (CCD sensor) or in the form of voltage (APS sensor). When these moving charges are created by the incidence of photons, the photodiode output signal gives information on the energy or the power of light.

FIG. 2B represents the structure of FIG. 2A in which a PN diode referenced 20 has been added in accordance with the invention. When a current is circulated in the positive direction of polarization, a certain amount of mobile charges is injected into the substrate 30. If this injection of mobile charges takes place in the vicinity of photodiode 10, the charges will be collected without distinction by the latter of their origin. In this case, we can simulate the effect of incident photons by this charge injection. The amount of mobile charges collected by the photodiode 10 is directly linked to the amount of electrical charges injected, therefore to the electric injection current.

In other words, if a PN diode called the injection diode is placed near a photodiode of an active pixel and a controlled electric current is injected, the photodiode collects these injected mobile charges as if a spot of light was present nearby, even when the pixel was placed in total darkness. When this arrangement is made in a matrix of pixels whose pitch is less than the length of diffusion of the mobile charges in the substrate, one can create a density of mobile charges very uniform in the matrix of pixels thanks to the diffusion of the mobile charges.

This arrangement thus perfectly meets the need for NUC imaging. With the injection of charges into the pixels carried out in accordance with the invention, there is no longer any need to use programmable and uniform light sources. Just place the sensor in the dark and take an NUC image with different levels of current injection into the substrate. At each injection level, we take an image and we can take as many images as necessary for the construction of the coefficients of the NUC correction.

A sensor with such pixels, associated with an electrically controlled shutter, makes it possible to update the NUC coefficients in any operating and installation condition.

The PN junction diode can only inject one type of charge into the substrate. For example, an N scattering in a P-type substrate only allows electrons to be injected into the substrate when a negative voltage is applied to the N scattering. When a positive voltage is applied to the N scattering, it will absorb electrons instead of injecting it. A negative voltage is therefore applied to the N scattering of the injection diode in order to inject electrons to simulate an illumination during the NUC imaging for the construction of the NUC coefficients. On the other hand, a positive voltage will be applied to it in order to stop this injection during normal use of the sensor to take images. The absorption of free electrons in the substrate by the injection diode competes with the photodiode during shooting. This drawback can be minimized by using an injection diode smaller than that of the photodiode.

FIG. 3 schematically represents a matrix sensor 40 comprising a plurality of active pixels 41 each comprising a photodiode 10 produced on a semiconductor substrate 30 and a charge injection diode 20, of smaller dimensions than the photodiode.

The ratio r between the surface Sp of the photodiode 10 and that Sd of the diode 20 is for example greater than 1.

The distance d between the photodiode 10 and the diode 20 is for example such that d <lo, where Id denotes the length of diffusion of the charges in the substrate 30.

In Figure 3 there is shown the supply of the injection diodes 20. These are for example, as illustrated, all supplied in parallel by the same line 42 under the same injection voltage.

FIG. 4 is an exemplary embodiment of an active pixel 41 based on a photodiode 10 in voltage mode, that is to say that the voltage on the photodiode 10 is read by an amplifier 43 as an output signal from the pixel. In this case, the photodiode 10 and the injection diode 20 can be in the same substrate 30 as the amplifier 43. They can also, in a variant, be produced in another substrate than that of the amplifier 43. The first case concerns monolithic image sensors. The second relates to hybrid sensors such as infrared sensors where the photodiode 10 and the injection diode 20 are located in a substrate made of a semiconductor material capable of converting infrared photons into electrons or free holes, and are connected to a CMOS read circuit.

FIG. 5 represents an alternative embodiment with a pixel 41 known as charge transfer.

The photodiode 10 is then produced by a buried diffusion, connected to a reading node 44 by a grid 45. A control signal TX from the grid 45 makes it possible to transfer the charge accumulated in the buried photodiode 10 to the reading node 44. The injection diode 20 is placed near the photodiode 10 in order to inject charges when it is directly biased by the injection voltage.

The read node 44 is connected to an amplifier 43 which can be produced in the same substrate or not as the photodiode.

The two examples in FIGS. 4 and 5 use N diffusions in a substrate P for the photodiode and the injection diode. A reverse symmetrical structure, with diffusion P in a substrate N, is of course possible without departing from the scope of the present invention.

FIG. 6 schematically shows a device 100 comprising a sensor 40 according to the invention, associated with an optic 101.

A control circuit 102 makes it possible to control a shutter 103 so as to intercept or not the incident light and place the sensor 40 in the dark for taking NUC images.

During this image capture, a Vinjection voltage is applied to the injection diodes 20 so as to simulate uniform illumination. The application of the Vinjection voltage is carried out according to a program so as to collect all the values necessary for calculating the coefficients of the NUC correction.

The control circuit 102 can be connected to a temperature sensor 105 which makes it possible to detect a change in the environment of the sensor 40 and thus automatically carry out a new calibration when this is desirable, in the event of too large a temperature variation, for example.

In FIG. 6, the shutter 103 is shown in its retracted position, where it does not block the light flux incident on the sensor 40.

In this configuration of normal operation, the control circuit 102 maintains reverse polarization on the diodes 20, in order to block the injection of the charges towards the photodiodes 10.

The control circuit 102 can be made in the form of a component separate from the sensor 40. However, the control circuit can be integrated into the same component as the sensor, this component then comprising for example at least one connection to an interface shutter control 103. The shutter 103 can be integrated with the optics 101 or the sensor 40, if necessary.

There are many other ways to have the injection diodes.

For example, as shown in FIG. 7, an injection diode 20 for several photodiodes 10, as shown in FIG. 7. On the latter, we see that an injection diode 20 is associated with four surrounding photodiodes arranged like the vertices of a square, the injection diode being in the center. The charges injected by this diode during the calibration phase diffuse towards the photodiodes, as illustrated.

Claims (15)

1. Method for calibrating the pixels (41) of a sensor (40) with at least two pixels, in particular a matrix sensor, comprising the steps consisting in: - placing the sensor in the dark, injecting charges in a controlled manner at least one pixel to simulate an illumination, measure the response of the pixel (s) to this charge injection, calculate a non-uniformity correction to be made to the sensor as a function of the measured response. 2. Method according to claim 1, in which the charges are injected with the same intensity in all the pixels to simulate a given uniform level of illumination. 3. Method according to one of claims 1 and 2, wherein the charges are successively injected with different intensities in each pixel, to simulate respective levels of illumination, by performing a measurement of the pixel response after each injection. 4. Method according to any one of the preceding claims, in which a shutter (103) is triggered to place the sensor (40) in the dark before performing the calibration. 5. Method according to any one of the preceding claims, in which the temperature of the environment of the sensor (40) is measured, and a new calibration is carried out when the measured temperature differs from a predefined deviation from the temperature at which the previous calibration has been performed. 6. Method according to any one of the preceding claims, the sensor (40) comprising a substrate (30) InGaAs. 7. Method according to any one of the preceding claims, the injection of charges into the substrate (30) of a pixel being carried out using a diode junction (20) polarized directly during this injection, the diode then being used with reverse bias when the sensor is used normally. 8. The method of claim 7, the pixels each comprising a photodiode, the sensor comprising as many injection diodes as photodiodes. 9. The method of claim 7, the pixels each comprising a photodiode, the sensor comprising more photodiodes than injection diodes. 10. Device (100) comprising a sensor (40) with at least two pixels (41), in particular a matrix sensor, comprising: for at least one pixel (41), a substrate (30) and a diode junction (20) allowing to inject, when polarized live, charges into the substrate (30), - a sensor calibration control circuit (40), in order to o inject charges into the pixel or pixels via said junctions, the or pixels (41) being during this injection placed in a dark condition, o measure the response of the pixel (s), o calculate from the measured responses the correction of non-uniformity (NUC) to be made to the sensor. 11. Device according to claim 10, the substrate being in InGaAs. 12. Device according to one of claims 10 and 11, the diode junction being that of an anti-glare diode, reverse biased during normal operation of the sensor. 13. Device according to any one of claims 10 to 12, comprising a shutter (103) for placing the pixels in the dark. 14. Device according to claim 13, the shutter (103) being controlled by the control circuit (102). 15. Device according to any one of claims 10 to 14, each pixel comprising a photodiode and the sensor comprising as many injection diodes as there are photodiodes.