WO2019166720A1

WO2019166720A1 - Dynamic detection of stray light in a digital image

Info

Publication number: WO2019166720A1
Application number: PCT/FR2019/050411
Authority: WO
Inventors: Joaquin Cornejo Bautista; Matthias Mutschler
Original assignee: Continental Automotive France; Continental Automotive Gmbh
Priority date: 2018-02-27
Filing date: 2019-02-22
Publication date: 2019-09-06
Also published as: FR3078427A1; FR3078427B1

Abstract

The invention discloses a method and a device for dynamically detecting a trace of stray light in a digital image acquired by an optical device. The method consists of extracting and determining, without complex calculations, values derived from known parameters of the image and comparing said values in real time with thresholds determined empirically to confirm or deny the presence of a trace of stray light in one or more images.

Description

Dynamic detection of stray light in a digital image

The present invention generally relates to digital processing performed, when acquiring images by optical devices, to compensate for defects or overcome limitations intrinsic to the devices in question.

It relates more particularly to a method and a device for detecting the presence of stray light in an image.

The invention finds applications, in particular, in the processing of images acquired by video cameras equipping systems for driving assistance of motor vehicles.

Advanced Driver Assistance Systems (ADAS) are becoming more common in today's motor vehicles. They continue to evolve by offering an increasing number of features with the aim of improving the comfort and safety of vehicle users. In the long term, it is conceivable to produce self-driving vehicles.

Many of the features offered are based on a real-time analysis of the immediate environment of the vehicle in which the ADAS is embedded. Whether it is for example to help emergency braking or to improve the control of trajectories of a car, the taking into account at any moment of characteristics of this environment is necessary to ensure the functionalities such systems.

In this context, one of the possibilities for acquiring information relating to the immediate environment of a motor vehicle is to use optical means. A vehicle can thus be equipped with one or more video cameras intended to observe its environment, at any moment, in one or more observation directions. The images collected by the cameras are then processed by the ADAS. It can use them to potentially trigger a driver assistance action (for example trigger an audible and / or visual alert, or even trigger a braking or a change of trajectory of the vehicle ...).

This type of operation assumes that the images collected by the camera (s) can be considered as exploitable, that is to say sufficiently representative of the reality of an observed scene. However, when acquiring images by a video camera under natural conditions (that is to say without mastering the different ambient light sources), it is common to see the quality of the images acquired degraded by the acquisition. unintentional stray light. This is the case, for example, if light picked up comes from an expected source but has taken an unintended path (for example, linked to internal reflections) in the optical device (ie the camera ) due to imperfections of said device. This is also the case if the light comes from an unexpected source but still "involuntarily" captured by the device, because, again, imperfections of said device.

Stray light pickup (which may be referred to as "flare") is inherently related to the imperfections of a real video camera in on-site operation (ie not under created lighting conditions). artificially). It causes the addition of a noise to the signal corresponding to one or acquired images, and is therefore at the origin of a degradation of the contrast, the luminance and / or the balance / saturation of the colors of the images. acquired images. Therefore, the faithful reproduction, by the camera, of the characteristics of a real scene is difficult or impossible in real conditions of use. At least with standard cameras on the market compatible with applications in the automotive field.

However, to be able to detect, for example, the proximity of another vehicle, the occurrence of a specific object in the path of a vehicle or the color of an object present in the field of vision of its driver, the image must be as faithful as possible to the scene actually observed. In other words, the acquisition of poor quality images by a video camera of an ADAS may be responsible for its inability to trigger a useful action, or conversely be responsible for triggering an action unsuited to the actual situation.

The image parameters of contrast, luminance, and true color reproduction are therefore essential to the performance of an ADAS using video. It is for this reason that a lot of effort is made to ensure the most accurate image acquisition possible.

These efforts include the physical optimization of optical image acquisition devices. Some approaches aim, for example, through treatments applied to the optical elements (the lenses in particular) of the device (for example reflective, polarizing or absorbent surface treatments), to minimize the risks of capturing stray light. However, the high cost of this type of treatment is often incompatible with an application in the automotive field.

These efforts also involve digital processing intended to compensate, a posteriori, the possible defects of the images acquired. This is particularly the case for high dynamic range imaging (HDR) Dynamic Range "), which relies on the (almost) simultaneous acquisition of images with different exposure times to improve the rendering / fidelity of these images. More specifically, by combining different images acquired with respective exposure times of varying lengths, a single image is formed which avoids overexposure and / or underexposure of certain areas of the image. The very dark areas and the very bright areas can thus be represented simultaneously and faithfully in a single image.

By way of example, US Pat. No. 9,307,207 B2 discloses an HDR acquisition method in which the glare of certain areas of the sensor of a camera (in English "glare") is reduced thanks to the acquisition and the combination of two images acquired with respectively long and short exposure times.

Such an approach imposes strict conditions of stability both to avoid the possible blur associated with a long exposure time and so that the two images can be combined correctly. Moreover, the solution proposed in this document also involves a determination of the parts of an overexposed image based on the observation of a predetermined brightness level associated with a known source. However, because of the diversity of real situations, since it is a question of overexposure and not of saturation, the threshold at which a portion of the image is considered to be overexposed can be very difficult. to be determined.

This technique is therefore only suitable for the correction of saturated areas of an image, or possibly, for overexposed areas in certain specific and restrictive cases. However, it does not correct a flare effect. An image affected by this type of effect will remain degraded, especially in terms of contrast or color balance, despite the use of this type of HDR method.

Finally, the two approaches discussed above are limited since they do not in any way to relieve some of all the effects associated with the capture of stray light. As a result, the issue remains crucial, in the context of the use of digital images by an ADAS, to know if an image is affected or not by this type of effect and, subsequently, if it is exploitable or not as such by the ADAS.

US Patent 9,497,427 B2 discloses in particular a method comprising an algorithm for detecting the presence of flare in an image. The algorithm relies on an analysis of the differences / color changes of neighboring pixels (in the digital image acquired by the camera sensor) for recognize shapes or contours characteristic of the presence of stray light in the image. He is also working to discriminate these changes from those possibly related to the presence, in the image, of an edge of any object that is likely to produce similar effects. However, this method presupposes the ability to clearly establish a "geometric signature" of the presence of stray light in the image.

In addition, the algorithm used here performs its various operations on a two-dimensional matrix (2D), namely the pixel matrix of the camera sensor, in order to take into account the notion of neighborhoods in X and / or in Y In other words, each operation is potentially expensive in computing resources and in memory for carrying out the operations of the algorithms associated with this type of method.

The invention aims to eliminate, or at least mitigate, all or part of the disadvantages of the prior art mentioned above.

For this purpose, a first aspect of the invention proposes a method of dynamically detecting a trace of stray light in a digital image acquired by an optical device during an exposure time t, said image being represented in a YUV color space. deducible from the RGB color space and giving a triplet of components R, G and B to all the pixels of the image, by a luminance component Y, a first chrominance component U resulting from the difference between the components B and G , and a second chrominance component V resulting from the difference between the components R and G, and with an additional component S describing the chromatic saturation and resulting from the difference between the maximum of the components R, G and B and the minimum components R, G and B, said method comprising, during the acquisition of at least one digital image by the optical device:

Calculating, for all the pixels of the digital image, a statistical distribution Sh of the values of the chromatic saturation S and the determination of the width of this distribution named Sh _w at a determined height;

• the determination, on all the pixels of the digital image, of the maximum value of the luminance Y named Y _max ;

Time filtering, for a determined number of digital images successively acquired by a determined filtering function, values of the width of this distribution Sh _w , the maximum of the luminance Y _max and the exposure time t; • checking the detection of a trace of stray light, the detection being considered positive if, and only if:

the filtered value of the width of this distribution Sh _w is less than a determined chromatic saturation threshold; and

the filtered value of the maximum of the luminance Y _max is greater than a first determined luminance threshold.

Embodiments taken alone or in combination, further provide that:

The detection of a trace of stray light is further considered positive if:

the filtered value of the exposure time t is less than a determined threshold,

The method further comprises:

the determination, on all the pixels of the digital image, of the minimum value and the maximum value of the first chrominance component U respectively named U _min and U _max ;

the determination, on all the pixels of the digital image, of the minimum value and the maximum value of the second chrominance component V respectively named V _{min and} V _max ;

the determination of the maximum value S _max of the chromatic saturation component S;

calculating a difference value of the first component of the chrominance U _D m by subtracting the minimum value U _min from the first chrominance component U with the maximum value U _max of the first chrominance component U;

calculating a difference value of the second component of the chrominance Vom by subtracting the minimum value V _min from the second chrominance component V with the maximum value V _max from the second chrominance component V;

- the application of time filtering to each of the previous values; and

when checking the detection of a trace of stray light, the filtered value of the maximum of the luminance Y _max is greater than a first determined luminance threshold, said detection being considered positive furthermore if: The filtered value of the difference value of the second component of the chrominance V _D i is less than a first determined chrominance threshold; and,

The filtered value of the difference value of the first component of the chrominance U _D m is less than a second determined chrominance threshold; and,

The filtered value of the maximum value of the chromatic saturation component S _max is less than a determined threshold; and,

The filtered value of the maximum of the luminance Y _max is greater than a second luminance threshold determined below the first luminance threshold,

All determinations of minimum or maximum values of a component are only obtained for only part of the pixels of the digital image corresponding to a predetermined region of interest,

The method further comprises checking the negative detection of a trace of stray light, the detection being considered negative if, and only if:

the filtered value of the exposure time t is greater than a determined threshold; and

the filtered value of the maximum of the luminance Y _max is less than a determined threshold,

• the temporal filtering of the filtering function is a function of low-pass type where the equation governing the calculation of a filtered value for the n ^th image of a sequence of images acquired is of the type:

filtered_value (n) = filtered_value (n-1) x a + non-filtered_value (n) x (1- a) where n is an integer and where a is a predetermined value between 0 and 1,

If an acquired digital image results from a combination of several images with several different exposure times, then the value of the exposure time t used by the method to analyze this digital image is the value of the exposure time the shortest of the different acquisitions that are at the origin.

The optical device is a digital video camera composed of at least one lens and a photographic sensor. • the digital video camera is embedded in a motor vehicle and is part of a system for assisting the driving of said motor vehicle.

According to a second aspect, the invention also relates to a device for dynamically detecting a trace of stray light in a digital image acquired by an optical device during an exposure time t, said image being represented in a YUV color space deductible from the RGB color space which gives a triplet of components R, G and B to all the pixels of the image, by a luminance component Y, a first chrominance component U resulting from the difference between the components B and G , and a second chrominance component V resulting from the difference between the components R and G, and further an additional component S describing the chromatic saturation and which results from the difference between the maximum of the components R, G and B and the minimum of the components R, G and B, said device comprising means for, at the time of acquisition of at least one digital image by the optical device:

Compute, for all the pixels of the digital image, at least one statistical distribution, from at least one of the values corresponding to a component of the image, and determine the width of this at least one distribution to a specific height;

• determining, on all the pixels of the digital image, at least one minimum value and one maximum value for at least one component of the image;

• filtering, for a determined number of digital images acquired successively, by a determined filtering function, determined values for components of the image; and

• check the detection of a trace of stray light by comparing filtered values or not with determined thresholds and considering the positive detection according to the result of these comparisons,

and the device being adapted to implement all the steps of a method according to the first aspect.

According to a third aspect, the invention finally relates to a driving assistance system comprising a device, according to the device according to the second aspect of the invention, capable of implementing all the steps of the method according to the invention. any of the embodiments of the method according to the first aspect of the invention. Other features and advantages of the invention will become apparent on reading the description which follows. This is purely illustrative and should be read in conjunction with the attached drawings in which:

- Figure 1 is a schematic representation of a motor vehicle equipped with a video camera for driving assistance;

FIG. 2 is a diagram of steps of an embodiment of the method according to the invention; and

- Figure 3 is a step diagram of another embodiment of the method according to the invention.

In the following description of embodiments and in the figures of the accompanying drawings, the same or similar elements bear the same numerical references.

Figure 1 schematically shows the operation of a video camera 102 embedded in a car 101 to provide images to an automotive assistance system, ADAS (not shown).

In the example shown, the camera scans the immediate environment of the motor vehicle in its forward direction of travel. Since the video camera provides digital images, these can be analyzed by an ADAS processing unit to detect specific objects or events. The ADAS can thus exploit the information contained in the images acquired by the camera 102 to take a decision or to trigger a driving assistance action.

For example, if a pedestrian suddenly enters the field of view 103 of the camera 102, and if it is recognized as such, then the ADAS can trigger an emergency brake. In another example, if a traffic light is in the field of view 103 and the processing unit of the ADAS is able to recognize the color, the ADAS may pass on this information to the driver of the vehicle or, if necessary, block the vehicle ahead in the event of a red light.

These examples are only part of the situations in which the use of one or more camera (s) can provide images to an ADAS causing the realization of some of these features.

Furthermore, it is clear that in such a situation the circumstances leading to the capture of stray light by the camera 102 can be numerous. Multiple causes of stray light in camera-acquired images can occur. It may be, for example, a reflection of the sun outside the field of view 103 but captured not intentionally. It may also be the light of a lighthouse observed intentionally but using an alternative optical path in the camera 102. These examples are not limiting.

Referring to Figure 2, we will now describe an embodiment of the method according to the invention.

The method described here aims to determine whether an image is affected by traces of stray light or not. By image is meant here a digital image, acquired by an optical and electronic device, composed of a matrix of pixels covering the entire surface of the image. The method is therefore applicable to a digital image acquired by different types of optical devices when it is ultimately acquired by a digital sensor (CMOS or CCD imager, for example) able to convert the optical signal into a digital image , that is to say in numerical values. This is for example the case of a digital video camera or a digital camera.

In a first step, the digital image resulting from this type of optical device and to which the invention applies is conventionally represented in the RGB (Red, Green and Blue) or RGB color space. French (for Red, Green, Blue). This is a representation, known to those skilled in the art, implying that each pixel of the image is composed of three different components respectively red, green and blue, where the intensity of each component determines the color of the image. a pixel while the cumulative intensity of these different components determines its luminance (that is to say the sensation of perceived brightness for this pixel).

However, for practical reasons which will appear more clearly in the light of the description of the steps of the method which will follow, the method relies, in its different variants, on the use / analysis of digital images represented in a another color space called YUV space. It is indeed advantageous to use this other color space than the RGB space, in which the components (values) of each pixel can be directly obtained by transformation from its components in the RGB space. More specifically, the components Y, U, V and S as described below are for example obtained from the components R, G and B of a color image reconstructed by the application of a "demosaicing" algorithm or "Demosaicing" ("demosaicing" in English) known in itself to those skilled in the art, on the initial raw image in RGB format.

Thus, the components of the images exploited, in the context of the invention, are the following: • Y is a luminance component resulting from the sum of the original components R, G and B;

• U is a first so-called chrominance component resulting from the difference between the original components B and G; and,

• V is a second so-called chrominance component resulting from the difference between the original components R and G.

In addition to these three components, related to the colorimetric representation space of the image used in the process, a fourth component is used for the implementation of the process steps. This fourth component is also directly related to the original components R, G and B. In fact, the component S, called chromatic saturation component results, for its part, from the difference between the maximum of the three components R, G and B and the minimum of these same components.

Finally, in addition to all the parameters described so far, another parameter can be taken into account for carrying out the steps of the method: it is the exposure time t of each digital image. Each image, acquired for example by a video camera, may indeed have been during a variable exposure time depending on the lighting conditions.

Moreover, a person skilled in the art will appreciate that, in the case where the image analyzed itself is derived from a combination of images, for example in the case of a high dynamic range image acquisition HDR, the the value of the exposure time used as input value of the process in its different implementation variants will be the shortest value of the different exposure times of the images at the origin of the combination.

Thus, in the embodiment of the method presented hereinafter with reference to FIG. 2, the parameters, resulting from a digital image acquired and used as input to the algorithm according to the method, are represented in step 201. These are exposure time t, chromatic saturation S and luminance Y.

Step 202 then consists, in a first step, in obtaining a statistical distribution Sh of all the chromatic saturation values S for all the pixels of the processed image. For example, for a digital image of 2 megapixels, the entire pixel is scanned to determine the value of the chromatic saturation S associated with each pixel in order to obtain this distribution. Thus the distribution Sh gives, for each value of S possible (for example from 0 to 255), the number of pixels of the image having this value. In a second step, step 202 also consists in determining the width Sh _w of this distribution Sh at a given height. Such a distribution, obtained from an image acquired under real conditions, has indeed a maximum and a typical shape for determining a width at a certain height of the distribution. The choice of this height, according to modes of implementation, may depend directly on the lighting profile of the scene captured by the acquired image. More specifically, the experiment shows that a color image affected by the presence of stray light sees the shape of its distribution become narrower than when no trace of stray light affects this image. If the distribution is wide, we can conclude that the image is not affected by a trace of stray light. On the other hand, other image profiles may have a narrow distribution form without being affected by the trace of stray light. This is why step 203, based on another criterion, is introduced to detect the trace of stray light.

Step 203 consists in determining the maximum value Y _max , on the set of pixels of the image, of the luminance parameter Y. In the same way as to be able to establish the distribution Sh from the chromatic saturation S, l The set of pixels in the image is scanned in order to extract the highest luminance value Y from the entire pixel matrix. This value makes it possible in particular to establish, through its proximity to the maximum possible value which corresponds to the saturation of a pixel, the fact that at least one zone of the image is overexposed or saturated, for example. In this sense, it gives an additional indication of the presence or absence of a trace related to the presence of stray light in an image. In particular, it makes it possible to discriminate that a narrow distribution, possibly observed in step 202, is linked to natural lighting conditions (for example cloudy or rainy weather) or to the presence of stray light in the picture.

Those skilled in the art will appreciate that in another embodiment of the method, the order of the two steps 202 and 203 may be reversed without changing the rest of the process flow. In addition, the maximum luminance value Y _max may also, in another embodiment, be determined for only part of the pixels of the image. We speak of a region of interest (in English "Region Of Interest" or ROI) of the image which is analyzed preferentially by the ADAS, especially in the case where it is expected that the camera observes an object or an event accurate in a specific area of his field of vision.

Step 204 consists of interrogating several conditions, relating to the parameters used or determined during the preceding steps, to check whether a trace of stray light is detected or not in the image. In the case described here, the exposure time values t, of width Sh _w of the statistical distribution Sh of the values of chromatic saturation and maximum luminance Y _max are compared with thresholds for determining whether a trace of stray light is detected or not.

Regarding the value of the exposure time t, its use is based on the fact that, typically, for a camera automatically adjusting its exposure time to the brightness of a scene, the information of a time exposure means that the observed scene is dark. Thus, the probability of having stray light is zero.

The comparison of the three different values t, Sh _w and Y _max to given threshold values, which are determined according to the specific characteristics of the sensor used, thus provides three complementary information in the detection of a trace of stray light. This makes this approach all the more robust. That being so, the comparison of the exposure time value t can be considered as not essential to the verification performed by the method on an image. This is why a variant of the method may consist in carrying out the same steps without exploiting the exposure time value t.

Finally, in the embodiment illustrated in FIG. 2, it is considered that an image is affected by the trace of stray light if it fulfills three conditions simultaneously. Specifically, if the exposure time value t is less than a threshold t _th , if the width value Sh _w is smaller than a threshold Sh _w-th and the luminance value mamimum Y _max is greater than a threshold Y _max-thi . If so, the algorithm considers that a trace of light has been detected for the analyzed image. Otherwise, the image is considered blank of any stray light.

Step 205 then resumes the result of step 204 in the form of the result delivered by the method indicating to the ADAS or to any user of the image, whether or not the image is affected by the presence of stray light.

Advantageously, the method does not perform complex operations on a 2D image to obtain this result but relies on 1D signals, that is to say a series of N x M pixels of the image considered. in sequence, for an image, originally N lines and M columns (where N and M are integers). In addition, the method uses values that can be readily derived from a given color space to make comparisons, namely the color space associated with the color space. acquisition camera as the RGB system in the example. As a result, the method does not require any significant calculation or memory resources and can all the more easily be implemented in real time to process each digital image acquired, for example by a video camera embedded in a motor vehicle.

Furthermore, another advantage of the method is that it does not require to assume the shape of a trace left by stray light in an image or, in general, its impact on the image in two dimensions. The determination and exploitation of 1D signals and single values from the image can detect whether it is affected by stray light or not.

Figure 3 illustrates another embodiment of the method according to the invention.

In this embodiment, the parameters of a digital image acquired and represented in step 301 which are used at the input of the algorithm include, in addition to the parameters of luminance Y, chromatic saturation S and exposure time t of the mode of implementation of FIG. 2 already described above, the parameters of the chrominance components U and V.

Step 302 is a step of obtaining a statistical distribution Sh of chromatic saturation values S and its width Sh _w identical to step 202 of the embodiment of FIG. 2.

On the other hand, during step 303, a set of values used in the remainder of the method is determined and comprising:

The maximum value Y _max of the luminance Y obtained on all the pixels of the image;

• the minimum values L in and maximum U max, obtained on all the pixels of the image for the first chrominance component U;

• the minimum values V _mm and maximum V _max , obtained on all the pixels of the image for the second chrominance component V; and

The maximum value S _max of the chromatic saturation S obtained on all the pixels of the image.

Similarly to the embodiment of FIG. 2, the order of the two steps 202 and 203 can be reversed, and the determination of the minimum and maximum values can be done only for a given region of interest in the image.

Step 304 simply consists in calculating U _uiff values by subtracting U _min from U _ma x and Vom by subtracting V _m m from V _ma x. Indeed, these differences reflect the color dynamics of the image. In this sense, they are an additional indication of the presence or absence of a trace of stray light in the image (or region of interest considered, if any) since it tends to reduce these dynamics.

During step 305, all the values calculated during the previous steps and intended to be used in the following steps can advantageously be filtered in time over several images acquired successively. More precisely, a temporal filtering function can be applied to the Y _max , S _max , t, Sh _w , Uiff and \ L values. By performing this type of filtering on several determined values for successive acquisitions, the effects of potential artifacts that would appear on a single image within a sequence can be "smoothed" or even deleted, so that the algorithm is not not affected by such artifacts.

This may be achieved, for example, with a function of temporal filtering of low-pass type where the equation governing the calculation of a filtered value for the n ^th image of a sequence of images acquired is of the type:

filtered_value (n) = filtered_value (n-1) xa + non-filtered_value (n) x (1- a) where n is an integer and where a is a predetermined value between 0 and 1, depending on whether the filter is more or less "Stiff".

Those skilled in the art will appreciate that this type of temporal filtering can be used irrespective of the embodiment of the implemented method and, in particular, whatever the different parameters used for determining the detection of a trace of stray light. by the process in question.

Step 306 consists, in the same way as for step 204 relating to the embodiment described with reference to Figure 2, to interrogate several conditions. These conditions relate to the parameters used or determined during the preceding steps and make it possible to check whether a trace of stray light is present in the image. In this embodiment, in addition to the values t, Sh _w and Y _max already used in the embodiment of FIG. 2, the values U _Diff , V _{D 1} and S _max are also compared with thresholds. In addition, their comparison is completed by a second comparison of the Y _max value, already compared with a first threshold value as in the embodiment of Figure 2, but with a second threshold value less restrictive. In other words, a second set of parameters is compared with threshold values to further enhance the robustness of the algorithm in detecting a trace of stray light in the image.

In particular, the conditions verified in step 306 are:

• the value of the width Sh _w is less than a determined threshold Sh _w-th ; and • the filtered value of the exposure time t is less than a determined threshold t _th ; and

The filtered value of the maximum luminance Y _max is greater than a first determined threshold Y _thi ; and by the way,

if the latter condition is not fulfilled, then the following conditions are examined:

The filtered value of the difference V _D i is below a determined threshold; and

The filtered value of the difference U _diff is below a determined threshold; and

The maximum filtered value of the chromatic saturation S _max is below a determined threshold; and

The maximum filtered value of the luminance Y _max is greater than a second determined threshold Y _th 2 lower than the first threshold Y _thi .

Assuming the verification of all these conditions simultaneously, the detection of the presence of a trace of stray light in the image is validated. In other words, at the end of step 306, the detection of a trace of stray light is positive (logic value 1, or TRUE). This state (the detection of parasitic light considered positive) is then stored in memory, for example through a variable, until the possible verification of the condition 307 for a next image. Thus, when the conditions of step 306 are not all verified, step 307 is performed.

Step 307 is a step of invalidating this same detection. In particular, when the detection of a trace had been confirmed for a previous image, if the conditions of step 307 are checked for a current image, then the detection is invalidated. In an example where a variable reflects the state of this detection, it can thus, after taking a state corresponding to a positive detection, resume the state corresponding to a negative detection.

In step 307, the following conditions are true:

• the filtered value of the exposure time t is greater than a determined threshold; and

• the filtered value of the maximum luminance Y _max is below a determined threshold.

Assuming the verification of all these conditions simultaneously for an image, the detection of the presence of a trace of stray light in the image is invalidated. In the opposite case, the state of the detection remains identical to that it was for the previous image. It will therefore be noted that the conditions examined during step 307 are not simply inverse, that is to say symmetrical, of the conditions examined during step 306. More specifically, these conditions are different in that they are more demanding, that is to say once the detection of the presence of stray light in an image has been considered positive so that the device goes into a state where the image is considered to be affected by In the case of stray light, it is more difficult to invalidate this detection with respect to a subsequent image in a series of successively acquired images in order to return the device to the state in which the images acquired are considered to be blank of stray light. Specifically, this means that on a sequence of images analyzed in real time, after determining that one or more images were affected by stray light, stricter conditions must be met in order to be able to switch back to a situation where images are considered reliable (ie without stray light) and can therefore be exploited by the ADAS. Those skilled in the art will appreciate that the latter advantage is valid not only when an ADAS exploits the images but, more generally, regardless of the user or the system exploiting the images acquired.

The hysteresis produced by the association of two groups of unsymmetrical conditions to validate and invalidate, respectively, the detection of stray light in the images, thus reinforces the guarantees for the use of images faithful by the ADAS. for triggering specific driving assistance operations.

Finally, in step 308, the result from steps 306 and 307 is delivered to the ADAS and / or is returned to a user in any suitable form.

The present invention has been described and illustrated in the present detailed description and in the figures of the accompanying drawings with reference to non-limiting embodiments. The present invention is not limited, in fact, to the embodiments presented. Other variants and embodiments may be deduced and implemented by the person skilled in the art upon reading the present description and the drawings.

In the claims, the term "include" does not exclude other elements or other steps. A single processor or several other processing units may be used to implement the invention. Similarly, several memories, possibly of different types, can be used to store information. The various features presented and / or claimed can be advantageously combined. Their presence in the description or in different dependent claims do not exclude this possibility. The reference signs can not be understood as limiting the scope of the invention.

Claims

A method of dynamically detecting the presence of a trace of stray light in a digital image acquired by an optical device, said image being represented in a YUV color space deductible from the RGB color space and giving a triplet of R components, G and B to all the pixels of the image, by a luminance component Y, a first chrominance component U resulting from the difference between the components B and G, and a second chrominance component V resulting from the difference between the components R and G, and with an additional component S describing the chromatic saturation and which results from the difference between the maximum of the components R, G and B and the minimum of the components R, G and B, said method comprising, during the acquiring at least one digital image by the optical device:

Calculating, for all the pixels of the digital image, a statistical distribution Sh of the chromatic saturation values S and the determination of the width of this distribution Sh _w at a given height;

Time filtering, for a determined number of digital images acquired successively, by a determined filtering function, values of width Sh _w , and maximum luminance Y _max ;

• the verification of the detection of the presence of a trace of stray light, the detection being considered positive if, and only if:

the filtered value of the width Sh _w is less than a determined chroma saturation threshold; and,

the filtered value of the maximum luminance Y _max is greater than a first determined luminance threshold.

2. Method according to claim 1, wherein the digital image is acquired by an optical device during an exposure time t, the time filtering step further comprising filtering the exposure time t for a predetermined number digital images acquired successively, the detection of the presence of a trace of stray light being further considered positive if:

the filtered value of the exposure time t is less than a determined threshold.

3. Method according to one of claims 1 and 2, further comprising: The determination, on all the pixels of the digital image, of the minimum value and the maximum value of the first chrominance component U respectively named U _mm and U _max ;

• determining, on all the pixels of the digital image, the minimum value and the maximum value of the second chrominance component V respectively named V _min and V _max ;

Calculating a difference value U _D m by subtracting the minimum value U _min from the maximum value U _max of the first chrominance component;

Calculating a difference value Vom by subtracting the minimum value V _min from the maximum value V _max of the second chrominance component;

• the application of time filtering to each of the previous values; and

When the detection of the presence of a trace of stray light is verified, the filtered value of the maximum luminance Y _max is greater than a first determined luminance threshold, the said detection being considered positive if:

the filtered value of the difference of the second chrominance component L is less than a first determined chrominance threshold; and,

the filtered value of the difference of the one of the first chrominance component U _D m is less than a second determined chrominance threshold; and,

the filtered value of the maximum chromatic saturation S _max is below a determined threshold; and,

the filtered value of the maximum luminance Y _max is greater than a second determined luminance threshold less than the first luminance threshold.

The method of any one of claims 1 to 3, wherein all minimum or maximum value determinations of a component are only obtained for only part of the pixels of the digital image corresponding to a predetermined region of interest. .

The method according to any one of claims 1 to 4, further comprising checking the negative detection of the presence of a trace of stray light, the detection being considered negative if and only if:

• the filtered value of exposure time t is greater than a determined threshold; and

The filtered value of maximum luminance Y _max is below a determined threshold.

6. A method according to any one of claims 1 to 5, wherein the filtering function of the temporal filtering is a type of low-pass function, where the equation governing the calculation of a filtered value for the n ^th image of a sequence of acquired images is of the type:

filtered_value (n) = filtered_value (n-1) x a + non-filtered_value (n) x (1- a) where n is an integer and a is a predetermined value between 0 and 1.

The method of any one of claims 1 to 6, wherein if an acquired digital image results from a combination of multiple images with several different exposure times, then the exposure time value t used by the method for analyzing this digital image is the value of the shortest exposure time of the various acquisitions which are at the origin thereof.

The method of any one of claims 1 to 7, wherein the optical device is a digital video camera composed of at least one lens and a photographic sensor.

9. The method of claim 8, wherein the digital video camera is embedded in a motor vehicle and is part of a system for assisting driving of said motor vehicle.

10. A device for dynamically detecting the presence of a trace of stray light in a digital image acquired by an optical device during an exposure time t, said image being represented in a YUV color space deductible from the RGB color space which gives a triplet of components R, G and B to all the pixels of the image, by a luminance component Y, a first chrominance component U resulting from the difference between the components B and G, and a second chrominance component V resulting from the difference between the components R and G, and furthermore an additional component S describing the chromatic saturation and which results from the difference between the maximum of the components R, G and B and the minimum of the components R, G and B,

said device comprising means for, when acquiring at least one digital image by the optical device: Compute, for all the pixels of the digital image, at least one statistical distribution, from at least one of the values corresponding to a component of the image, and determine the width of this at least one distribution to a specific height;

• check the detection of the presence of a trace of stray light by comparing filtered values or not with determined thresholds and considering the positive detection according to the result of these comparisons,

and the device being adapted to carry out all the steps of a method according to any one of the preceding claims.

11. A driving assistance system comprising a device according to the device according to claim 10, adapted to implement all the steps of the method according to any one of claims 1 to 9.