Generating daylight images in a vehicle being driven at night

Abstract

The invention relates to an image-generating method implemented in a motor vehicle, comprising a current phase comprising the following steps: - obtaining (201) a first image representative of a scene facing the vehicle; - obtaining (202) additional information relating to the scene facing the vehicle; - determining (203) a second image depending on the first image and on the additional information by applying an image-generating model stored in the vehicle, the second image having a greater brightness than the first image; - transmitting (204; 205) the at least one second image to a module in charge of a function of the vehicle, with a view to implementing the function at least depending on the at least one second image.

Description

Generation of daytime images in a vehicle during a night driving situation

[0001] The present invention relates to the field of assistance in driving a motor vehicle.

[0002] It is particularly advantageous in enabling an improvement in driving conditions when the motor vehicle is driven at night.

[0003] Driving at night can be difficult and even stressful for a driver, especially on poorly lit or unlit roads. Identifying elements outside the vehicle can be complex due to limited visibility compared to daytime lighting conditions. The lighting functions implemented by the motor vehicle do not always adequately compensate for this lack of visibility.

[0004] Furthermore, certain vehicle functions, in particular adaptive lighting or driving assistance functions, are based on images acquired by a vehicle camera, positioned so as to obtain images representative of a driving scene facing the vehicle, i.e. in front in the front half-space of the vehicle oriented in the direction of travel of the vehicle. The detection of relevant elements on the basis of which the aforementioned functions are implemented is less reliable at night than during the day, which can lead to misinterpreting or even ignoring an object or a person in the scene in front of the vehicle.

[0005] Thus, the deterioration of visibility conditions at night can lead to safety problems, due to driver errors or errors in functions implemented in a motor vehicle.

[0006] There is thus a need to improve the identification of objects in a scene facing a motor vehicle during night driving.

[0007] To this end, a first aspect of the invention relates to an image generation method, implemented in a motor vehicle, comprising a current phase comprising the following steps: - obtaining at least one first image representative of a scene facing the vehicle; - obtaining at least one additional information relating to the scene facing the vehicle; - determining at least one second image as a function of said at least one first image and as a function of said at least one additional information by applying an image generation model stored in the vehicle, said second image having a luminance greater than said first image; - transmitting said at least one second image to a module in charge of a function of the vehicle, for implementation of said function at least as a function of said at least one second image.

[0008] Thus, the invention makes it possible to generate a second brighter image, therefore closer to daytime brightness, than a first image previously available to the vehicle. Such a second brighter image makes it easier to detect elements of the scene facing the vehicle and thus makes it possible to improve the precision associated with the function implemented in the vehicle.

[0009] According to embodiments, said at least one first image may come from an image acquisition device capable of obtaining images of the scene facing the vehicle.

[0010] Thus, the second image is brighter than an image acquired in real time by an image acquisition device. This makes it possible to have a higher brightness in the second image than in the first image, but also higher than the ambient brightness outside the vehicle.

[0011] According to embodiments, said at least one second image can be transmitted:- to a driving assistance module capable of implementing at least one driving assistance function;- to a lighting module responsible for implementing a lighting function;- to a control module of a lighting module capable of determining a photometry to be transmitted to the lighting module to perform a lighting function; and / or- to a human-machine interface of the vehicle, comprising at least one screen, for displaying said at least one second image on said at least one screen.

[0012] Thus, the generation of the second image allows for the improvement of several functions implemented in the vehicle, including functions that improve the comfort and safety of driving the vehicle.

[0013] According to embodiments of the invention, the image generation model may be an artificial neural network.

[0014] This makes it possible to generate second realistic images, representative of the scene facing the vehicle, and by increasing the brightness of the first image.

[0015] According to embodiments of the invention, the method may further comprise a preliminary phase of constructing said image generation model comprising training the image generation model from training data.

[0016] Thus, it is made possible to build the image generation model by machine learning, which strengthens the accuracy of the second image, and therefore the accuracy of the function(s) implemented on the basis of the second image.

[0017] Additionally, the image generation model may be constructed by a generative adversarial network comprising a generation module and a discrimination module, the model being implemented by the generation module, wherein the training data comprises:- training night images acquired in night driving situations and representative of driving scenes in front of vehicles, and associated additional training information, and- training day images acquired in day driving situations.

[0018] This simplifies the training phase, particularly the creation of the training database.

[0019] According to one embodiment, said at least one additional information may comprise a thermal image of the scene facing the vehicle.

[0020] The additional information thus makes it possible to improve the identification and localization of elements of the scene facing the vehicle, in the second image, in particular other vehicles or living elements, such as pedestrians, cyclists or animals.

[0021] According to one embodiment, said at least one additional information may comprise at least one motion vector associated with the first image.

[0022] Thus, the additional information makes it possible to improve the identification of an element of the scene facing the vehicle, in the second image.

[0023] Additionally, said at least one additional information may comprise a set of motion vectors, each motion vector being associated with an area of ​​the first image.

[0024] Thus, the additional information makes it possible to improve the identification and localization of elements of the scene facing the vehicle, in the second image.

[0025] According to embodiments, said at least one additional information may comprise a luminance image representative of the scene facing the vehicle.

[0026] Thus, the additional information makes it possible to improve the identification and localization of elements of the scene facing the vehicle, in the second image.

[0027] According to one embodiment, the first image may be acquired when a brightness outside the vehicle is below a predetermined threshold.

[0028] Thus, the method can be implemented automatically at night, i.e. when the outside brightness is below the predetermined threshold.

[0029] A second aspect of the invention relates to a computer program comprising instructions for implementing the method, when these instructions are executed by a processor.

[0030] A third aspect of the invention relates to an image generation device for a vehicle, comprising:- a first interface capable of obtaining at least one first image representative of a scene facing the vehicle;- a second interface capable of obtaining at least one additional information relating to the scene facing the vehicle;- a processor capable of determining at least one second image as a function of said at least one first image and as a function of said at least one additional information, by applying an image generation model stored in the vehicle, said second image having a luminance greater than said first image;- a third interface capable of transmitting said at least one second image to a module in charge of a function of the vehicle, for implementation of said function at least as a function of said at least one second image.

[0031] Other characteristics and advantages of the invention will appear on examining the detailed description below, and the appended drawings in which:

[0032] illustrates a motor vehicle according to embodiments of the invention;

[0033] illustrates the steps of a common phase of an image generation method according to embodiments of the invention;

[0034] illustrates the steps of a preliminary phase of an image generation method according to embodiments of the invention;

[0035] illustrates the structure of a generative adversarial network for training an image generation model according to one embodiment of the invention;

[0036] illustrates an image generation device according to embodiments of the invention.

[0037] The description focuses on the characteristics that distinguish the method and device from those known in the state of the art.

[0038] Illustrates a vehicle 100 according to embodiments of the invention.

[0039] The vehicle 100 comprises an image generation device 101 according to embodiments of the invention, the device 101 being capable of receiving a first image from an image acquisition device 102 of the vehicle, as well as additional information, and of generating a second image corresponding to the first image and having a luminance greater than the first image. The first and second images are representative of a scene facing the vehicle 100.

[0040] The luminance of an image corresponds to the ratio between the light intensity and the surface area of ​​the image.

[0041] The luminance of the image can be determined from the individual luminance values ​​of the image pixels. The luminance of the image can, for example, be an average value of the individual luminance values ​​of the image pixels.

[0042] So, a higher luminance indicates that, on average, pixels are brighter (have higher luminance values) in the second image than in the first image.

[0043] The image acquisition device 102 is capable of acquiring the first images representative of a scene facing the vehicle, that is to say in front of the vehicle in a direction of travel of the vehicle. In particular, the image acquisition device 102 may be located at the front of the vehicle, for example at the top of the windshield of the vehicle 100 or at the bumper of the vehicle 100.

[0044] The first images can be color images, coded for example in red-green-blue, or RGB in English for “Red-Green-Blue”, that is to say that the first images include colored pixels arranged in a matrix, and that each pixel is coded so as to correspond to one color among strictly more than two colors. For example, each pixel can be coded in red-green-blue, or RGB for “Red Green Blue”, that is to say that the set of colors is represented by three coordinates respectively red, green and blue. Each pixel can thus be coded on several bits, for example on a byte coding 256 colors.

[0045] Alternatively, the first images acquired by the image acquisition device 102 may be grayscale images, i.e. each first image comprises pixels, arranged in a matrix, each pixel being coded on strictly more than two levels, with at least one intermediate grayscale between white and black. Each pixel may thus be coded on several bits, for example on a byte coding 256 grayscale levels.

[0046] According to the invention, the first image is acquired in so-called “night” driving situations. Such a driving situation can, for example, be detected by a vehicle sensor when the light intensity outside the vehicle is below a given, predetermined threshold, the given threshold being, for example, between 800 and 1200 Candelas, in particular equal to 1000 Candelas.

[0047] The second image generated by the image generation device 101 thus has a luminance greater than the first image, and preferably a luminance such that the second image corresponds to a daytime driving situation, in which it is easier for the driver or for a function of the vehicle to identify the objects in the scene facing the vehicle.

[0048] Preferably, the first image has a luminance corresponding to a first light intensity value lower than the given threshold and the second generated image has a luminance corresponding to a second light intensity value higher than the given threshold.

[0049] The vehicle 100 may further comprise a human machine interface, or HMI, 106 capable of exchanging information with the driver and with any passengers of the vehicle 100.

[0050] The HMI 106 may in particular comprise at least one screen, capable of displaying a fixed image or a series of images received from another module of the vehicle. In particular, the HMI 106 is capable of displaying the second image generated by the image generation device 101.

[0051] Preferably, the image generation device 101 receives a series of first images, or first images continuously and in real time from the image acquisition device 102, and thus transmits a series of second images in real time to another module of the vehicle, for example to the HMI 106 for real-time display of the second images. The driver can thus consult on the HMI 106 the second images representative of the scene facing the vehicle and having a light intensity greater than the actual scene facing him. It is thus made possible to assist the driver during night driving situations, the driver being able to consult the HMI in order to ensure the presence or absence of an object in the scene facing the vehicle 100.

[0052] The screen may also be a touch screen, making it easier for the driver to input tactile inputs, in correlation with the image displayed on the touch screen. The touch screen may in particular be of the resistive or capacitive type.

[0053] The vehicle may further comprise at least one additional sensor 103 capable of acquiring data representative of the environment of the vehicle, in particular of the scene facing the vehicle. The additional sensor 103 may be:- a radar or lidar directed towards the front of the vehicle 100;- a thermal or infrared camera, capable of obtaining an image, or a series of images, representative of the infrared radiation of objects located in a scene facing the vehicle 100. Preferably, the thermal camera is oriented according to the orientation of the image acquisition device 102, so that their respective fields of vision overlap. The image acquisition device 102 and the thermal camera thus acquire images of the same scene;- a luminance camera, capable of obtaining a luminance image, also called a luminance map, or a series of luminance images.As with the thermal camera above, the luminance camera is oriented so that its field of view overlaps with the field of view of the image acquisition device 102.

[0054] Luminance is a quantity that designates the luminous flux coming from an illuminated surface and which is reflected in the eye, which is expressed in candela per square meter, or cd / m2. Luminance thus translates a visual sensation of brightness of a surface.

[0055] More precisely, luminance is the power of visible light passing through or being emitted by a surface element in a given direction, per unit area and per unit solid angle.

[0056] In outdoor lighting, the measurement of luminance is normalized according to the distance from the surface, the position and height of the observer, the angle of observation, etc.

[0057] Thus, the luminance map is distinct from a grayscale image, in which the grayscale of a pixel illustrates the brightness perceived by a camera sensor at a given location, but which does not correspond to luminance as defined above.

[0058] According to embodiments, the vehicle 100 comprises several additional sensors 103, in particular any combination of the aforementioned additional sensors. The vehicle can thus comprise:- a lidar / radar and a luminance camera;- a lidar / radar and a thermal camera;- a luminance camera and a thermal camera; or- a lidar / radar, a thermal camera and a luminance camera.

[0059] Optionally, the vehicle may further include a driving assistance module 105, also called ADAS, for “Advanced Driver Assistance System” in English.

[0060] The ADAS module 105 is capable of implementing at least one driving assistance function. No restriction is attached to said at least one function, which depends in particular on the level of autonomy of the vehicle.

[0061] At least one of the functions implemented by the ADAS module 105 may receive as input at least one data item relating to an object detected in a scene facing the vehicle. This is the case in particular for functions for automatically regulating the longitudinal speed of the vehicle, which may cause sudden braking upon detection of a fixed object on the road. In the case where such a function is implemented by the ADAS module 105, the image generation device 101 may transmit the second generated image, or the series of second images, to the ADAS module 105, which may process said second image to detect objects in the scene facing the vehicle. Such detection is carried out by the ADAS module 105 with greater certainty in the second images compared to a detection which would be implemented in the first images, due to the greater luminance of the second images.The driving assistance function can thus be carried out with greater precision, which avoids untimely braking and increases the safety associated with driving the vehicle 100.

[0062] The vehicle 100 further comprises a pair of lighting devices each comprising at least one lighting module 104 capable of performing at least one lighting function of the vehicle 100.

[0063] In the example of the, the vehicle 100 is shown in profile so that a single lighting module 104 integrated in the left lighting device is visible in the figure. It will be understood, however, that the vehicle 100 further comprises a lighting module located to the right of the vehicle, arranged symmetrically to the lighting module 103, relative to a front-rear axis of the vehicle.

[0064] The lighting module 104 may be dedicated to a first lighting function, such as a dipped beam function, also called LB for “Low Beam” in English. Alternatively, the lighting module is capable of implementing two lighting functions, such as the LB function and a high beam function, also called HB for “High Beam” in English.

[0065] The vehicle 100 further comprises a control module for the lighting module 104, not shown in the, and capable of activating / deactivating the first function and / or the second function of the vehicle. When the light sources of the lighting module 104 are of the matrix type with a plurality of individually controllable light elements, the control module can further transmit pixelated lighting photometry, from which the lighting device is capable of selectively activating / deactivating light elements of the source of the lighting module 104.

[0066] The control module may be dedicated to controlling the lighting devices or may further provide control of other vehicle functions. The control module may in particular be a central control module of the vehicle 100.

[0067] It is thus made possible to achieve adaptive lighting, i.e. depending on the elements of the scene facing the vehicle 100.

[0068] The adaptive lighting may be based on the first images acquired by the image acquisition device 102. However, the elements of the scene facing the vehicle 100 may be subject to detection errors in the first images, as for the human eye. The invention thus advantageously makes it possible to use the second images generated to allow the control module to detect with greater precision and greater certainty the elements of the scene facing the vehicle, thus improving the adaptation of the lighting of the vehicle.

[0069] The image generation model may be stored in a vehicle memory not shown in the. The memory may be integrated into the image generation device 101.

[0070] This is a diagram illustrating the steps of a common phase of an image generation method according to embodiments of the invention.

[0071] The current phase corresponds to a phase during which the vehicle 100 is in circulation, in particular at night, that is to say when the light intensity outside the vehicle is lower than the predetermined threshold described previously.

[0072] We thus distinguish the current phase from a preliminary phase during which the image generation model is constructed, the preliminary phase being described in the following with reference to the.

[0073] The steps of the current phase can be implemented by the image generation device 101 described previously.

[0074] The current phase comprises a step 201 of obtaining at least one first color or grayscale image representative of a scene facing the vehicle 100. A series of first images can be obtained in step 201.

[0075] The first image may come directly from the image acquisition device 102 of the vehicle 100, or may come from image processing applied to an image captured by the image acquisition device 102.

[0076] The current phase further comprises a step 202 of obtaining additional information, representative of the scene facing the vehicle 100.

[0077] There are no restrictions on the additional information, its format, the manner in which it is obtained, or the module of the vehicle 100 capable of acquiring the additional information. The adjective “additional” indicates that the additional information is additional to the information contained in the first image. In other words, the first image does not include the additional information.

[0078] In a first embodiment, the additional information may be a thermal image representative of the scene facing the vehicle 100. The thermal image may be acquired by the additional sensor 103, or by one of the additional sensors 103, which may be a thermal camera as described previously.

[0079] Such additional information is advantageous in that it makes it possible to increase the level of certainty associated with the detection of elements of the scene facing the vehicle, in particular animated, living or moving elements, such as animals, other vehicles, pedestrians or cyclists, etc.

[0080] In a second embodiment, the additional information may comprise at least one motion vector associated with the first image obtained in step 201. The additional information may in particular comprise a map of motion vectors, each motion vector of the map corresponding to an area of ​​at least one pixel of the first image, each area preferably comprising several pixels. In the second embodiment, the additional information thus indicates moving or stationary areas in the scene facing the vehicle 100, which makes it possible to improve the precision associated with the detection of elements in the scene facing the vehicle 100.

[0081] In a third embodiment, the additional information may be a luminance image. The luminance image may be acquired by the additional sensor 103, when it is a luminance camera, or acquired by one of the additional sensors 103. Alternatively, the luminance image is obtained by a calculation module from the first image from the image acquisition device 102, by applying a luminance map generation model. It should be noted that the luminance map generation model is distinct from the image generation model capable of generating a second image from a first image and at least one additional information.

[0082] Such a luminance map generation model encompasses any model capable of receiving a color or black and white image as input, and obtaining a luminance image as output.

[0083] The luminance map generation model can for example be an artificial neural network and can be obtained by machine learning.

[0084] For this purpose, the luminance map generation model may be previously trained by supervised learning from pairs of training data from a training database, each pair comprising a color or black and white image, and an associated luminance image (or associated luminance map). The luminance map generation model is thus trained so as to reduce the error between the luminance image predicted or generated from a training image, and the luminance image associated with the training image in the training database. The training database used for training the luminance map generation model may be distinct from the training database described below and used for training the image generation model according to the invention.

[0085] According to a fourth embodiment, the additional sensor 103, or one of the additional sensors 103, is a lidar or a radar, and the additional information is information captured by the lidar or the radar. The information captured by the radar or the lidar may for example be the presence of an element in the scene facing the vehicle, such as another vehicle for example.

[0086] In another embodiment, several additional information from among the additional information associated with the above four embodiments are obtained at step 202.

[0087] In a step 203, the image generation device 101 generates a second image from said at least one first image obtained in step 201 and from said at least one additional information obtained in step 202. For this purpose, the image generation device 101 accesses the image generation model stored in the vehicle, said image generation model being capable of:- receiving as input a color or black and white image, of the same nature and format as the first image obtained in step 201, and information of the same format and nature as the additional information obtained in step 202;- and generating as output an image having a higher luminance than the image received as input.

[0088] The image generation model can be an artificial neural network, such as a recurrent neural network for example.

[0089] The image generation module thus depends on the implementation method considered, namely the format and nature of the additional information.

[0090] Following step 203, the second generated image can be transmitted:- to the HMI 106 at a step 204, for display of the second image on the screen of the HMI 106. The user is thus allowed to consult the second image which is brighter than the scene facing him, which improves the driving conditions and reinforces the safety associated with driving. The HMI 106 can be considered as a module of the vehicle implementing a display function; and / or- to at least one module capable of implementing at least one function of the vehicle 100 other than the display function, at a step 205, for taking into account the second image in the implementation of the function. The module receiving the second image can be the ADAS module 105 for identification of elements in the scene facing the vehicle 100, and taking into account the second image in the implementation of at least one driving assistance function.Alternatively or additionally, the module receiving the second image may be the control module of the lighting device, which is thus able to take the second image into account for adaptive control of the lighting, in particular for adapting the photometry transmitted to the lighting device.

[0091] Alternatively, the image generation device 101 is capable of detecting the elements of the scene in the second image, before transmission to the HMI 106 and / or to the module capable of implementing the function of the vehicle 100.

[0092] Steps 201 to 205 have been described as implemented by the image generation device previously described. Alternatively, the image generation device 101 may be replaced by a module, hardware or software, having the same functionalities as those described previously, and implemented in another hardware entity of the vehicle 100, for example in the ADAS module 105 or in the image acquisition device 102.

[0093] Furthermore, steps 201 to 205 of the current phase may be repeated, so as to generate a series of second images, transmitted to successive steps 204 and / or 205. In the case where the current phase is iterated, the frequency of obtaining the first image in step 201 may differ from the frequency of obtaining the additional information in step 202. For example, the additional information is obtained only once for several first images obtained at several iterations of step 201.

[0094] This is a diagram illustrating the steps of a preliminary phase of an image generation method according to embodiments of the invention.

[0095] The preliminary phase makes it possible in particular to obtain, or construct, the image generation model stored in the vehicle for the implementation of step 203 of the current phase described previously.

[0096] The preliminary phase is described in the case of a construction of the image generation module by generative adversarial network, also called GAN for “Generative Adversarial Network” in English, and the principle of which is illustrated with reference to the. According to a variant not described, the image generation module is not constructed from a GAN, but by supervised learning from training data comprising pairs of training images, each pair comprising a night image and a day image of the same scene as the night image, as well as additional training information associated with the pairs of images. However, in practice, such pairs of images are complex to obtain, and construction by GAN makes it possible to overcome these difficulties.

[0097] In a step 301, a training database is obtained, the training database comprising training data for training the image generation model. No restriction is attached to the manner in which the training data according to the invention are obtained. They may in particular be collected on a fleet of vehicles comprising respective sensors and image acquisition devices, or may be synthesized or simulated.

[0098] In the context of a construction of the image generation model by GAN, the training data comprises:- associations between training night images and additional training information, each association comprising a night image of a driving scene and at least one additional information on the same driving scene as the night image. Each training night image can be acquired by a camera in a vehicle of a fleet of vehicles, similar or even identical to the image acquisition device 102 previously described. The additional training information can be captured in a coordinated manner with the training night images, by an entity of the vehicle of the fleet of vehicles, in the same way as the additional information is obtained in step 202 previously described (which depends on the embodiment considered);- training day images, representing daytime driving scenes.Such training day images can be acquired during daytime driving situations.

[0099] The training day images and the training night images can be acquired in various driving situations, which makes it possible to improve the richness of the training database and therefore the robustness of the image generation model, which is thus accurate in various driving situations. Furthermore, the training day images and the training night images can be acquired by cameras of several types, or on the contrary, of a single type corresponding to the same type as the image acquisition device 102 previously described and integrated in the vehicle 100 according to the invention. “Camera type” means a category defined on the basis of technical criteria, such as a sensor resolution, a sensor format, a focal length and / or any other well-known technical characteristic.

[0100] In a step 302, a training cycle is implemented to train the image generation model by machine learning, from the training data obtained in step 301.

[0101] At each training cycle, the preliminary phase comprises a step 303 of evaluating a performance criterion, which makes it possible to determine whether the image generation model resulting from the current training cycle satisfies the performance criterion.

[0102] If this is the case, that is to say if the performance criterion is satisfied by the image generation model at the end of the current training cycle, the method moves on to a step 304 of storing the image generation model resulting from the current training cycle, in a memory of the vehicle 100.

[0103] Otherwise, i.e. if the performance criterion is not satisfied by the image generation model at the end of the current training cycle, the method returns to step 302 for a new training cycle, thus making it possible to continue training the image generation model until the performance criterion is satisfied.

[0104] Illustrates the structure of a generative adversarial network 400 capable of constructing the image generation model according to embodiments of the invention.

[0105] The generative adversarial network 400, or GAN 400, comprises a generation module 401 and a discrimination module 402.

[0106] The generation module 401 is configured to receive, during a training cycle, a series of training images 403 corresponding to night scenes facing a vehicle, as well as a series of associated additional training information 404. For each training image received and each additional training information, the generation module 401 generates a second image, brighter than the received training night image.

[0107] The 401 generation module implements the image generation model being built. The parameters of the image generation model are optimized as training progresses. In particular, the parameters are recalculated at the end of each training cycle.

[0108] Thus, several second images are generated during a training cycle by the generation module 401.

[0109] The second generated images as well as one or more training day images 405 are submitted during the training cycle to the discrimination module 402. Preferably, each second generated image is submitted in association with a training day image, so that the discrimination module 402 determines which of the received images is a real image, and which is a generated (or synthetic) image.

[0110] For this purpose, the discrimination module 402 is able to implement, or implement, a classification model configured to determine, for each pair of images received, a score for each image of the pair of images received. The highest score is assigned to the image of the pair which is classified by the discrimination module 402 as being a real image.

[0111] For each series of image pairs, a GAN evaluation module 400, not shown in the figure, determines the number of errors made by the discrimination module 402. The higher the number of errors, the more efficient the generation module 401 is, since the second generated image is then interpreted as a real image.

[0112] At least some of the GAN 400 parameters are redefined at each training cycle based on an output from the evaluation module, the output depending on the number of errors made by the discrimination module

[0113] The generation module 401 is thus trained in such a way that the discrimination module 402 is unable to identify the real image among the second image and the training day image. The performance criterion can be defined as a condition for stopping the training process. For example, the performance criterion can be achieved when the discrimination module 402 assigns the same score to the second image and the training day image, for all pairs of images in a training cycle. Other performance criteria can be implemented in the GAN 400.

[0114] The present invention shows the structure of the image generation device 101 according to embodiments of the invention.

[0115] The image generation device 101 comprises a processor 501 configured to communicate unidirectionally or bidirectionally, via one or more buses or via a direct wired connection, with a memory 502 such as a “Random Access Memory” type memory, RAM, or a “Read Only Memory” type memory, ROM, or any other type of memory (Flash, EEPROM, etc.). Alternatively, the memory 402 comprises several memories of the aforementioned types.

[0116] The memory 502 is capable of storing, permanently or temporarily, at least some of the data used and / or resulting from the implementation of the steps of the method according to the invention illustrated with reference to the.

[0117] In particular, the memory 502 may be able to store the image generation model during step 304, for execution during the current phase. However, such storage is optional, the image generation model being able to be stored in another entity of the vehicle 100.

[0118] The processor 501 is capable of executing instructions, stored in the memory 502, for implementing steps 201 to 205 of the current phase, described with reference to the. Alternatively, the processor 501 can be replaced by a microcontroller designed and configured to carry out steps 201 to 205 of the current phase, described with reference to the.

[0119] The image generation device 101 comprises a first interface 503 capable of receiving the first images during the step 201 described previously, either directly from the image acquisition device 102, or from a module responsible for processing the images acquired by the image acquisition device 102 and for obtaining the first images.

[0120] The image generation device 101 comprises a second interface 504 capable of receiving the additional information, obtained in step 202 described previously, from another module of the vehicle depending on the embodiment.

[0121] The image generation device 101 comprises a third interface 505 capable of transmitting each second generated image. For example, the third image may be transmitted to the ADAS module 105, to the HMI 106, to the lighting module 104 or to the lighting module control module 104 described above.

[0122] The present invention is not limited to the embodiments described above as examples; it extends to other variants.

Claims

Image generation method, implemented in a motor vehicle (100), comprising a current phase comprising the following steps:- obtaining (201) at least one first image representative of a scene facing the vehicle;- obtaining (202) at least one additional information relating to the scene facing the vehicle;- determining (203) at least one second image as a function of said at least one first image and as a function of said at least one additional information by the application of an image generation model stored in the vehicle, said second image having a luminance greater than said first image;- transmitting (204; 205) said at least one second image to a module in charge of a function of the vehicle, for implementation of said function at least as a function of said at least one second image. Method according to claim 1, wherein said at least one first image comes from an image acquisition device (102) capable of obtaining images of the scene facing the vehicle (100). Method according to claim 1 or 2, wherein said at least one second image is transmitted:- to a driving assistance module (105) capable of implementing at least one driving assistance function of the vehicle (100);- to a lighting module (104) responsible for implementing a lighting function;- to a control module of a lighting module capable of determining a photometry to be transmitted to the lighting module to perform a lighting function; and / or- to a human-machine interface (106) of the vehicle, comprising at least one screen, for displaying said at least one second image on said at least one screen. Method according to one of the preceding claims, in which the image generation model is an artificial neural network. Method according to one of the preceding claims, further comprising a preliminary phase of constructing said image generation model comprising training the image generation model from training data. The method of claim 5, wherein the image generation model is constructed by a generative adversarial network (400) comprising a generation module (401) and a discrimination module (402), the model being implemented by the generation module, wherein the training data comprises:- training night images acquired in night driving situations and representative of driving scenes in front of vehicles, and associated additional training information, and- training day images acquired in day driving situations. Method according to one of the preceding claims, wherein said at least one additional information comprises a thermal image of the scene facing the vehicle (100). Method according to one of the preceding claims, wherein said at least one additional information comprises at least one motion vector associated with the first image. The method of claim 8, wherein said at least one additional information comprises a set of motion vectors, each motion vector being associated with an area of ​​the first image. Method according to one of the preceding claims, in which said at least one additional information comprises a luminance image representative of the scene facing the vehicle. Method according to one of the preceding claims, in which the first image is acquired when a brightness outside the vehicle (100) is lower than a predetermined threshold. Computer program comprising instructions for implementing the method according to one of the preceding claims, when these instructions are executed by a processor (501). Image generation device (101) for a vehicle (100), comprising:- a first interface (503) capable of obtaining at least one first image representative of a scene facing the vehicle;- a second interface (504) capable of obtaining at least one additional information relating to the scene facing the vehicle;- a processor (501) capable of determining at least one second image as a function of said at least one first image and as a function of said at least one additional information, by applying an image generation model stored in the vehicle, said second image having a luminance greater than said first image;- a third interface (505) capable of transmitting said at least one second image to a module in charge of a function of the vehicle, for implementation of said function at least as a function of said at least one second image.

Images