Method for reducing noise in a camera image of a vehicle camera of a vehicle, control device, vehicle and computer program

Abstract

The invention relates to a method for reducing noise in a camera image of a vehicle camera (3) of a vehicle (1), comprising the steps of: - providing the camera image and a reference image which was generated at a preceding time, - calculating a difference image between the camera image and the reference image, wherein the pixels of the difference image each contain at least one difference value according to a difference between the pixel in the camera image and the pixel in the reference image in at least one color channel, - generating an alpha mask from the difference image by comparing the difference values of the difference image with a limit value that is inversely proportional to a speed of the vehicle (1), the alpha mask containing, for the pixels having at least one difference value above the limit value, an alpha value corresponding to complete transparency and, for the other pixels, an alpha value which is dependent on the difference value and which describes at least partial transparency, - generating a correction image by applying the alpha mask to the reference image, - generating a noise-reduced camera image from the correction image and the camera image.

Description

[0001] Description

[0002] Method for noise reduction in a camera image from a vehicle camera, control unit, vehicle and computer program

[0003] The invention relates to a method for noise reduction in a camera image from a vehicle camera. The invention further relates to a control unit, a vehicle, and a computer program.

[0004] Modern vehicles typically feature driver assistance systems that allow the driver to see areas of the vehicle's surroundings that are difficult to see or obscure from the driver's seat. These systems often utilize cameras mounted on the vehicle and pointed towards its surroundings, with the images displayed on a monitor inside the vehicle. The camera images can show, for example, an area directly in front of, beside, and / or behind the vehicle. To display larger areas of the vehicle's surroundings and / or to provide views from a virtual vantage point, it is common practice to combine multiple, possibly transformed, camera images to create a comprehensive view of the vehicle's environment.

[0005] For example, DE 10 2016 225 066 A1 describes a surround-view system for a vehicle. The surround-view system comprises at least one camera for capturing an image of the vehicle's surroundings, an image processing unit, and a display unit. The display unit is configured to show a projected representation of the image captured by the real camera.

[0006] The cameras, which can be wide-angle cameras for example, typically have an image sensor that detects electromagnetic radiation or light and captures it as an image. The image quality generally depends on the lighting conditions of the area of ​​the vehicle's surroundings being recorded. Particularly in poor lighting conditions, such as darkness, image noise can occur, which can be perceived as distracting when viewing the camera images.

[0007] The invention is therefore based on the objective of providing a method that reduces image noise in a camera image from a vehicle camera.

[0008] To solve this problem, a method of the type mentioned at the outset is provided according to the invention to comprise the following steps:

[0009] - Providing the camera image and a reference image generated at a previous time,

[0010] - Calculating a difference image between the camera image and the reference image, wherein the pixels of the difference image each contain at least one difference value as a function of a difference between the pixel in the camera image and the pixel in the reference image in at least one color channel,

[0011] - Generating an alpha mask from the difference image by comparing the difference values ​​of the difference image with a limit value inversely proportional to the speed of the vehicle, wherein the alpha mask contains an alpha value corresponding to complete transparency for the pixels with at least one difference value above the limit value and an alpha value for the other pixels that depends on the difference value and describes at least partial transparency,

[0012] - Generating a correction image by applying the alpha mask to the reference image,

[0013] - Generating a noise-reduced camera image from the correction image and the camera image.

[0014] The camera image, in which image noise is reduced, originates from a vehicle camera. This vehicle camera can be, in particular, a surround-view camera that captures at least a portion of the vehicle's surroundings, with the corresponding portion of the vehicle's environment then being depicted in the camera image. For example, the vehicle camera can be a front camera capturing the area in front of the vehicle, a side camera capturing the area beside the vehicle, or a rear camera capturing the area behind the vehicle. Other camera configurations on the vehicle are also possible. The camera image can be displayed on a vehicle display device. In this case, the camera image can be displayed directly.Additionally or alternatively, an environment view generated using at least part of the camera image can also be displayed.

[0015] To perform noise reduction in one of the camera images, a camera image and a reference image generated at a previous, earlier time are provided in a first step of the process, which is particularly computer-implemented. The camera image can be, in particular, a currently captured camera image, especially an image from a continuously generated image stream or video stream. The reference image can be a noise-reduced camera image generated in a previous iteration of the process. Alternatively, a camera image captured at the previous time in which no noise reduction was performed can be used, especially if no noise-reduced camera image is yet available, for example, at the start of the process.

[0016] A difference image is then generated between the camera image and the reference image. The camera image and the reference image have the same resolution. The difference image specifies, for each pixel, at least one difference value depending on the difference in at least one color channel between the pixel in the camera image and the pixel in the reference image. Specifically, a difference value can be determined for each color channel. In other words, the difference image indicates, for each pixel, the amount by which the values ​​in the at least one color channel of the respective pixel have changed. An alpha mask is then generated from the difference image by comparing the pixels of the difference image, each containing the difference value, with a threshold value that is inversely proportional to the vehicle's speed.For each pixel, the difference value is compared to a limit value, where the value of the limit value is inversely proportional to the vehicle's current speed. Inversely proportional means that the limit value decreases as the vehicle speed increases and increases as the vehicle speed decreases. A stationary vehicle can also be assigned a speed, particularly 0, and a corresponding limit value. The change in the limit value depending on the speed can be continuous or discrete.

[0017] The alpha mask determined in this way contains an alpha value corresponding to complete transparency (for example, alpha value 0) for pixels with one or more difference values ​​above the threshold, and an alpha value dependent on the difference values ​​for pixels with difference values ​​below the threshold. This alpha value can be greater than 0 and less than or equal to 1. In other words, pixels with at least one difference value above the threshold are not assigned transparency, while pixels with a difference value below the threshold are assigned an alpha value dependent on the difference value.Pixels in which a large change has occurred between the difference image and the camera image are therefore not assigned any transparency in the alpha mask, whereas pixels in which only a small change, defined by the threshold value, has occurred between the difference image and the camera image are assigned at least partial transparency.

[0018] Using the alpha mask, a correction image is then generated by applying the alpha mask to the reference image. This is done, for example, by multiplying the respective alpha values ​​of the pixels in the reference image, which correspond to complete opacity and are typically 1, by the alpha values ​​assigned to each pixel from the alpha mask. In other words, each pixel in the reference image can be assigned the alpha value described by the alpha mask.

[0019] The noise-reduced camera image is then generated from the correction image and the camera image, in particular by superimposing the correction image with the camera image or an intermediate image generated depending on the camera image.

[0020] In a further step of the process, the noise-reduced camera image and / or a derived view of the surroundings can be displayed on a screen in the vehicle. Additionally or alternatively, the image content of the noise-reduced camera image can also be analyzed by a driver assistance function of the vehicle.

[0021] By using a threshold value dependent on the vehicle speed, noise reduction only occurs in image areas where there is no excessively large difference between the camera image and the reference image. For example, in image areas containing one or more moving edges between the reference image and the camera image, no noise reduction takes place. This advantageously prevents edge blurring and thus a reduction in the image quality of the camera image. Conversely, in image areas with essentially homogeneous surfaces, where even when the vehicle is moving there are no significant differences between the camera image and the reference image, noise reduction occurs, as image noise in these areas can be particularly disruptive when displaying the camera image.

[0022] The invention thus advantageously improves the image quality of the camera image, particularly when it is displayed in the vehicle and / or subsequently processed by a driver assistance system, especially under poor lighting conditions in a part of the vehicle environment recorded by the vehicle camera.

[0023] According to the invention, the difference value can be the magnitude of the difference between the color channel of the pixel in the camera image and the color channel of the pixel in the reference image, or the difference value can be proportional to the magnitude of the difference between the color channel of the pixel in the camera image and the color channel of the pixel in the reference image. In other words, the difference value assigned to a pixel can thus be the (unsigned) magnitude of the difference between an amplitude in a color channel of the camera image and the amplitude of the corresponding color channel in the reference image, or it can depend proportionally on this magnitude, for example by multiplication by a factor or the like.

[0024] In a preferred embodiment, the limit value can also be proportionally dependent on a noise figure, which describes the image noise in the camera image. The noise figure is higher the greater or more pronounced the image noise in the camera image. Thus, in camera images with high noise levels, noise reduction is applied to larger image areas than in camera images with low noise levels.

[0025] According to the invention, the noise figure can be determined by the vehicle camera during image acquisition and / or based on a histogram of previously determined difference images. A noise figure determined by the vehicle cameras can, for example, be measured directly by the camera during image acquisition and transmitted to a control unit designed to carry out the method according to the invention. The histogram, which is determined in particular based on several difference images from previous iterations of the method, can, for example, be stored in a memory device of such a control unit and evaluated by the control unit. The histogram can, for example, be a standard deviation of the determined difference values ​​between the camera image and the reference image for the individual pixels.

[0026] In a preferred embodiment of the invention, the alpha value, which depends on the difference value, for pixels below the threshold value corresponds to the magnitude of the difference in at least one color channel multiplied by a scaling factor, wherein the scaling factor depends in particular on the speed of the vehicle and / or on a noise measure. In the case of differences in multiple color channels, the scaling factor can be applied to the magnitude of the difference in the color channel with the largest difference, i.e., the difference with the greatest absolute value. Alternatively, the difference across multiple color channels can be considered, for example, by averaging the differences between several color channels, and the average value can then be multiplied by the scaling factor to determine the alpha value.

[0027] Using the scaling factor, the alpha mask can, for example, be normalized to a range of alpha values ​​between 0 (completely transparent) and 1 (completely opaque). This can be achieved, for example, by subdividing the image into 16 bits, 32 bits, or similar. Based on the scaling factor, which depends on the vehicle speed, the influence of the difference image on the resulting noise-reduced camera image can be adjusted depending on the speed and / or a noise figure, without altering the areas of the camera image defined by the limit value where noise reduction takes place.

[0028] According to the invention, the noise-reduced camera image can be generated by adding the correction image to create a semi-transparent camera image, wherein the semi-transparent camera image is generated by modifying the pixels of the camera image depending on the alpha values ​​of the alpha mask. In other words, a semi-transparent camera image can also be generated using the alpha mask by reducing the alpha values ​​of the pixels of the camera image by the alpha values ​​of the alpha mask. The alpha values ​​of the pixels of the camera image can, in particular, initially be 1, i.e., completely opaque, so that by reducing the alpha values, at least some of the pixels are assigned partial transparency. In other words, partial transparency is assigned to those pixels which, in the difference image, exhibit a change below the vehicle speed-dependent threshold.

[0029] The noise-reduced camera image can then be generated by adding the correction image and the semi-transparent camera image. This addition is performed for each pixel, taking into account its respective transparency. The semi-transparent pixels in the semi-transparent camera image and the transparent pixels in the correction image overlap in such a way that their alpha values ​​in the noise-reduced camera image sum to 1. This advantageously preserves the brightness and color saturation of each pixel.

[0030] According to the invention, the noise-reduced camera image can be generated by superimposing the correction image and the semi-transparent camera image using a spatial filter function. In particular, the spatial filter function allows the color values ​​of individual points in the sum of the correction image and the semi-transparent camera image to be adjusted based on the color values ​​of one or more neighboring pixels. The spatial filtering can, for example, be a blurring effect or the like. By additionally applying the spatial filter function, the image quality of the noise-reduced camera image or a derived environmental view can be further improved.

[0031] According to the invention, it can be provided that a kernel of the spatial filter function is set depending on the determined alpha mask and / or that the spatial filtering is carried out depending on sections in the alpha mask that can be assigned to an edge in the camera image. The kernel of the spatial filter function can define which neighboring pixels are used to determine the color value of a pixel. In particular, the kernel can be set depending on the determined alpha mask such that only pixels are processed and / or only pixels are used to process pixels to which a non-zero alpha value is assigned in the alpha mask.

[0032] Additionally or alternatively, the spatial filtering function can also be applied in areas where the difference value of individual pixels has exceeded the limit value used to determine the alpha mask. This enables spatial filtering, for example, of edges moving between the camera image and the reference image.

[0033] According to the invention, the vehicle camera can capture at least a partial area of ​​the vehicle's surroundings, wherein the noise-reduced camera image and / or a representation of the surroundings generated from the noise-reduced camera image are displayed on a display device in the vehicle. The display device can, for example, be a display unit located in the interior of the vehicle, such as a screen, a heads-up display, or the like.

[0034] A control device according to the invention is provided that it is configured to carry out a method according to the invention.

[0035] A vehicle according to the invention is provided to comprise at least one vehicle camera and a control unit according to the invention.

[0036] A computer program according to the invention is provided to include instructions which cause a computing device to execute a method according to the invention.

[0037] The computer program can be stored on a data carrier, particularly a non-transient one, such as a CD-ROM, a hard drive, flash memory, or the like. It is also possible for the computer program according to the invention to be accessible via a communication connection, for example, via the Internet, from a data storage device, such as a server or the like.

[0038] All advantages and embodiments described above in relation to the method according to the invention also apply analogously to the control device, the vehicle according to the invention, and the computer program according to the invention, and vice versa. The advantages and embodiments described in relation to the control device, the vehicle according to the invention, and the computer program according to the invention are also transferable analogously to the other subject matter of the invention.

[0039] Further advantages and details of the invention will become apparent from the exemplary embodiments described below and from the drawings. These are schematic representations and show:

[0040] Fig. 1 shows an embodiment of a vehicle according to the invention.

[0041] Fig. 2 shows a block diagram of an embodiment of a method according to the invention and

[0042] Fig. 3 shows a functional block diagram of an implementation of the exemplary embodiment of the method according to the invention.

[0043] Figure 1 shows an embodiment of a vehicle 1. The vehicle 1 can be, for example, a motor vehicle, in particular a passenger car, a truck, or another type of commercial vehicle. It is also possible that the vehicle 1 is an unmotorized vehicle such as a trailer or that it is a combination of a towing vehicle and a trailer. The vehicle 1 can also be a rail-bound vehicle, such as a tram or the like, or a robot. The vehicle 1 comprises an image acquisition device 2, which includes one or more vehicle cameras 3 and a control unit 4. The control unit can, for example, be implemented as or include a microcontroller, a processor, or the like.

[0044] In the present embodiment, the image acquisition device 2 comprises four vehicle cameras 3, wherein a first camera 5 is configured as a front camera of the vehicle 1, a second camera 6 as a rear-view camera of the vehicle 1, and a third camera 7 and a fourth camera 8 are each configured as a side camera of the vehicle 1. The cameras 3 form, in particular, a surround-view camera system that can capture the entire area around the vehicle 1.

[0045] Each of the vehicle cameras 3 can capture a portion of the vehicle 1's surroundings. The vehicle cameras 3 can, for example, each be a fisheye camera, allowing each camera to have a horizontal field of view of 180° or more. In this way, the entire area around vehicle 1 can be captured using the vehicle cameras 3.

[0046] The vehicle cameras 3 continuously generate image data and transmit it to the control unit 4. For this purpose, the vehicle cameras 3 are connected to the control unit 4 via a communication link, for example, a bus system of the vehicle 1 or similar. The control unit 4 can be configured to generate a surround view from one or more of the camera images from one or more of the vehicle cameras 3. This surround view can be, for example, a front view, a side view, or a rear view. It is also possible for the control unit 4 to generate a surround view, such as a perspective side view or a top view of the vehicle and at least a portion of its surroundings, from the camera images of several of the vehicle cameras 3.The vehicle 1 further includes a speed sensor 9, which determines the speed of the vehicle 1, for example by measuring a component of a drivetrain of the vehicle 1 and / or using satellite data or the like. The speed sensor 9 transmits a speed value describing the current speed of the vehicle 1 to the control unit 4.

[0047] Vehicle 1 includes a display device 10, which can show graphical information to a driver or user of vehicle 1. The display device 10 could, for example, be a screen or similar device located in the interior of vehicle 1. The camera images from one or more of the vehicle cameras 3 can be displayed on the display device 10. The camera images can be processed by the control unit 4, for example, to remove distortions or similar adjustments. Environmental displays generated by the control unit 4 can also be shown on the display device 10.

[0048] To improve image quality, for example in poor lighting conditions in the vehicle environment, the control unit 4 is equipped to carry out a noise reduction procedure in a camera image of one of the vehicle cameras 3 of the vehicle 1.

[0049] Figure 2 shows a flowchart of the noise reduction process in the camera image.

[0050] In a first step S1, a camera image and a reference image generated at a previous time are provided. The reference image can be a camera image taken at the previous time or a noise-reduced camera image determined in a previous iteration of the process. The reference image can, for example, be stored in a memory device of the control unit 4, so that the control unit 4 can access it. The camera image and the reference image can, in particular, have the same resolution, i.e., the same number of pixels in the horizontal and vertical directions. Each pixel is assigned at least one value in a color channel. The alpha values ​​of the individual pixels in the camera image and the reference image are each 1, i.e., the pixels are completely opaque.

[0051] In step S2 of the process, a difference image is then calculated between the camera image and the difference image. The resolution of the difference image corresponds to the resolution of the camera image or the reference image. For each pixel of the difference image, the magnitude of the difference between the value in at least one color channel of the pixel in the camera image and the corresponding color channel of the corresponding pixel in the reference image is determined. In other words, for each pixel of the camera image, the magnitude of the difference between the corresponding value in the color channel of the pixel in the camera image and the value in the color channel of the respective corresponding pixel in the reference image is determined and assigned as a value to the corresponding pixel in the difference image. The difference can, in particular, be the mean value of a respective difference between a plurality of color channels, for example, RGB channels or YUV channels.As an alternative to the amount of the difference, a value proportional to the amount of the difference, a value determined from the difference or the amount of the difference using a linear or non-linear assignment rule, or the like, can also be used as the difference value.

[0052] In step S3, an alpha mask is generated from the difference image by comparing the difference value of each pixel in the difference image with a limit value that is inversely proportional to the speed of vehicle 1. The resolution of the alpha mask also corresponds to the resolution of the camera image and the reference image, and thus also to the resolution of the difference image.

[0053] The alpha mask specifies the pixels whose assigned value is...

[0054] If the difference in the color channel between the values ​​of the pixel in the camera image and the reference image is above the threshold, an alpha value of 0 is assigned, which corresponds to complete transparency. For pixels whose assigned value representing the difference in the color channel between the pixel in the camera image and the corresponding pixel in the reference image is below the threshold, the alpha mask contains an alpha value that depends on the respective difference value or the value of the difference. This alpha value is, in particular, greater than 0 and less than 1, and thus describes partial transparency of the respective pixel.

[0055] The threshold value used to compare the respective differences between the values ​​of the individual pixels in the color channels is inversely proportional to the speed of vehicle 1 as determined, for example, by speed sensor 9. In other words, the threshold value decreases as vehicle 1 moves faster.

[0056] It is possible that the limit value also depends proportionally on a noise figure, which describes image noise in the camera image. The dependency is such that the limit value is higher with high noise and lower with low noise. The noise figure can, for example, be determined by the camera 3 generating the camera image and transmitted to the control unit 4. Alternatively, the noise figure can also be determined by the control unit 4 from a camera image transmitted by camera 3. The noise figure can be determined, for example, using a histogram of generated difference images, where the histogram contains, for example, for a plurality of previously generated difference images, a mathematical distribution of the contained difference values ​​or the contained magnitudes of the differences between the values ​​in the color channels of the individual pixels in the camera image and the reference image.

[0057] The alpha value, which depends on the difference value or magnitude and is assigned to pixels below the threshold, can be determined from the magnitude or absolute value of the difference using a fixed or variable scaling factor. A variable scaling factor can also be inversely proportional to the vehicle speed and / or proportional to the noise figure. For difference values ​​in multiple color channels, the scaling factor can be applied to the magnitude of the difference in the color channel with the greatest difference, i.e., the difference with the largest absolute value. Alternatively, the difference across multiple color channels can be considered, for example, by averaging the differences between several color channels, with the average then being multiplied by the scaling factor to determine the alpha value.

[0058] In step S4, a correction image is generated by applying the alpha mask to the reference image. Each pixel of the reference image is assigned the alpha value specified in the alpha mask for that pixel. This is done, for example, by multiplying the alpha values ​​of the reference image, which are all 1, by the alpha values ​​specified in the alpha mask, which are between 0 and 1. Since pixels whose difference exceeds the limit were assigned an alpha value of 0 in the alpha mask, the alpha value of these pixels in the correction image is also 0, meaning the underlying pixels of the reference image are completely transparent. Pixels that were previously assigned a non-zero alpha value also contain a non-zero alpha value in the correction image, which describes complete or partial transparency.

[0059] In step S5, a semi-transparent camera image is then generated by modifying the alpha values ​​of the camera image pixels, which represent complete opacity, by the alpha values ​​of the alpha mask. The pixels of the semi-transparent camera image, which also corresponds to the camera image or the reference image in resolution, thus encompass the values ​​of the camera image in the color channel, whereby at least some of the pixels are assigned at least partial transparency depending on the alpha mask. This can be achieved, in particular, by subtracting the alpha value of the corresponding pixel in the alpha mask from each of the camera image alpha values, which are each 1.

[0060] When generating the semi-transparent camera image, the sum of the correction image and the semi-transparent camera image can optionally be processed using a spatial filter function. This spatial filter function assigns a modified color value to at least some of the pixels in the semi-transparent camera image being generated. This modified color value depends, at least partially, on the color values ​​of surrounding pixels in the sum of the correction image and the semi-transparent camera image. It is possible for the kernel of the spatial filter function to be set based on the determined alpha mask, and / or for the spatial filtering to be performed based on sections in the alpha mask that correspond to an edge in the camera image.The kernel of the spatial filter function can, for example, be set depending on the determined alpha mask in such a way that only pixels are processed and / or only pixels are used for processing pixels to which a non-zero alpha value is assigned in the alpha mask.

[0061] Additionally or alternatively, the spatial filtering function can also be applied only in areas where the difference between individual pixels exceeds the threshold value used to determine the alpha mask. This enables spatial filtering, for example, of edges moving between the camera image and the reference image. For this purpose, a memory location in the control unit 4 can store information about which pixels exceeded the threshold value when transmitting the alpha mask, and the spatial filtering can then be performed based on this stored information.

[0062] In step S6, a noise-reduced camera image is then generated by adding the correction image to the semi-transparent camera image. The color values ​​of the individual pixels are superimposed depending on the transparency or alpha value assigned to each pixel. Due to the prior reduction of the alpha values ​​in the semi-transparent camera image, the image properties for the areas where semi-transparent pixels are superimposed are maintained compared to the areas where nothing changes, thus advantageously resulting in a homogeneous, noise-reduced camera image.

[0063] The noise-reduced camera image and / or an environment representation derived from the noise-reduced camera image can then be displayed on the display unit 10 of the vehicle 1 in step S7. The noise-reduced camera image generated in this way can then be used as a reference image for a new camera image in a further iteration of the process. In this way, continuously acquired camera images, which are provided by the cameras 3, for example as a video stream, can be processed with the noise reduction method, in particular to improve the image quality of a camera image displayed on the display unit 10 or a camera image stream or an environment representation derived from it.

[0064] Figure 3 shows a functional block diagram of an implementation of the exemplary embodiment of the method for noise reduction in the camera image.

[0065] Block B1 corresponds to the camera image I [x,y,t], which was recorded at time t and contains a two-dimensional arrangement of image points or pixels [x,y].

[0066] Block B2 corresponds to the reference image R[x,y,t-1], which is a noise-reduced camera image l'[x,y,t-1] acquired at a previous time t-1 and stored, for example, in a buffer memory of the control unit 4. In block B3, the difference image AbsDiff[x,y] is determined, whose pixels each contain the magnitude of the difference between the color channel of the respective pixel in the camera image l[x,y,t] and the color channel of the respective pixel in the reference image R[x,y,t].

[0067] To generate the alpha mask Alpha_mask[x,y], different alpha values ​​are assigned to the respective pixels in blocks B4 and B5 depending on the respective value of the amount of the difference.

[0068] In block B4, each pixel of the difference image AbsDiff[x,y] is compared to a threshold value T, and the alpha value for those pixels that lie above the threshold value T is set to zero. In block B5, the respective pixels of the difference image AbsDiff[x,y] processed in block B4 are multiplied by a scaling factor s, which only affects pixels with a non-zero alpha value, i.e., pixels with a difference value below the threshold value T. For example, if a difference value exists in multiple color channels, the scaling factor s is multiplied by the largest absolute value of each pixel.

[0069] As described above, the limit value T is inversely proportional to the speed v of vehicle 1 and optionally also proportional to the noise figure r; thus, T~(1 / v) and optionally additionally T~r. The scaling factor s can either be fixed or also inversely proportional to the speed v of vehicle 1 and / or proportional to the noise figure r.

[0070] In block B6, the reference image R is multiplied by the alpha mask Alpha_mask[x,y] to generate a correction image C[x,y]. The alpha value of each pixel in the reference image R is multiplied by its corresponding alpha value in the alpha mask Alpha_mask[x,y]. Since the alpha values ​​of the pixels in the reference image R[x,y,t-1] are all 1, the pixels in the correction image C all have an alpha value that corresponds to the alpha value of the pixels in the alpha mask Alpha_mask[x,y].

[0071] Furthermore, in block B7, a semi-transparent camera image Z is generated by multiplying the camera image l[x,y] by a factor (1-alpha_mask), which reduces the alpha values ​​of the camera image I [x,y], which describe complete intransparency, depending on the transparency of the individual pixels described by the alpha mask.

[0072] In block B8, the sum (Z[x,y]+C[x,y]) is formed, in which the respective color values ​​of the pixels of the semi-transparent camera image Z[x,y] and the correction image C[x,y] are added depending on the respective assigned alpha value.

[0073] Subsequently, in block B9, a spatial filter function f is applied to the sum (Z[x,y]+C[x,y]) to obtain the noise-reduced camera image l'[x,y,t] in block B10. This image can then be further processed and / or displayed as described above. The spatial filter function f can be applied depending on the alpha mask, as previously described.

[0074] Block B11 describes the provision of the vehicle speed v and the noise figure r.

Claims

Patent claims 1. Method for noise reduction in a camera image from a vehicle camera (3) of a vehicle (1), comprising the steps: - Providing the camera image and a reference image generated at a previous time, - Calculating a difference image between the camera image and the reference image, wherein the pixels of the difference image each contain at least one difference value depending on the difference between the color channel of the pixel in the camera image and the color channel of the pixel in the reference image, - Generating an alpha mask from the difference image by comparing the difference values ​​of the difference image with a limit value inversely proportional to a speed of the vehicle (1 ), wherein the alpha mask contains an alpha value corresponding to complete transparency for the pixels with at least one difference value above the limit value and an alpha value for the other pixels that depends on the difference value and describes at least partial transparency, - Generating a correction image by applying the alpha mask to the reference image, - Generating a noise-reduced camera image from the correction image and the camera image.

2. Method according to claim 1, characterized in that the reference image is a camera image taken at the preceding time or a noise-reduced camera image determined in a preceding iteration.

3. Method according to claim 1 or 2, characterized in that the difference value is the amount of the difference between the color channel of the pixel in the camera image and the color channel of the pixel in the reference image, or that the difference value is proportional to the amount of the difference between the color channel of the pixel in the camera image and the color channel of the pixel in the reference image.

4. Method according to one of the preceding claims, characterized in that the limit value additionally depends, in particular proportionally, on a noise measure which describes image noise in the camera image.

5. Method according to claim 4, characterized in that the noise level is determined by the vehicle camera (3) during the recording of the camera image and / or that the noise level is determined on the basis of a histogram of previously determined difference images.

6. Method according to one of the preceding claims, characterized in that the alpha value dependent on the difference value for the pixels with the difference value below the limit value corresponds in each case to the amount of the difference in at least one color channel multiplied by a scaling factor, wherein the scaling factor depends in particular on the speed of the vehicle (1 ) and / or on the or a noise measure.

7. Method according to one of the preceding claims, characterized in that the noise-reduced camera image is generated by adding the correction image to a semi-transparent camera image, wherein the semi-transparent The camera image is generated by changing the pixels of the camera image depending on the alpha values ​​of the alpha mask.

8. Method according to one of the preceding claims, characterized in that, to generate the noise-reduced camera image, a superposition of the correction image and the semi-transparent camera image is processed with a spatial filter function.

9. Method according to claim 8, characterized in that a kernel of the spatial filter function is set depending on the determined alpha mask and / or that the spatial filtering is carried out depending on sections in the alpha mask that can be assigned to an edge in the camera image contained in the alpha mask.

10. Method according to one of the preceding claims, characterized in that the vehicle camera (3) captures at least a partial area of ​​the vehicle environment, wherein the noise-reduced camera image and / or a generated on the basis of the noise-reduced camera image The environment is displayed on a display device (10) of the vehicle (1).

11. Control device, configured to carry out a procedure according to any of the preceding claims.

12. Vehicle comprising at least one vehicle camera (3) and a control unit (4) according to claim 11.

13. Computer program comprising instructions which cause a computing device to execute a method according to any one of claims 1 to 10.

Images