Correction of images from a camera in case of rain, incident light and contamination

1 . A method for machine learning of an image correction of input image data, which are negatively influenced by rain, incident light and/or dirt, from a camera, into output image data by utilizing an artificial neural network, wherein the learning is effected with a plurality of pairs of training images in such a way that, in each case, a first image negatively influenced by rain, incident light and/or dirt is provided at an input of the artificial neural network and a second image of the same scene without negative influence is provided as a nominal output image,

wherein the artificial neural network is configured to determine a certainty measure c which is dependent on the degree of wetting by water, incident light and/or contamination for an input image and, following the conclusion of the machine learning, the artificial neural network may establish and output the certainty measure c for a new input image, and

wherein at least one factor d is determined as a measure of a difference between the nominal output image and the negatively influenced input image of a pair of training images and is provided to the artificial neural network.

2 . The method according to claim 1 , wherein the pairs of training images are produced such that, in each case, a first image negatively influenced by rain, incident light and/or dirt and a second image without negative influence are acquired simultaneously or immediately after one another.

3 . The method according to claim 1 , wherein the pairs of training images include at least one sequence of consecutive input and output images.

4 . The method according to claim 1 , wherein the artificial neural network has a common input interface for two separate output interfaces, wherein the common input interface has shared feature representation layers, wherein corrected image data are output at the first output interface, wherein ADAS-relevant detections of at least one ADAS detection function are output at the second output interface and wherein the outputs of both output interfaces are optimized as part of the training.

5 . The method according to claim 1 , wherein the camera is a vehicle-mounted environment-capturing camera.

6 . A method for correcting input image data, which are negatively influenced by rain, incident light and/or dirt, from a camera, said method comprising:

providing input image data captured by the camera, the images being negatively influenced by rain, incident light and/or dirt, to a trained artificial neural network,

converting the input image data negatively influenced by rain, incident light and/or dirt into output image data without negative influence with the trained artificial neural network,

determining a certainty measure c which is dependent on the degree of wetting by water, incident light and/or contamination for an image of the input image data and characterizes the certainty of the network that the image correction of the network is accurate,

outputting the output image data and the determined certainty measure c from the trained artificial neural network, and

additionally providing a factor d to the trained artificial neural network and, wherein the converting the input image data is controlled as a function of the factor d.

7 . The method according to claim 6 , wherein the input image data contain at least one sequence of input images captured one after another as input image data.

8 . An apparatus having at least one data processing unit configured to correct input image data, which are negatively influenced by rain, incident light and/or dirt, from a camera, into output image data, comprising:

an input interface which is configured to receive the input image data, from the camera,

a trained artificial neural network configured to convert the input image data into output image data without negative influence depending on a factor d, which is added to the neural network as an additional input value, and wherein the converting the input image data is controlled as a function of the factor d and to determine a certainty measure c which is dependent on the degree of wetting by water, incident light and/or contamination for an image of the input image data and characterizes the certainty of the network that the image correction of the network is accurate, and

a first output interface configured to output the converted output image data and the determined certainty measure c.

9 . The apparatus according to claim 8 , wherein the data processing unit is implemented in a hardware-based image pre-processing stage.

10 . The apparatus according to claim 8 , wherein the camera is a vehicle-mounted environment-capturing camera and the trained artificial neural network for correcting images is part of an onboard ADAS detection neural network having a shared input interface, and two separate output interfaces, wherein the first output interface is configured to output the corrected output image data and the second output interface is configured to output the ADAS-relevant detections.

11 . The apparatus according to claim 8 , wherein the input image data include at least one sequence of input images captured one after another as input image data, and the artificial neural network has been trained with the aid of at least one sequence of consecutive input and output images.