Correction of images of a camera in the presence of rain, incident light and dirt

Abstract

The present invention relates to a machine learning method, a method, and an apparatus for correcting input image data (Ini) of a camera device (K), such as an in-vehicle ambient environment detection camera, which is affected by rain, incident light, and / or dirt. A method for correcting input image data of a camera device (K) includes the following steps: a) providing input image data (Ini) detected by the camera device (K) that is affected by rain, incident light, and / or dirt to a trained artificial neural network (CNN1, CNN10, CNN11, CNN12); b) configuring the trained artificial neural network (CNN1, CNN10, CNN11, CNN12) to convert the input image data (Ini) affected by rain, incident light, and / or dirt into interference-free output image data (Opti), and determining a deterministic metric c, which depends on the degree to which the image of the input image data is affected by water, incident light, and / or dirt, and characterizes the determinism of the network, i.e., the network's image correction is accurate; and c) configuring the trained artificial neural network (CNN1, CNN10, CNN11, CNN12) to output the output image data (Opti) and the determined deterministic metric c. Advantageously, the method is able to identify objects while the camera device is fogging up and generate image data streams from the network for human and computer vision to perform optimized correspondence search.

Images