Image enhancement system and method based on generative adversarial network (GAN) model

An image enhancement system based on a generative adversarial network (GAN) model, comprising:

• an acquiring unit;
• a training unit;
• and
• an enhancement unit;
• wherein the acquiring unit is configured to acquire a first image of a driving environment captured by a camera of a first vehicle and a second image of the driving environment captured by a camera of a second vehicle;
• the training unit is configured to acquire a training image based on the first image, wherein the training image comprises a first training image and a second training image, and an image quality of the second training image is lower than that of the first training image;
• configured to train a discriminative model in the GAN model by using the first training image and the second training image;
• and configured to train a generative model of the GAN model by using the second training image through a trained discriminative model to obtain an image enhancement model;
• the enhancement unit is configured to enhance the second image by inputting the second image into the image enhancement model, so as to improve an image quality of the second image;
• wherein the training unit comprises a processing module, a discriminative model training module and a generative model training module;
• the processing module is configured to process the first image to obtain the first training image and the second training image;
• the discriminative model training module is configured to train the discriminative model by using the first training image and the second training image;
• the generative model training module is configured to train the generative model by using the second training image based on the trained discriminative model to obtain the image enhancement model;
• wherein the generative model training module comprises an output sub-module, a feature extraction sub-module, a comparison sub-module and a training sub-module;
• the output sub-module is configured to output an output image after the second training image is input into the generative model;
• the feature extraction sub-module is configured to extract a feature from feature areas of a desired image, the second training image and the output image;
• the comparison sub-module is configured to obtain a brightness difference of a specific area between the second training image and the output image, and obtain a texture difference of the specific area between the desired image and the output image;
• and
• the training sub-module is configured to train the generative model according to the brightness difference and the texture difference.